Zipline Expands US Drone Deliveries as FAA Finalizes BVLOS Rules

FAA敲定BVLOS规则之际 Zipline扩大美国无人机配送业务 除非监管机构从指令性认证转向允许硬件快速迭代的基于统计和性能的安全规则，否则美国在全球无人机行业中可能会面临输给中国的风险。 Zipline最初从应对东非的关键医疗物流和血液短缺起步，最终在卢旺达实现了经实证核实的51%的产妇死亡率下降。此后，该公司扩展至美国，利用其新型Platform 2混合垂直起降（VTOL）飞行器应对严格的监管环境，并在得克萨斯州推动了需求导向的郊区家庭送货爆发式增长。通过采用具有成本效益的组件实现航天级冗余并将制造集中在加州，Zipline专注于提供自动化、多模态的物流平台，而不仅是销售无人机硬件。 Zipline在八个国家运营着自动化医疗物流系统，累计飞行里程超过1.1亿英里，完成超过100万次自主飞行。 (0:00) 即时配送的现状效率极低，往往使用重达4,000磅的燃油车配送平均仅重5磅的包裹。 (0:00) 商用无人机配送的长期延迟是由“幻灭低谷期”造成的，事实证明，在现实世界中扩大自主生产和自主导航的规模远比构建早期原型困难得多。 (4:10) Zipline通过与卢旺达集中化的医疗系统合作，绕过了碎片化的美国监管环境，将重点直接转向血液物流以应对产妇死亡问题。 (7:41) 宾夕法尼亚大学的一项学术研究实证表明，Zipline的即时配送基础设施使卢旺达的产妇死亡率降低了51%。 (11:35)

Zipline co-founder and CEO Keller Rinaudo Clifton contends that automated drone delivery succeeds by treating flight as an invisible logistics service rather than selling hardware, eliminating the environmental and financial costs of conventional ground transport. He argues that operational deployment—scaling from national medical supply chains in East Africa to US suburban commercial deliveries—demonstrates that high safety and reliability can be achieved through deep vertical integration and rapid real-world iteration. Reimagining Last-Mile Delivery as an Invisible Service (0:00) Clifton challenges modern parcel logistics, pointing out the inefficiency of dispatching a 4,000-pound internal combustion vehicle to transport an average five-pound package. In his view, this mismatch creates congestion, air pollution, and excessive delivery expense. Clifton also rejects sidewalk robots, arguing that pedestrian walkways should not be surrendered to commercial machines. Instead, Zipline treats aerial robotics as an invisible infrastructure layer—comparable to municipal running water or instant teleportation—where end customers care only about rapid, reliable transit rather than the aircraft itself. Because enterprise and government clients want solutions rather than fleet maintenance burdens, Zipline explicitly avoids selling standalone drone hardware. Clifton views the drone purely as an enabling technology platform, comparing its role to how Amazon utilized the early internet. He maintains that long-term viability requires direct service contracts with end users—such as national ministries of health or retail partners like Walmart—shifting supply chains away from disjointed, donor-driven push programs toward responsive, demand-driven systems. Clinical Impact and Demand Stabilization in Rwanda (7:41) Facing restrictive US commercial drone regulations in 2014, Zipline initiated operations in Rwanda after the country's Minister of Health directed them to focus strictly on emergency blood delivery. Blood logistics present acute operational bottlenecks because blood components have rigid cold-chain requirements and brief usable windows, such as platelets expiring within six days. Despite early operational setbacks where the system initially served just a single facility five miles away, the network eventually expanded to more than 500 public health facilities. A University of Pennsylvania study confirmed a causal 51% reduction in maternal mortality from postpartum hemorrhage, accompanied by a 70% reduction in vaccine waste and a 32 percentage point decrease in zero-dose children. This reliability transformed local administrative behavior across 400 to 500 rural health posts. Under prior quarterly road delivery schedules, local clinics routinely placed inflated orders to offset expected administrative rationing, obscuring true demand signals. With instant aerial fulfillment, health workers stopped defensive hoarding and ordered exact clinical quantities, stabilizing overall supply chain volume while expanding access. Consumer Electronics Economics with Aerospace Redundancy (28:57) Zipline accumulated over 120 million autonomous commercial flight miles with zero reported safety incidents by combining aerospace failover design with consumer electronics components. Clifton notes that traditional aviation architectures rely on expensive, legacy parts; Zipline replicates these safety margins using modern mobile hardware at roughly one-thousandth of the cost. The aircraft feature redundant battery nodes, dual flight computers, and a dual CAN bus wiring layout capable of withstanding in-flight vibrations. When commercial off-the-shelf servos proved prone to failure, Zipline vertically integrated its engineering to build custom, automotive-grade servos rated for a million cycles. Hardware development was driven by iterative field deployment rather than theoretical design. To eliminate a 20-minute GPS satellite reacquisition delay during rapid battery swaps, engineers placed the GPS sensor directly on the battery pack, doubling turnaround speed despite the higher bill-of-materials cost of purchasing multiple sensors per drone. Field maintenance was similarly streamlined by redesigning the airframe to reduce fastener varieties from 43 types down to two. Suburban Integration and Performance-Based Regulatory Pathways (38:15) To serve residential US markets, Zipline introduced Platform 2, moving away from Platform 1's catapult-and-parachute framework to a quiet hybrid vertical-takeoff-and-landing aircraft. The main vehicle hovers at 300 feet to mitigate neighborhood noise, lowering a tethered, motorized droid that dynamically maneuvers along the X and Y axes to deposit packages onto driveways with high precision. Deployed across roughly 10 Walmart Supercenters in the Dallas-Fort Worth metroplex, the regional system handles thousands of daily deliveries and achieved 25% to 30% week-over-week growth, prompting Zipline to pause outbound marketing due to operational capacity constraints. Clifton identifies legacy regulatory structures as a key friction point. He observes that while the 1930s Douglas DC-3 cost roughly $75 million in inflation-adjusted dollars over two years to develop, modern airliners require tens of billions of dollars and decades due to prescriptive, component-level reviews. While Zipline secured Federal Aviation Administration exemptions through its 100-million-mile safety record, Clifton advocates for performance-based frameworks—such as proposed Part 108 beyond-visual-line-of-sight rules—where regulators enforce statistical safety thresholds while allowing rapid technological iteration. While Zipline aims to scale metro-level capacity toward 10,000 daily deliveries and produce up to 50,000 aircraft annually in its South San Francisco plant, Clifton highlights broader macroeconomic and geopolitical vulnerabilities. China manufactures significantly more drones weekly than the United States produces annually, supported by deep consumer electronics supply chains. Although Clifton believes the US maintains an advantage in autonomy, artificial intelligence, and complex systems integration, long-term leadership remains vulnerable to domestic manufacturing atrophy and municipal permitting friction for ground infrastructure.

Zipline 联合创始人兼首席执行官 Keller Rinaudo Clifton 阐述了将无人机作为自动化物流基础设施的核心战略，主张通过在真实环境中快速迭代、结合消费电子级硬件与航空级冗余设计，重构从发展中国家医疗急救到美国郊区即时零售的全球履约网络。 重塑即时配送与解决医疗供应链浪费 (0:00) Clifton 指出，当前即时配送市场存在严重结构性失衡：以美国为例，每年约有 55 亿次即时配送，平均包裹重量仅约 5 磅，却普遍使用重达 4000 磅的人工驾驶燃油车来完成，从而造成交通拥堵、环境污染和高昂成本；同时他反对占用行人空间的地面轮式机器人。对全球约 50 亿至 60 亿缺乏可靠物流保障的人口而言，由于缺乏统一的基础设施，医疗供应链往往极度分散，传统模式被迫在扩大覆盖面与控制浪费之间艰难权衡。 在非洲早期的调研中，Clifton 曾亲眼目睹后院堆满一整年过期药品的援助物资仓库。Zipline 打破了这种折中：通过高频即时配送，系统不再需要依赖季度性囤货与预测，使诊所能够按需下单并消除过度采购。不仅如此，Zipline 坚持商业化可持续路线，直接与东道国卫生部签订服务合同，凭借更低成本与更高时效帮助各国公共卫生系统节约巨额资金，彻底摆脱了非政府组织捐赠模式常伴随的低效与脱节弊端。 卢旺达破局与公共卫生成效实证 (7:41) 2013 至 2014 年创立初期，因美国缺乏大规模自主无人机飞行的监管途径，Zipline 将目光投向拥有单一国家医疗体系且治理高效的卢旺达。尽管初登门时抛出了运送多种物资的宏大蓝图，但时任卫生部长要求其聚焦于急需冷链且保质期极短（如血小板仅存活 6 天）的血液制品。首年运营极度艰难，团队不仅缺乏航空硬件研发和民航对接经验，前 9 个月更只能勉强服务距离 5 英里的 Kabgayi 医院。 随着卢旺达政务团队的高效协同与系统逐步稳定，Zipline 迅速实现规模化，从 1 家医院扩充至首期合同的 21 家，再扩展到如今覆盖该国 500 多家公共医疗机构及 400 至 500 个基层卫生哨所，单一分发超过 500 种健康产品。宾夕法尼亚大学等学术机构的研究证实了该系统的因果效应：在其覆盖区域内，产妇死亡率显著下降了 51%，疫苗浪费减少了 70%，未接种任何疫苗的零剂量儿童比例下降了 32 个百分点。这一验证促成了公司自 2020 年起向加纳、科特迪瓦、尼日利亚和肯尼亚等国的快速扩张。 自研整合与低成本航空级硬件架构 (28:57) 在硬件工程上，Zipline 拒绝直接采购或销售无人机硬件，定位自身为纯粹的物流服务提供商。针对用于广域医疗物流的固定翼无人机 Platform 1，团队借鉴了航母阻拦索设计，研发出依靠 1 厘米尾钩、以时速 100 公里在空中被钢索精准捕获的回收机制，并采用弹射起飞与纸降落伞投递。为跳出传统航天依赖昂贵过时组件的窠臼，Zipline 引入智能手机零部件来实现多重冗余，包括备用电池节点、冗余飞行计算机以及双 CAN 总线架构，即便飞行中剪断任一电线或遭遇剧烈震动也不会坠机，将航天级安全成本降至原本的千分之一。 在追求 99.9% 可靠性的过程中，团队通过一线运营反复试错并实施垂直整合：为消除电池更换导致 GPS 重启耗时 20 分钟的瓶颈，团队将 GPS 直接置于电池模块；为降低野外维修复杂度，将机身 43 种紧固件精简为 2 种；针对市场上廉价玩具电机与千元级航天舵机的极端分化，自主研发出可承受百万次运转的汽车级高可靠舵机。全机 43 个核心子组件几乎全部由南旧金山 16 万平方英尺的工厂自研生产，配合 24/7 满负荷三班倒运作，年产能可达 5 万架。 二代静音平台与美国郊区零售飞跃 (38:15) 为了契合美国等郊区商业场景对无专门起降场、极高落点精度与静音的需求，Zipline 推出了垂直起降与固定翼混合构型的 Platform 2。该机型悬停在居民住宅上方 300 英尺的高空以保持近乎无声，再通过系留缆绳垂降一具具备 X-Y 轴自主动力纠偏的微型机器人，在强风下亦能将包裹精准放至餐盘大小的门前区域。目前该平台正服务于达拉斯-沃思堡地区，并以每周新增一家沃尔玛超级中心的速度推进，单店配备 8 至 20 架飞机，全网单日运力已达数千单并向万单迈进。 在达拉斯上线的三月中，系统保持了 25% 至 30% 的周复合增长，迫使公司在零营销状态下应对涌入的需求。高频用户展现出极高黏性，如有 78 岁的地方官员一年下单多达 340 次，平均配送时长仅约 8 分钟。与加州等因地方许可繁琐而推进迟缓的地区相比，得克萨斯州对地面充电和发射基建的批文流转极为迅速，有力承接了联邦航空管理局从 Part 107 视距内限制迈向 Part 108 超视距（BVLOS）立法的红利，这背后的核心支撑正是 Zipline 累积的 1.2 亿英里商业自主飞行且零安全事故的纪录。 制造业纵深、监管演进与硬科技方法论 (45:52) Clifton 剖析了传统航空监管带来的高昂成本，指出 1930 年代 DC-3 客机研发仅耗时 2 年、花费折合现值约 7500 万美元，而波音 787 耗时 20 余年耗资 320 亿美元，成本膨胀近 500 倍；他呼吁监管机构应从繁琐的部件审查转向基于统计安全门槛的结果导向监管。在软件与安全验证上，Zipline 通过以超实时 10 至 100 倍运行的纯软件测试、覆盖 95% 软硬件集成隐患的硬件在环（HITL）模拟，以及 24 小时外场实测，达成了全球机队每 30 天推送一次完整更新的敏捷节奏。 针对全球供应链竞争，Clifton 认为虽然中国依托苹果公司历史上数十年的供应链投入而在千元级消费级塑料四轴机产能上具有压倒性规模，但在复杂的软硬件协同与高端自主物流系统方面，美国仍保持领先。回顾硅谷过去十余年从纯软件向物理世界硬科技的野心变迁，他强调硬科技初创企业必须避免长期闭门研发，尽早将极度简化、哪怕令人略感简陋的产品推向一线实战，通过真实世界的反馈完成快速进化。 Zipline 证明了结合软硬件纵向一体化与消费电子级成本控制能够在全球范围内落地自主航空物流，但其下一步全国化乃至全球化扩张，仍取决于各地地面基建许可审批效率、联邦适航审定制度的现代化演进，以及抵御宏观经济波动与地缘制造业供应链断裂风险的能力。

[0:00] Keller Clifton: My personal opinion is that we should be giving space back to humans, not taking space away from humans [4:10] Keller Clifton: When people say it's a stupid idea and it's never going to work, that's when the actual hard work begins [7:41] Minister of Health of Rwanda: Keller, shut up. Just do blood. [28:57] Keller Clifton: reach into the aircraft, while it's flying and cut any wire with scissors [33:38] Keller Clifton: None of our customers care about drones [33:38] Keller Clifton: the first logistics system that serves all people equally [38:15] Keller Clifton: dinner plate level accuracy [38:15] Keller Clifton: fundamentally jaw-dropping [54:24] Keller Clifton: science fiction becomes entitlement [57:11] John Collison: Why can we not have this in California? Why are you in Texas and not California? [66:18] Keller Clifton: In a week, China makes more drones than the US makes in a year. [66:18] Keller Clifton: The bad news is like, yeah, it's true. You know, China has like massive manufacturing advantages and massive scale in this area that the US does not currently have. But that is mainly for like, you know, plastic quadcopters that cost $1,000. [70:38] Keller Clifton: we don't have the ability to build our own smartphones period [71:30] Keller Clifton: designing something like incredibly simple and getting it into the real world quickly and learning by doing

[0:00] How many countries today [0:01] do you run your medical logistics network? [0:04] Eight. Why is that not 80? [0:06] You sound like our board, John. [0:07] You're really putting the pressure on. [0:09] And then we walk out the back of the warehouse [0:11] and it's just a field of boxes as far as the eye can see. [0:15] And I was like, "What is this?" [0:17] And they're like, "That's all the medicine [0:19] that expired last year." [0:21] What is Guinness? [0:22] Guinness is a stout. [0:24] Irish people would not call it beer. [0:27] Does anyone know what the difference between stout and beer is? [0:29] Keller Clifton is the cofounder and CEO of Zipline, [0:32] the drone delivery company. [0:33] They started out with medical deliveries in Rwanda, [0:36] but they're now expanding in many countries, [0:37] including the United States, [0:39] and have completed more than a million autonomous flights. [0:41] Cheers. Cool. Cheers. [0:45] Okay. I'm very excited to talk about Zipline. [0:46] So, previously you were a professional rock climber. [0:51] You studied biochemistry and computer science at Harvard. [0:54] You had a previous robotics startup. [0:55] Then in 2014 you started Zipline. [0:58] Most people who do know Zipline probably know you [1:00] for the medical deliveries in Rwanda and broader in Africa. [1:04] You now also have operations. [1:06] You just announced Chipotle [1:07] and burrito deliveries in the United States, [1:09] and you have a delivery takes place every minute. [1:12] And you have 100 million miles flown. [1:14] Is that a good— [1:15] I think like 110 or 120 now. Yeah. [1:16] Okay. Yeah. So, that the good potted Zipline summary? [1:19] What would you add? [1:20] What would I add? [1:21] I mean, you know, when we were starting [1:23] to build Zipline in 2013, 2014, [1:25] our backgrounds were in robotics [1:27] and automation and software. [1:28] And it really seemed to us like it should be possible [1:33] to build an automated logistics system for Earth. [1:36] Like when it comes to robotics, [1:37] you really want boring and repetitive. [1:42] Like that's ideal. [1:43] You know, constrained problems, boring and repetitive. [1:47] And logistics is about as boring and repetitive as it gets, [1:50] but like it's this underlying infrastructure [1:53] that affects everything in the world. [1:55] It affects how people get access to healthcare. [1:57] It affects like how people even choose where they live. [1:59] It affects how much time people get with their families [2:02] versus having to spend time driving around [2:04] and, you know, buying things. [2:06] And so, yeah, it was...the idea was simple. [2:11] The execution has turned out to be [2:12] a bit more complicated than we were expecting. [2:14] Yes. And you've made the point that like [2:15] what does a UPS truck weigh? [2:17] Like, you know, 5,000— 5, 6,000 pounds. [2:19] Yeah, yeah, yeah. [2:20] It's like we've taken for granted [2:22] that we just have these massive vehicles [2:24] trundling around our suburban streets [2:25] as the best answer for delivering things. [2:27] And it's not obvious it couldn't be improved from there. [2:29] I think the best way to think about that is that [2:31] when you're batch processing things [2:33] and delivering things over the course of a day, [2:35] that works pretty well. [2:37] You can load tons of packages into a car [2:38] and then the car, you know, the truck over time [2:40] can deliver things slowly to lots of homes. [2:42] But obviously, the thing we've seen [2:44] over the last five or seven years [2:46] is the rise of instant delivery. [2:48] There are now five and a half billion instant deliveries [2:51] happening every year just in the US. [2:54] And this is where we're using [2:55] a 4,000 pound gas combustion vehicle [2:57] driven by a human sitting in the front [2:59] moving the steering wheel back and forth [3:01] to deliver something to your home [3:02] that weighs on average five pounds. [3:04] Yes. So actually, [3:05] you do not have to be a physicist [3:07] to realize this is a really weird way. [3:09] It's a bit over-engineered. Yeah. [3:10] Or under-engineered, [3:11] depending on how you think about it. [3:13] But like this is, you know, [3:14] it creates traffic in our neighborhoods, [3:16] it creates pollution in our neighborhoods, [3:18] it creates safety issues in our neighborhoods. [3:22] It's phenomenally expensive actually, and it's pretty slow. [3:26] And obviously, competing vision [3:27] is the sidewalk robots that some people are pursuing, [3:29] but I presume you think for many of these applications, [3:31] they're just too slow [3:32] and you actually, can't fit that many of them [3:34] before they start clogging up sidewalks. [3:36] Is that basically, the argument you make? [3:38] My personal opinion is that we should be [3:39] giving space back to humans [3:41] not taking space away from humans. [3:43] So, you know, we have already, I mean, [3:45] I grew up, maybe you grew up, you know, [3:47] we played in the streets like we played street hockey. [3:49] I wouldn't really let my kids [3:50] like go play street hockey today. [3:51] So, something has mentally changed [3:53] where that doesn't seem like a safe a space anymore. [3:56] Yes. [3:56] Sidewalks maybe are like the last place [3:59] that are meant to be for humans. [4:01] The idea that, okay, [4:02] now the robots are going to take those over too [4:04] and like humans are going to have to jump outta the way [4:06] as these sidewalk robots go back and forth, [4:08] doesn't make a lot of sense to me. [4:10] Why has the drone delivery vision taken so long? [4:15] Where, you know, people were talking about [4:16] tacocopter on Hacker News in 2012. [4:19] The big Amazon announcement was what year? [4:22] 2013. 2013. [4:23] And so, everyone has felt like it's just around the corner [4:27] for, well, longer than a decade now. [4:30] Is it, you know, the FAA beyond visual line of sight rules? [4:33] Is it like the tech was surprisingly hard, [4:35] just why is this happening now? [4:37] You guys just announced Chipotle burrito deliveries. [4:41] Why is that happening in 2025 and not 2015? [4:44] I think, it's similar to the question of [4:47] "why do we not have autonomous cars." [4:49] When, you know, certain leaders were saying, [4:51] "Oh, it's right around the corner," in 2015 or 2016. [4:53] I think that with any advanced technology, [4:56] there's always, what do they call it? [4:57] The trough of disillusionment curve. [5:01] It's kind of like you have like the early hype people [5:02] realize something should be possible. [5:04] It's really exciting and then it completely craters, [5:07] Because people realize it's really hard [5:09] and then everybody basically is like, [5:10] "Well, that was a stupid idea. It's never going to work." [5:12] And when people say it's a stupid idea [5:14] it's never going to work, [5:15] that's when the actual hard work begins. [5:17] And then there's like a decade of building [5:19] all of the fundamental underlying infrastructure [5:21] and technology that's required to actually, [5:24] make something work in the real world. [5:28] And, you know, work in the real world [5:30] is kind of like an easy way of putting it, [5:32] but the reality is, whether you're talking about [5:33] autonomous cars or autonomous aircraft, [5:36] you have to figure out not just safety and reliability, [5:39] you have to figure out regulatory, [5:40] you have to figure out manufacturing [5:42] and maintenance and supply chain. [5:44] All of the software systems that are required [5:46] to run behind the hardware to really make it reliable. [5:49] Then you also need a good customer experience [5:51] on top of all that. [5:52] And then ideally, you also have to [5:53] make the unit economics work. [5:54] So, it's the classic thing of [5:56] producing a convincing demo in hardware is relatively easy, [5:59] but then getting something actually, useful operationally [6:02] is where the, you know, the last 10% of the work [6:06] is another 2x or 3x from there. [6:07] I think, I mean, Elon, I remember watching him [6:10] for 10 years say again and again, [6:12] Because people would always say like, [6:13] "Oh, Ford's going to do what you do," [6:16] you know, "GM is just going to do it and like crush you." [6:19] And he just kept saying like, [6:20] "Prototypes are easy, production is hard. [6:22] Prototypes are easy, production is hard." [6:23] Yep. [6:24] I think it's deeply true in hardware. [6:26] It's very easy for people to... [6:28] and we've seen it again and again and again, right? [6:30] Like in specifically, with drone delivery, [6:32] someone will buy a quadcopter off the shelf, [6:34] duct tape a Snickers bar to the bottom of it, [6:36] manually pilot it, [6:37] like when the weather is good, two miles land it. [6:39] It's like, "We just did drone delivery," [6:41] and then get a journalist to cover that. [6:43] But the reality is that [6:46] what customers really want is teleportation. [6:50] And they don't care at all about drones. [6:53] This is kind of a counterintuitive thing about [6:55] that we've learned. [6:56] It's like when we're working with [6:57] all these ministries of health, [6:58] or these governments, or big companies, [7:00] like Walmart or big hospital systems in the US, [7:03] like Cleveland Clinic or many others. [7:05] They all want to be able [7:07] to just teleport things directly to patients [7:09] or directly to a customer's home [7:11] in a way that either saves a life [7:13] or transforms a customer experience. [7:15] You know, we could be using whatever— [7:17] They're not interested in drones, [7:21] they don't care about drones, [7:21] they don't want to deal with any of that complexity. [7:24] All they want to know is something [7:24] goes from point A to point B fast enough [7:26] to solve a real problem. [7:28] And so, I think that a lot of the technology [7:30] kind of sits behind a curtain [7:32] in order to enable a customer experience [7:34] that is just jaw dropping [7:36] and you know, changes the way people live their lives. [7:38] So, demos are easy, production is hard. [7:41] You got started in production [7:43] and really ramped up in Rwanda. [7:46] How did that come about where, you know— [7:48] people, I can understand maybe [7:49] you might want to go outside the US [7:51] for a more favorable regulatory environment, [7:52] but most people will go to Mexico, or Canada. [7:54] How did you get to Rwanda? [7:57] And maybe you can tell the story [7:58] of scaling up operations there. [8:00] Yeah, hard to frame. [8:03] I mean, I think that when we were [8:05] starting to build Zipline in 2013, [8:07] we spent about two years [8:08] building the initial version of the technology, [8:10] and everybody told us this was a stupid idea. [8:13] Everybody told us there was no way it was going to work. [8:15] Was it hard to get funding for? [8:16] Oh, incredibly difficult, [8:18] like I remember we were begging people to invest, [8:21] you know, $250,000 in the company. [8:24] What was the first big break in terms of who invested? [8:27] Sequoia led our series A and took this bet on us. [8:31] This was actually, [8:31] before we started building Zipline though. [8:34] Oh, wow. Yeah. [8:35] This is for your previous startup? [8:36] Yeah, well, same company actually. [8:38] Oh. But it pivoted. I didn't know that. [8:40] Yeah, most people don't. [8:42] And then, as we were starting to build Zipline, [8:44] I just think there was so much skepticism. [8:46] I mean it's like such a crazy idea. [8:48] Like, what do you know it about aviation? [8:50] Nothing. Right. [8:50] What do you know about healthcare? [8:52] Nothing. Yeah. [8:52] What do you know about operating in Africa? [8:53] Nothing. Okay. [8:54] Well, like, let's stay in touch. [8:56] Hard tech operating in Africa. Yeah. [8:58] Let's stay in touch, you know? [8:59] Smile and nod. [9:00] Yeah, exactly. Yeah. [9:02] I think that, our intuition was [9:05] we needed a country that was as desperate [9:07] as we were desperate as a startup. [9:09] Maybe that sounds bad. [9:10] Which is like threshold question. [9:11] Why didn't you start operating in the United States? [9:13] It just wasn't a regulatory path... [9:14] regulatory pathway to do so. [9:15] Okay. Just the ban on drones. Yeah. [9:17] Totally impossible to do it in the US. [9:19] And also, we really wanted to find, you know... [9:22] we had this instinct that [9:24] to convince a country to do this at scale, [9:26] we needed to focus on lifesaving use cases. [9:28] And so, we really wanted to find [9:29] a national healthcare system. [9:31] Rather than a fragmented private one [9:33] like you have in the US. [9:35] So, we were looking for a national healthcare system, [9:38] healthcare challenges that we felt [9:40] we could really clearly solve. [9:42] And a country that would be entrepreneurial [9:44] and fast moving alongside us. [9:46] And so, that to us, you know, [9:48] had us looking at countries in East Africa. [9:50] We're also talking to countries in Latin America. [9:53] And interestingly, I remember this meeting [9:55] with the minister of health of Rwanda, [9:58] I mean we were beyond clueless. [10:00] Like it's just like, [10:03] 24-year-old showing up in East Africa, [10:05] like, "We build robots, [10:06] we think that this could work." [10:07] And I remember this meeting with her where I was saying, [10:10] "Hey, we'll deliver, blood and vaccines [10:14] and cancer products and all these different things. [10:15] We could do it all instantly [10:16] in a way that would be transformative [10:18] and reduce waste and save lives." [10:19] And I remember she looked at me and just said, [10:20] "Keller, shut up. Just do blood." [10:23] And as she was explaining to me, you know, [10:25] 50% of blood transfusions are going toward [10:28] moms with postpartum hemorrhaging, [10:29] 30% are going toward kids under the age of five. [10:32] And it is a super hard problem from a logistics perspective. [10:35] Because you have packed red blood cells, platelets, [10:38] plasma, crop precipitates, all these different types [10:41] of each AB, A, B, and O. [10:43] They have all different storage requirements [10:45] and shelf lives. [10:45] Platelets only last six days, for example. [10:47] When you get them out of a donor into a patient. [10:51] So, they gave us 21 hospitals and they said just do blood. [10:55] And that's what we did for the entire first year. [10:57] And it was a disaster by the way. [10:59] Why? [10:59] Because we had no idea what we were doing. [11:02] You know, I think apropos of the flag behind us, [11:05] we didn't know how to integrate [11:06] with a national healthcare system. [11:08] We didn't know how to integrate [11:10] with a civil aviation authority. [11:12] And we definitely didn't know how to build [11:13] an aerospace-grade product that was reliable yet. [11:16] So, we launched and we were pulling constant all-nighters. [11:20] We were flying back and forth between the US and Africa. [11:22] You know, we had the entire Rwanda team hired [11:25] and the system only served one hospital [11:28] for the first nine months reliably. [11:30] And we were just trying everything we could do [11:32] to make the system work reliably for that one hospital. [11:35] Why Rwanda out of all the Sub-Saharan African countries [11:38] you could have picked? [11:41] We can answer that question now, [11:43] because we have a much better understanding. [11:44] The reality is then we weren't smart enough. [11:46] We just got lucky. [11:47] The answer now is, [11:48] Rwanda really is kind of like the Singapore of Africa. [11:51] It has like a very technocratic leader [11:53] with a very technocratic leadership team [11:56] where they're taking really smart bets [11:58] on technology and moving quickly. [11:59] And it's a very, very entrepreneurial country. [12:01] It's got a much stronger civil service. Yeah. [12:03] Super strong civil service. [12:04] Like I can't even describe to you, [12:06] I mean many times I would be on calls with a minister [12:10] or a chief of staff at like nine o'clock on a Saturday, [12:13] 11:00 pm on a Saturday. [12:16] I mean moving lightning fast. [12:17] That's kind of what I mean. [12:18] You know, there are many countries [12:20] that are moving very fast, [12:21] because they know that the only way they're going to [12:22] get ahead [12:24] is by being aggressive and entrepreneurial and making bets. [12:28] So, it was messy at first. [12:30] When did you crack operations in Rwanda [12:32] and you felt like, wow, this is really working. [12:34] So, after nine months of all-nighters, [12:36] total desperation, everything that could break did break, [12:40] it started working reliably for that one hospital, [12:42] which was called Kabgayi Hospital. [12:45] By the way, that hospital was only like five miles away. [12:47] So, it was like we were doing [12:49] the easiest possible thing and failing at it. [12:51] It didn't feel great for nine months. [12:52] But like when did Zipline get to the point [12:54] where Rwanda would definitely not let you rip it out, [12:56] because they're like, oh, we rely on this, you know? [12:58] So, after nine months it started [12:59] working well for that hospital [13:00] and then we rapidly added the other 20. [13:02] So, we were actually, serving all 21 hospitals [13:04] in the contract by the end of the first year. [13:06] And then they asked us to expand to 50 hospitals, [13:09] and then 200, and then 500. [13:11] Then we went from delivering blood to vaccines, [13:14] cancer products, infusions, transfusions. [13:16] Today, we're like the sole distributor [13:18] of 500 plus health products [13:20] in the public healthcare supply chain. [13:22] And then we started expanding outside Rwanda. [13:24] Today, we serve 5,000 hospitals and health facilities. [13:27] It's become, yeah, [13:28] like the largest commercial autonomous system on Earth. [13:31] If this model makes total sense, [13:33] which I think what you're saying makes sense. [13:35] There are various healthcare products [13:37] where you may need them anywhere, [13:39] you can't predict where you're going to need them [13:41] and you need to get them in a time-critical fashion. [13:44] Why haven't other countries followed... [13:46] I mean you're starting to do this in other countries, [13:48] but why as of 2020 was every country [13:51] not beating down a path to your door saying, [13:52] "Oh yeah, we'll take 10 Zipline systems, [13:55] you know, for our health service." [13:56] It's kind of interesting given that [13:58] all countries healthcare needs are kind of the same, [14:00] as a lot of national healthcare systems. [14:03] You would think more people [14:03] would want to replicate the system. [14:05] Yeah, I mean, I would say they kind of were, [14:07] because it starting in '22 so it took us like, you know, [14:10] three years to stabilize and grow it in Rwanda. [14:12] And then it was also very helpful. [14:14] We had a number of big academic institutions [14:17] come in and want to do studies. [14:18] So, like University of Pennsylvania came in [14:20] and did this big study across all the hospitals we serve [14:23] and showed a 51% reduction in maternal mortality. [14:26] This is like, you know— [14:27] and that's pretty causal. A pretty profound thing. [14:29] Very causal. Yeah. [14:30] I mean they, they have their control group [14:32] and their non-control group. [14:33] There are like interesting different reasons for it. [14:36] A lot of it is around blood availability. [14:38] But also part of it is just [14:40] increased trust in the health system. [14:41] More moms give birth in hospitals [14:43] when they know that the hospital's going to have [14:44] access to the products they need. [14:45] By the way, you know, when we were [14:47] starting to build Zipline, [14:48] we were told this was such a stupid idea. [14:49] But if we had known we were going to get [14:51] a 5% reduction in maternal mortality, [14:54] we would've told you, "Hell yes." [14:56] Yeah, yeah, yeah. [14:57] This is like completely worth doing. [14:58] Victory condition. Yeah. [14:58] Victory condition. Yeah. [14:59] So, the idea of a 51% reduction is, [15:02] you know, is something... [15:05] don't listen to the experts [15:06] I guess is like a good takeaway. [15:08] And so, you know, from there, [15:10] We also got like these additional studies showing, [15:13] you know, 70% reduction in vaccine waste, [15:15] 32 percentage point decrease in zero dose children. [15:18] So, these are kids who never get vaccinated. [15:21] And so, we actually did see, you know, [15:23] Zipline then launched in Nigeria, [15:25] and Cote d'Ivoire, and Ghana, and Kenya. [15:27] We started launching a lot of different countries... [15:30] Around 2020, the system started to scale very fast. [15:33] Have you seen any induced demand [15:35] when healthcare systems have Zipline available to them? [15:42] So, you know, you can imagine [15:43] we can fulfill existing needs better, [15:45] we need blood for a surgical procedure and we don't have it, [15:48] and you know, now with Zipline we have it. [15:50] But have you seen healthcare systems start to do things [15:52] that they weren't previously doing, [15:53] because they have this cool logistics system [15:55] available to them? [15:56] For sure. I mean, that's like the main thing [15:58] that we are observing happening. [15:59] I think there's a lot going on that's all pretty interesting [16:04] and maybe we didn't anticipate all of it. [16:05] I would say one big thing to realize is like [16:08] with a health system you're typically balancing [16:10] access against waste. [16:11] So, if you want to make access really high, [16:14] you send a ton of medicine out to the last mile [16:16] and then access gets better. [16:17] But now you start throwing tons of stuff out, [16:19] because all this stuff has shelf lives [16:20] and a lot of it is like very expensive [16:22] and cold chain-dependent. [16:23] If you want to reduce waste, [16:25] you keep more stuff centralized, [16:26] but now you have more people dying [16:27] when they can't get access to the thing that they need. [16:29] So, one of the really crazy things with Zipline was, [16:31] you know, we were able to show [16:33] that by keeping things centralized [16:35] and turning the entire logistics system [16:38] over to a just-in-time system, [16:40] you can actually, move both up simultaneously. [16:43] That's a big idea. Yeah. [16:45] Another idea is what you're saying, [16:46] which is like the induced demand, [16:47] you know, you make it possible to vaccinate kids [16:51] for example at a different layer of the healthcare system. [16:55] In Rwanda, we now serve I think it's like [16:58] 400 or 500 health posts in the country. [17:00] And so, these are like [17:01] the layer below primary care facilities. [17:03] It means that kids can get vaccinated [17:05] way closer to where they live. [17:06] More kids get vaccinated, because it's just more convenient. [17:09] The last interesting thing that we see is, [17:12] and this is kind of on the opposite side of induced demand, [17:14] is actually, when you had a logistics system [17:16] that wasn't very reliable, you actually, [17:18] had a lot of systems [17:20] that started to exhibit hoarding behavior. [17:22] Which is kind of weird. [17:23] But basically, what would happen is that [17:26] there was never enough blood to go around [17:28] and these health systems would have [17:29] to place their order for blood every quarter. [17:31] So, they would over-order. They over-order, [17:33] because they know they're going to be cut back. [17:35] And so, they would order twice what they needed [17:38] on the assumption that the government's going to send them [17:40] half of what they actually asked for. [17:41] And then that's just like a brinksmanship kind of thing. [17:44] The numbers get bigger and bigger [17:45] and you lose like all signal in the system. [17:47] And so, the really cool thing that we've seen is actually, [17:49] a lot of the demands have gone down, [17:52] because they're actually confident [17:53] that they're going to get exactly what they're asking for. [17:55] Yeah. It's the just-in-time framework, [17:57] but for a healthcare system. Yep. [18:00] What is the business of what you're doing in Africa? [18:02] Are you paid by the healthcare system? [18:04] Is this NGO funded? [18:06] How does the business side of things work? [18:08] 10 years of, [18:09] building the infrastructure in Africa, [18:11] we always felt really strongly that, [18:13] we wanted this to be sustainable [18:14] and we want it to be led by the countries. [18:16] There's a lot of work that gets done in Africa by NGOs, [18:20] they're funded by the US [18:21] or by the international organizations. [18:22] I think what's kind of most amazing about Zipline [18:25] is it's been entirely country led [18:27] and almost entirely country funded. [18:30] So, we sign contracts directly [18:32] with the ministries of health, the governments themselves, [18:34] and then we help those governments [18:36] save lives and save money. Yep. [18:38] So typically we can, for any given ministry of health [18:41] or health system, we can save millions of dollars. [18:44] Plus, you can show extraordinary healthcare impact, [18:48] and that makes it a pretty easy decision for any country. [18:51] Yeah. It's one of these amazing, [18:52] better, faster, cheaper tech use cases [18:56] or tech value props where you go to a health ministry [18:59] and you say, "We'll give you a better logistics system [19:02] that delivers stuff faster and is cheaper [19:04] than how you're currently moving things around." [19:05] And I presume that's kind of hard to pass up. [19:07] Running a healthcare system is expensive. [19:08] Yeah. Running a healthcare— [19:09] yeah, exactly. [19:11] These products are super expensive. [19:13] You know, whether it's vaccine or cancer products, [19:16] they're all cold chain-dependent, [19:17] short shelf life, very expensive, [19:19] and there is phenomenal waste. [19:21] It's mind blowing. Yep. [19:22] Like I remember, and one of the things [19:24] that actually kind of inspired us, [19:25] on one of my first visits ever, [19:28] and I won't name the country, it wasn't Rwanda. [19:30] We went to this USAID funded [19:33] medical product distribution warehouse. [19:37] And we were kind of going inside and you can tell it's like, [19:39] whoa, these systems are not good. [19:40] By the way, like all, you know, donor funded [19:42] and there's a separation from the customer [19:45] I think is the way to think about it. [19:46] But we were in the warehouse, [19:47] it's kind of a disaster. [19:49] It doesn't look like things are going well [19:50] and then we walk out the back of the warehouse [19:52] and it's just a field of boxes as far as the eye can see. [19:56] And by the way, it all says, you know, [19:59] from the American people on it. [20:01] And I was like, "What is this?" [20:04] And they were like, "Oh, don't worry about that. [20:05] Don't worry about that." [20:06] I was like, "No, really, like what the heck?" [20:07] Like it was like vast scale. [20:10] Yeah. And they're like, [20:11] that's all the medicine that expired last year. [20:13] And so, I think we had this kind of like aha moment. [20:16] And the crazy thing is that was happening in a country [20:18] that was having massive shortages [20:19] and healthcare outcomes [20:21] where people couldn't get access to these products. [20:23] You could kind of see the two points [20:26] to just connect with a line. [20:28] So, what's the lesson, the generalized lesson here, [20:30] is it that existing aid supply chains [20:34] are pretty inefficient right now [20:35] and there's a lot of improvement possible? [20:37] I think maybe one big generalized takeaway [20:40] is that logistics is so boring [20:42] and so in the background we never think about it. [20:45] You know, if you're living in a, [20:46] I don't know, you know, the Bay area, [20:48] you're ordering stuff on Amazon, [20:49] you have it delivered to you 30 minutes later [20:51] or an hour later, whatever it is. [20:53] It's kind of like running water. [20:55] You don't really appreciate it until you don't have it. [20:57] Yes. [20:58] And there are parts of the world [20:59] that have crappy access to logistics. [21:01] Yeah. [21:02] And then there is about [21:03] five or six billion people on Earth [21:05] who have no access to logistics. [21:07] And I think that is maybe the big realization [21:10] is that like, maybe we think of it [21:12] as something that's just ubiquitous [21:13] or like something you don't worry about, [21:15] but the reality is that, [21:16] if you don't have like that strong layer of logistics, [21:19] everything else is phenomenally difficult [21:21] and expensive to do. [21:22] So, you mention the public healthcare institutions, [21:26] they're all building verticalized supply chains... [21:29] in these country, and it's many different supply chains. [21:32] It'll be a different supply chain for Aids drugs, [21:35] different supply chain for, maternal mortality, [21:38] different supply chain for kids, [21:41] or you know, neglected tropical diseases, [21:43] and those systems don't run very efficiently. [21:45] In how many countries today [21:46] do you run your medical logistics network? [21:49] Eight. Why is that not 80? [21:51] Again, it feels like something [21:52] every healthcare system should want, [21:55] and it's pretty proven at this stage [21:56] where it's been working well in Rwanda [21:58] and other places for several years. [21:59] You sound like our board, John. [22:01] You're really putting the pressure on. [22:02] You're like, what are our revenue growth numbers [22:05] and why are we not growing faster? [22:06] Exactly. [22:08] You know, I think that there are a few things. [22:10] I mean one is that we've always been pretty clear [22:13] that it's one thing to save the lives [22:16] of the people you're delivering to, [22:18] but it's really important that we be safe [22:20] for the people who we are flying over. [22:22] You know, Zipline now has about [22:23] 120 million commercial autonomous miles [22:25] and zero safety incidents, and that's definitely hard to do. [22:29] That has come from, [22:30] I think the company being very focused [22:33] and step by step. Doing things [22:35] in the right way, testing, validating, [22:37] and making sure never to get over our skis. [22:39] But the laws of aerodynamics are not different in Rwanda [22:41] versus Algeria versus the UK. [22:44] So, that's part of it, [22:46] but then also it's kind of like a, you know, I don't know, [22:48] sometimes we talk about it internally [22:49] as like a pig traveling through a snake. [22:51] It's kind of all about bottlenecks, [22:53] scaling up hardware. [22:54] I mean we were really only ready [22:55] for a lot of scale I would say, [22:58] 2020 was when we started expanding into new countries. [23:00] We then went from one to eight countries [23:03] Went from like 100 health facilities to 5,000 today. [23:08] So, the system has been scaling very, very fast. [23:11] The other thing that did happen is that 2022, [23:15] when you see interest rates spike in the US it's like, [23:17] oh man, that's really inconvenient [23:19] and annoying to us in the US. [23:20] In Africa, it creates economic Armageddon. [23:23] And it feels like if you're running a medical system, [23:25] what you're describing about having, [23:28] like... I view you as not being in the drone business, [23:30] but in this line of business [23:32] you are in the medical inventory management business [23:36] and you just have a much better [23:37] inventory management solution, [23:39] because you're able to do this just-in-time stuff, [23:40] and there are a few medical systems that don't want that. [23:42] Well, I just think it's becoming more and more clear [23:44] that like this is the infrastructure of the future. [23:46] This is a smart thing for countries to be investing in. [23:49] And so, there is a lot that's going on [23:52] even just this year that is leading to [23:54] massive growth both in terms of within countries, [23:57] us adding new distribution centers, [23:58] and us adding new medical products, [23:59] and then also adding new countries to the list. [24:06] This medical logistics system we're describing [24:08] is based on your Platform 1 aircraft. [24:10] And the way the system works, as I understand it, [24:12] is you have a fixed-wing, battery-powered drone [24:16] that is launched off a catapult. [24:18] You shoomph it off the catapult [24:19] and then it flies from a distribution center, [24:22] flies to the hospital, [24:23] it drops the payload with a parachute attached [24:26] and they go grab it. And then the drone returns home [24:29] and I presume just lands in the grass [24:30] and you take it and plug it back in. [24:33] That's roughly how Platform 1 works, right? [24:34] Yeah. It doesn't land in the grass. [24:36] I mean we actually use something [24:37] that is kind of inspired by aircraft carriers. [24:39] Oh, you catch it? [24:40] Yeah, we actually, catch it out of the air. [24:41] Okay. It looks kind of crazy. [24:42] Or it's like the thing in Batman, you know, [24:43] where they're recovering the— [24:45] What thing from Batman? [24:46] You mean the skyhook from Mission Impossible? [24:48] Yeah. There's also a skyhook in Batman. [24:49] Sometimes people think skyhook. [24:50] You know, it basically like the vehicle is flying by [24:53] at about 100 kilometers an hour [24:54] and we can snatch it out of the air. [24:56] Oh, I've seen that. Yeah. Called the snap catch. [24:58] Big pole. Yeah, yeah. Big poles. [24:59] And yeah, you know? [25:01] It's like something out of a cartoon. [25:02] You know, every time you see it. [25:03] Yeah. It really does look [25:03] kind of cartoonish. [25:04] Cartoon energy, yeah. [25:05] It does look weird. [25:06] And every time you see it you're kind of like, hallelujah. [25:08] You know, like it's amazing that, that worked, [25:10] but now it works 1,000 times. [25:11] Like we got away with another one. [25:12] Yeah. We got away with another one. [25:14] But the cool thing about, you know, [25:15] autonomous systems is that like, you know, [25:16] we're using guidance navigation control algorithms [25:20] that are hyper precise. [25:22] When you take the human out of it, [25:23] you can actually achieve levels... [25:24] Not that we want to take human pilots out, [25:27] I realize I was sitting with a talented pilot. [25:30] Thank you. None taken. [25:32] But robots are really good at flying vehicles. [25:34] And so actually, you can achieve levels of precision [25:37] that seem very counterintuitive. [25:39] The aircraft has a one centimeter [25:41] tail hook on the back of it [25:42] and we reliably catch that one centimeter tail hook [25:45] every single time as the vehicle's coming by. [25:47] When you say robots are better pilots, [25:50] I take no offense at that. [25:52] Have you read "The Right Stuff" by Tom Wolfe? [25:53] I haven't actually. I really enjoyed it. [25:55] Tom Wolfe fabulous writer all his stuff I would recommend [25:58] and certainly, "The Right Stuff" [25:59] maybe for listeners of this podcast [26:01] is on the history of the Mercury space program. [26:07] So, before the Apollo program. The initial space program. [26:10] But they describe how when they were making the spacecraft, [26:14] their initial plan that was [26:16] totally obvious to them was they would make them autonomous. [26:19] And this is back in the sixties, [26:21] but it was still kind of patently obvious to them [26:23] that they would make them autonomous [26:24] and they faced rebellion [26:26] from all the shit hot Air Force fighter jocks. [26:30] We can do it better. From the program. [26:32] Well, they just didn't want to be passive passengers. [26:34] You know, we sent monkeys up to space [26:36] and monkeys don't do better. [26:37] We're no better than monkeys [26:38] just like sitting there. [26:39] Sit there and hope you survive. [26:41] Exactly. So, they added [26:42] manual controls for the spacecraft, [26:44] afterwards to the design. [26:46] Because of the pilot— [26:48] The controls weren't connected to anything, [26:50] but they could at least like, [26:51] it's like in an elevator, right? [26:52] You can like push the close button. [26:53] Exactly, but it's very funny. [26:55] And so anyway, that was back in the sixties [26:58] and so, yeah, totally. [26:59] But on a slightly more serious, note on that, [27:02] which is that like today the majority, [27:04] I think you probably well know, the majority of aircraft, [27:07] airplane crashes are caused directly by humans. [27:10] Yeah. [27:11] It's actually, a bit sobering. [27:13] I mean, but also I think it's just a sign [27:15] that like the safety of these systems [27:16] has progressed so dramatically [27:19] that at this point the last unsafe thing in the cockpit [27:21] is the human themselves. [27:23] So, some questions on tech and safety [27:24] and how you do things. [27:25] How do you avoid birds? [27:27] We don't, birds avoid us. Okay. [27:29] Yeah. Turns out like birds— [27:30] Not a big practical issue. [27:31] Birds have pretty good [27:32] detect and avoid systems of their own. [27:34] And so, they hear us coming. [27:36] How fast are you flying? [27:37] About 100 kilometers an hour. [27:38] You're much smaller than airplanes. [27:40] We're not like sucking birds into they cells [27:43] in the same way that a jet would. [27:44] How do you avoid airplanes? [27:45] So, we avoid airplanes by basically, [27:47] it's like a multi-layer deconfliction strategy. [27:50] First of all, we're flying [27:51] below the commercial aviation floor. [27:53] So, you know, in the US that's about 400, 500 feet. [27:57] So, we're never conflicting with [27:59] the kinds of normal airplanes that, [28:02] people who are listening would ever be flying in. [28:04] We are also then on ADSB, so we're receiving ADSB. [28:08] So, as long as something is squawking on ADSB, [28:10] we know it's there and can avoid it. [28:12] Yep. We also then have [28:13] cameras on board and then we also do things like notam so, [28:17] you know, old term, but notice to airmen. [28:18] Basically telling people [28:19] when we're operating in certain places. [28:21] So, like all of these layers together [28:23] combine to get us a safety profile [28:26] that is super obvious that the FAA can approve. [28:29] Hmm. Super cool. [28:30] And then, can you tell, if a battery's caught on fire? [28:35] So, the cool thing about the way [28:36] the aircraft is designed is that like, you know, [28:39] there are 43 major sub-assemblies on the aircraft. [28:41] Every one of those sub-assemblies [28:43] is designed from scratch by Zipline. [28:44] So, super, you know, vertical integration. [28:47] Yes. [28:48] And we really take inspiration [28:50] from the way that planes are designed, [28:52] like we'll look at the way of flight compute on a, you know, [28:54] Boeing 787 is designed [28:56] and they're, of course, using like [28:57] multimillion dollar like 20-year-old components. [28:59] We can talk about why that's so screwed up [29:01] in a second, if you want, [29:02] but like the architecture is usually fundamentally sound. [29:04] And then what we'll usually do is [29:06] just use cell phone components to build that architecture [29:09] in a way that's about one thousandth of the cost, [29:11] but just as safe. Yeah. [29:13] And so, you're kind of using [29:14] a lot of those aerospace paradigms of [29:17] being able to fail over [29:18] and building redundancy into every flight critical system. [29:21] So, you know, the aircraft we have [29:22] redundant nodes of the battery. [29:24] One node can fail the vehicle can still fly itself home. [29:26] We have redundant flight compute architecture. [29:28] So, you can have a flight compute fail, [29:30] you can fail over the second flight computer and fly home. [29:33] You can actually, reach into the aircraft, [29:35] while it's flying and cut any wire with scissors, [29:39] the plane will still fly itself home. [29:40] Because it's wired, redundantly. [29:42] Redundant CAN bus. [29:43] So, basically we're passing, yeah, [29:45] redundant CAN bus on the aircraft means that we're not, [29:48] you know, you can lose any wire. [29:51] We also have redundant control services. [29:52] Did you do that because you're paranoid about mid-air things [29:53] like mid-air collisions [29:55] or because... just as a general redundancy strategy? [29:59] General redundancy, I mean like you're not actually, [30:01] worried about someone reaching into the airplane [30:02] with scissors usually. [30:04] That would be a hard thing to pull off, [30:06] but what you are worried about is like [30:07] vibration and connectors shaking loose. [30:09] That's like a core failure mode for aircraft. [30:12] So, going back to the operations in [30:14] like we're talking about these Platform 1 operations [30:16] at that you're doing in Africa [30:17] and the medical logistics use case. [30:19] What we described sounds pretty simple. [30:23] but as you say, demos are easy, [30:24] engineering or production is hard. [30:27] What was hard about getting, you know, [30:30] from 90% reliability to 99.9% reliability. [30:34] Concretely what was, [30:36] what kind of stuff were you solving? [30:39] I think that with hardware, [30:41] like everything that can go wrong does go wrong. [30:43] You know, first you have to get [30:46] a design that works and makes sense— [30:49] Because you have earlier design iterations that [30:51] were wrong in subtle ways. For sure. [30:52] Unbelievable, or not subtle ways, like in retrospect, [30:55] wow, we're like complete idiots. [30:56] Forward the silly early design. [30:58] Yeah. One really good example with Platform 1 was, [31:00] you know, when the aircraft would land [31:02] we would swap the battery. [31:04] The advantage of swapping the battery [31:06] was we could then have the aircraft back on the launcher [31:08] and back in the air three minutes later. [31:10] We put the GPS on the flight computer, [31:12] which would like seem to make sense as a place for GPS. [31:14] But it has to reacquire GPS every single time. [31:15] So, see you're smarter than we were basically. [31:17] So, the problem was that you have... [31:20] you get an order and it's like wow there's an emergency, [31:23] like a patient needs this, blood transfusion, [31:25] and so, every second counts. [31:28] So, we immediately grab the aircraft, we put the battery in- [31:30] Give me a minute. Plugs in the GPS turns on [31:33] and now the GPS is going to take 20 minutes [31:34] to acquire all the satellites necessary [31:36] for the vehicle to pass pre-flight and launch. [31:38] And so, we realized how stupid this was [31:40] and immediately iterated [31:41] and put the GPS on the battery itself. [31:43] So, it didn't make sense from a bomb cost perspective, [31:46] because now you need two GPS units [31:48] for every aircraft that's flying, [31:49] because one is sitting on the battery charger. [31:51] But it really makes up for it, [31:53] because you've just [31:54] doubled your cycle time on the aircraft. [31:56] So, there are a million examples. [31:57] I mean another example, this stuff is so unfancy, [31:59] it's almost like embarrassing to describe, [32:01] but like when we initially launched Platform 1, [32:04] the aircraft had 43 different kinds of fasteners on it. [32:07] Turns out that is extraordinarily difficult [32:09] to keep 43 fasteners in stock [32:11] at all these different distribution centers [32:13] where you're doing constant maintenance actions on them. [32:15] So, the next version of the aircraft [32:16] had two total kinds of fasteners on the entire aircraft. [32:20] So, a lot of this is like, [32:21] it's not fancy. Yeah. [32:23] But it's hard to learn without [32:25] you just kind of learn by doing. [32:27] Is there a component that's been really annoying? [32:29] Like if you could wave a magic wand [32:30] and get a better version of some component, [32:33] what would you ask for a better version of? [32:35] I mean every component we have iterated [32:37] on many, many times I would say like [32:39] today probably more than anything is like servos. [32:41] In fact Zipline went [32:42] and designed its own servo from scratch. [32:45] You know, we also ended up going and designing, you know, [32:49] I mean almost every component of the aircraft. [32:51] It was just kind of a matter of like [32:52] when it made sense to do it. [32:54] But over time almost every component of the aircraft, [32:57] every major subassembly, [32:57] has been designed completely from scratch. [32:59] Why do servos frustrate you? [33:00] Servos for context are actuators basically, [33:03] things that turn electrical power into mechanical movement. [33:05] I think servos are hard [33:07] just because like there was no industry... [33:09] I think, the way to think about, you know, [33:10] people think about drones and they're like, [33:12] oh yeah, it's like a drone, [33:14] but the reality is you have, you know, an a thousand dollar [33:15] quadcopter [33:18] that will take pictures and fly for 20 minutes [33:21] and the reliability and safety of it [33:23] really doesn't matter at all. [33:24] And then you have a $50 million Predator aircraft, [33:28] which is actually, surprisingly also very unreliable. [33:31] Like Predator, you know, famously I read this statistic, [33:33] which is like 50% of US military drones [33:38] have crashed due to operator error. [33:40] Oh, my God. Read these stories. [33:41] They're like, they didn't realize [33:42] they were flying upside down. [33:44] I mean it is like crazier than you think. [33:45] The DJI inbuilt collision avoidance is now much better. [33:48] Not working. I don't know. [33:49] The new ones will. [33:50] Yeah. I think that the reality is [33:52] those are two different sides of the spectrum. [33:53] And what Zipline really needed was like [33:56] we think a lot about automotive grade, [33:58] which is actually, almost think of it like [34:01] a sacred area of like cost and quality, [34:05] because automotive grade is super. [34:08] It's generally, very low cost [34:10] and very, very high quality and reliability, [34:12] and that's really where servos are. [34:13] Like you can get servos that are totally crappy [34:16] that you'd like use in consumer, [34:17] I don't know like toys. Yeah. [34:19] And then you can get like extremely expensive servos [34:22] designed in Germany and they cost $1,000 each [34:25] and they're way over-engineered. [34:27] But what we needed was something [34:28] that was like extremely cost effective [34:30] and that would last for like a million cycles, [34:32] and that doesn't exist in the world. [34:33] And so, you're now producing your own servos? [34:35] Yeah. And so, we designed it completely from scratch [34:36] and now that's what we use in Platform 2. [34:38] That's wild. [34:39] Will you end up just selling drones for other applica... [34:43] like right now you are a, you know, [34:45] did you know Boeing and United used to be the same company? [34:47] Like it was full vertically integrated aviation where— [34:49] I did not know that. [34:50] Yeah. Wow. [34:51] I feel like, I'm embarrassed. [34:53] The airframer and the airline. Yeah. [34:54] And then they were split up in that like it kind of [34:55] there was a big antitrust, [34:57] antitrust was emerging [34:58] as a domain kind of at the time. [35:00] And as part of that they were forced to split it up. [35:02] Anyway, you are the Boeing and United, right? [35:04] You know, you provide the service [35:05] and you produce the aircraft. [35:07] Might you ever just sell the aircraft? [35:08] I don't think so. [35:10] For a number of reasons. [35:11] I mean I think the most important one [35:12] is kind of what I was alluding to before, [35:14] which is none of our customers care about drones. [35:16] Like if we were out there trying to sell drones, [35:18] I mean it just is not solving the problem [35:19] that people actually care about. [35:21] Yes. Like Walmart, [35:22] Cleveland Clinic, the government of Rwanda, [35:24] none of these customers want to be [35:26] managing a fleet of autonomous aircraft. [35:29] Oh no, I don't think you'd sell the drones [35:30] to your current customers. [35:31] I think there'd be new customers [35:32] where maybe I want to build a crop dusting service [35:35] and your drones would be the best for doing crop dusting. [35:37] And so, I buy your drones [35:38] and they have all the safety equipment [35:40] and the high-quality servos and things like that, [35:42] but you're not currently in the crop dusting market. [35:44] Yeah. I mean the second reason I was going to say is, [35:46] I just don't think selling hardware in that way [35:47] is a very good business. Ah, okay. Yeah. [35:49] But I also think the third thing to consider is that like, [35:52] it's a bit like asking someone, like [35:54] oh, you know, Jeff Bezos like early days of the internet, [35:56] "Are you also going to like [35:58] build other internet businesses?" [35:59] You know, it's like the internet [36:01] is just a technology platform. Yeah. [36:02] And Amazon was building obviously, you know, [36:05] a new kind of retail company [36:07] and I think that, you know, [36:08] drones are a technology platform. [36:10] Zipline is a logistics company. [36:12] Like we are really focused on building automated logistics [36:15] for Earth in a way that [36:17] you want to build the first logistics system [36:18] that serves all people equally. [36:20] Like we would happily use any different kinds of tech. [36:22] Like if someone invents teleportation tomorrow, [36:23] we'll switch to that, because that would be better. [36:25] You're a logistics platform. [36:26] The big unsolved space was drones. [36:30] Will you ever do a modality that is not drones? [36:33] Very possibly. Huh. [36:34] What would that be? [36:37] We have lots of ideas. [36:37] Have a few more sips. Yeah, exactly. [36:42] I think that ultimately, if you really want to be [36:45] a new kind of logistics provider, [36:47] you probably have to be even more integrated, [36:51] like certain things are not going to go via the air, [36:53] certain things are going to go via the ground. [36:55] Okay. You get multimodal. [36:57] Yeah. You should be multimodal. [36:58] You know, UPS is multimodal. [37:00] Perhaps Zipline should be multimodal. [37:03] Keller is describing here [37:04] how Zipline got its start in Africa, [37:06] but they're now up and running [37:07] with drone deliveries in the United States. [37:09] And this is a technology pattern you see again and again. [37:12] Drone delivery, mobile payments, WhatsApp, QR codes, [37:14] these are all technologies [37:15] that started in emerging markets first, [37:18] and it's going to be the same with stablecoins. [37:20] So, here at Stripe we're all in on stablecoins. [37:22] We recently acquired Bridge [37:24] the leading stablecoin orchestration platform [37:26] and Privy, the best crypto wallet infrastructure. [37:29] We're already seeing a ton of stablecoin adoption, [37:31] especially with companies [37:32] that want to build cross-border financial products. [37:35] We're working with companies like Ramp [37:36] to build a global stablecoin backed spend card, [37:39] SpaceX, for global treasury management, [37:41] and Shopify, to enable payments [37:42] on Shopify stores from every country. [37:44] And then breakout startups like Chipper Cash and Dollar App, [37:47] they're offering dollar access [37:48] to people in Africa and Latin America, [37:50] but we don't think stablecoins are going to stop there. [37:53] Stablecoins are a programmable, low cost, [37:55] high throughput financial system [37:57] that'll have applications everywhere. [37:58] I know from my experience when I talk to CEOs, [38:01] they're all interested in the applications of stablecoins. [38:04] Stablecoin supply is up 50% in the past year alone. [38:06] They're definitely not a flash in the pan. [38:09] So, if you are a company building with stablecoins [38:11] or you're just thinking about your strategy there, [38:12] come talk to Stripe. [38:15] We have been describing Platform 1, [38:19] which is like we said, [38:20] fixed-wing, catapult-launched, sky hook recovered, [38:26] operating for the last decade in Rwanda. [38:30] You now have just launched Platform 2, [38:32] which is vertical takeoff and landing. [38:35] Looks more like a traditional multi-rotor drone [38:39] that people might picture when they hear a drone [38:40] though there's also a wing? [38:42] Yeah, it's a hybrid. Okay. [38:43] So, I forget what you call these, [38:45] but yeah, there's a wing and there's— [38:46] Vertical takeoff and landing, so VTAL. [38:48] Yeah. VTAL fixed wing thingy. [38:50] And you are using this for deliveries in suburban America. [38:56] Many questions on this. [38:58] First off, why construct a whole new modality? [39:00] Like it seems to me that [39:02] what you were doing with the precision airstrikes [39:05] of little packages under parachutes— [39:07] Yeah. [39:08] Would also work for my burrito. [39:10] So, why not do Platform 1 with my burrito? [39:15] I mean in fact it did. [39:16] The crazy thing is that when we, you know, [39:19] when we initially started building Zipline, [39:21] we kind of thought, oh, we're going to have to [39:22] build this very fancy complex technology. [39:24] It'll have to be a vertical takeoff [39:25] and landing fixed-wing vehicle. [39:27] We'll have to be able to deliver things [39:27] very precisely and gently. [39:29] And then we were talking to all these early customers, [39:31] it kind of, you know, we realized [39:33] and this is the power of just [39:34] going and talking to customers directly, [39:35] they only cared about cost. [39:36] Yeah. And speed. [39:38] They said they wanted to cover like maximum range. [39:40] They want it to be fast and they want it to be really cheap. [39:42] And so, we kind of ended up thinking like, [39:44] hey, you know, perhaps we should just design this [39:46] very simple fixed-wing system that would be, you know, [39:48] to a certain degree very unfancy. [39:50] Like when you see the way we're delivering there, you know, [39:52] it's just like this, you know, [39:54] cardboard box with a wax paper parachute. [39:59] It is amazing to me that, that system worked that well [40:02] and scaled for 10 years [40:03] to like 5,000 hospitals and health facilities [40:05] and became the largest commercial [40:07] autonomous system on Earth. [40:08] You know, when we say [40:09] 100 million commercial autonomous miles, [40:10] I don't think people fully get, [40:12] that's going to the Moon and back 200 times. [40:14] Yeah. And I presume actually, [40:16] I always find comparisons of to the Moon and back really. [40:19] Not that helpful. Okay. Not that helpful. [40:21] Here's another. [40:22] No one knows where the Moon is. [40:23] I presume your average journey is [40:25] order of 100 miles in Africa [40:27] and so, you're basically, talking a million flights. [40:29] Yeah, 1.5 million. Yeah. 1.5 million flights. [40:31] So, I think that's a pretty good number. [40:35] So, anyway like, not to, you know, digress, [40:39] but it's kind of amazing to me that Platform 1 [40:41] scaled as far as it did [40:43] with this delivered experience that was super unfancy. [40:45] Yeah. [40:46] And ultimately, as we were scaling [40:50] like crazy through 2020 and 2021, [40:52] there were all these big customers in the US [40:54] who all were kind of saying, [40:56] hey, like we want teleportation, [40:58] we would use that here in the US. [40:59] And so, I think that this was kind of our realization [41:01] that like it was time to expand basically, [41:04] and say, okay, if we were going to build a product [41:07] that can deliver directly to homes, [41:09] what does that need to look like? [41:10] And the design considerations for that are pretty different. [41:12] You know, in our opinions we need [41:15] to integrate really well with like existing buildings. [41:17] We couldn't build distribution centers [41:18] like we did with Platform 1. [41:20] We need to be able to basically, you know, [41:22] plug into the side of any existing hospital, [41:24] or primary care facility, or warehouse, or even restaurant. [41:27] And then from there you need to be able to deliver to homes [41:29] in a way that was fully automated, zero emission, [41:32] you know, 10 times as fast. [41:34] And most importantly, we want it to be silent. [41:36] Because we think that like people, [41:38] this is a kind of important thing. [41:39] I think sometimes people think drone delivery, [41:40] oh my God, it's going to be so annoying. [41:42] They have this nonstop buzzing of drones, [41:44] because that is how people experience drones today. [41:47] But the reality is that it doesn't have to be that way. [41:49] You can design systems that are actually, [41:51] way, way quieter than receiving a delivery via car. [41:54] And the crazy thing, sorry, that I forgot to mention [41:56] in my description of Platform 2 [41:59] is that the drone does not come... [42:02] the aircraft does not come anywhere near your house. [42:04] It sits up at 300 feet above your house. [42:07] And then long lines the payload down to you [42:11] in a little droid thingy and that drops it off, [42:14] but as a result the drone is hovering— [42:17] Far away. 300 feet above you. [42:19] Where I presume it's just inaudible basically. [42:20] Yeah, it's it's nearly inaudible. [42:21] Yeah. It looks crazy. [42:23] I was talking to an investor the other day. [42:25] It sounds crazy. Yeah, I know. [42:26] And I was talking to an investor the other day, [42:27] he wrote us this nice email and he said like, [42:29] "Hey, we really want to meet you, [42:30] because we think Zipline has built like [42:31] a fundamentally jaw-dropping consumer experience." [42:34] I do think when people come and see it for the first time, [42:36] it's pretty wild— Yeah. [42:38] Because you just never really experienced [42:39] something like zipping down to you [42:42] from like a football field above your head. [42:43] It's like out of science fiction. [42:44] Yeah. [42:45] It's kind of science fiction. [42:46] But the advantage is that droid that you described, [42:48] the little robot on the end is actually, [42:51] controlling its position on an x and y axis. [42:53] So, this is really important, [42:54] because like there, it's often windy, it's going to [42:55] be blowing. [42:56] You have different environmental conditions [42:57] you're trying to control for [42:58] and we want to be able to deliver [43:00] with like dinner plate level accuracy to any home. [43:03] And so, this enables us, [43:04] even when the wind is blowing really, really hard, [43:06] this enables us to like very reliably, safely, and precisely [43:10] deliver something even in a very, very tight area. [43:13] Okay. So, the reason that you needed [43:14] to develop a Platform 2 [43:16] for suburban deliveries in the United States was, [43:18] one, the place you're delivering from, the Chipotle, [43:21] you probably can't fit a catapult there [43:24] at least not in every one. [43:25] I'm pretty sure you can't. [43:26] And then, well, some of them you could, [43:27] there's a parking lot, you know, [43:30] but not reliably. [43:32] And then for the target, again, it might be accurate enough, [43:36] but the new system is definitely accurate enough. [43:38] Yeah. I mean, in fact like [43:40] the very first Walmart store that we ever integrated, [43:42] we actually integrated with Platform 1. [43:44] And that was even more shocking to me. [43:46] The customers loved like basically, [43:48] having like their stuff bombarded them, you know, [43:50] like from 30 feet up. [43:51] I mean it's like honestly crazy. [43:55] I think that was like the clearest signal to me of like, [43:57] man, if people were willing to put up [43:58] with that delivery experience [44:00] to get something delivered autonomously, [44:01] like wait until they see Platform 2. [44:03] So, the reason that we haven't had [44:05] drone delivery in the United States for a long time [44:07] is the FAA did not provide any rule making [44:12] beyond visual line of sight operations. [44:15] And you know, they're your regulators [44:17] so you probably can't criticize them in public. [44:18] I'll criticize them for you. [44:19] I think, you know, the FAA was slow [44:21] and roundly criticized for being slow [44:23] in not having any commercial drone frameworks. [44:25] since the old part 107 stuff for like hobby drone operators. [44:28] And line of sight is the test [44:30] where it has to be a light drone less than 55 pounds [44:32] and you have to be like there [44:33] controlling with your remote control, [44:35] which obviously, is not viable [44:36] for commercial drone operations. [44:37] So, for the longest time they sat on [44:40] commercial drone operations and didn't do anything. [44:45] Just now in the last month, we've seen Sean Duffy, [44:48] the Secretary of Transportation and acting head of the FAA, [44:51] announce new Part 108 beyond visual line of sight, [44:55] commercial drone rules [44:57] or at least, the notice of proposed rulemaking. [45:00] What do you think of the current FAA [45:03] and the rules they're proposing? [45:05] Yeah, I mean, first of all, you know, [45:07] Part 108 is kind of like codifying something [45:11] that has already been happening [45:12] over the last three to four years, [45:13] which is I think that about four years ago. [45:15] On an ad hoc basis. On an ad hoc basis. [45:16] And like via exemptions. Yeah. [45:19] Which is that about four years ago, [45:20] Zipline started seeking regulatory approval in the US. [45:23] We were able to go to the FAA and say, [45:24] "Hey, we have 100 million commercial autonomous miles [45:27] and zero safety incidents. [45:28] We think at this point you have the data you need." [45:30] And they agreed. [45:31] And so, very much to their credit, like, [45:33] you know, this is what really unlocked [45:34] like Zipline starting to scale aggressively in the US, [45:37] and we've done that via exemptions. [45:39] There are things called like 44807, [45:41] which is a different process [45:42] rather than getting an aircraft type certified. [45:44] If you really want to talk about, you know, [45:46] something that's broken, [45:47] we could talk about type certification. [45:49] Of drones or aircraft? Both. [45:50] Okay. Yeah, I mean so far. [45:52] Do you see the Mosaic stuff? [45:54] I have, I'm not an expert Mosaic. [45:57] Its amazing Part 23— [45:57] That is really cool. [45:59] Certification by making Part 22 much bigger. [46:01] But type certification has been, you know, [46:03] this interesting like blocker, [46:04] which is if you wanted to operate commercially in the US [46:07] you couldn't operate under 107. [46:08] Or if you wanted to operate [46:09] commercially, non-visual line of site, [46:10] you had to operate under Part 135. [46:13] And I fear that we're getting like, you know, [46:14] we're really in the weeds now, [46:16] but like, you know, I guess the main point here [46:18] is that like you needed to get an aircraft type certified. [46:20] And this is kind of the definition [46:22] of a very painful regulatory process. [46:24] Like it takes, you know, five to 10 years [46:26] to get aircraft type certified. [46:28] And it's really based on this idea of like [46:31] a regulator is going to look at every component [46:32] of the aircraft [46:35] and like independently verify that, that component [46:37] is meeting certain standards that were set by the regulator. [46:41] That might, you know, I mean, imagine trying to do that [46:45] in a world where like every year [46:46] you have a radically new and better version of technology. [46:47] It just doesn't work. [46:48] Yes, you guys are doing like [46:49] progressive hardware development [46:50] Super fast. And constantly figuring out [46:52] that, you know, two fasteners is better than 43 [46:54] and other such innovations. [46:56] You just have really, really fast iteration. [46:58] And so, in a world where type certification [46:59] takes five to 10 years, you're dead on arrival. [47:01] Yeah. And that's kind of what we saw [47:03] for like five or 10 years. [47:04] None of this innovation happened in the US. [47:06] I think the good news is that in like 2020, [47:08] the FAA kind of like woke up to this. [47:09] They realized like, hey, this is going to become [47:12] one of the most important industries in the world. [47:14] The US needs to be a front runner here. [47:16] We cannot be a laggard. [47:17] And the FAA actually, started to move super fast. [47:20] I mean, the FAA has been, you know, I know you think, [47:22] maybe I'm just saying this, [47:23] but like the FAA has actually, been a great partner. [47:25] Like there are so many people there [47:26] who want to win and who want to see the US win— [47:29] I think the FAA is full of [47:30] awesome people trying really hard. [47:31] I think the rule making has been slow [47:33] and I think exemptions and waivers [47:35] are not a substitute for a good regulatory framework. [47:37] Yeah. Well, there you go. [47:38] I mean, 108 really is like, I think, you know, [47:41] there have been these exemptions made [47:42] that luckily enabled Zipline [47:43] to like invest aggressively in the US [47:45] over the last four years [47:46] and experience pretty explosive exponential growth, [47:48] which we can talk about. [47:49] But like, you know, 108, [47:52] it is cool to see that now be codified [47:54] that is just the beginning of a process, right? [47:56] Because now the FAA has nine months to like actually, [47:59] operationalize and execute based on that rule making. [48:02] And we'll see how that goes. [48:04] But I think it's like, it's super important. [48:06] I mean, I think you and I probably both have grown up [48:08] reading about like the SR-71 program, [48:12] or a lot of these programs like to design airplanes. [48:13] I mean, there was a world where like the US [48:16] was iterating incredibly fast in aviation and aerospace, [48:19] and we've largely lost that. [48:21] Like our grandparents got to grow up in a world [48:23] where they could fly on supersonic jets. [48:26] We can't. Yep. [48:27] You know, I think people have this sense [48:28] that like the world moves inexorably forward. It doesn't. [48:32] Didn't you have a tweet about this? [48:33] Maybe like, you know, I think about it every day, [48:36] like actually, no, sometimes we literally move backwards. [48:39] Yes, yes. And in aviation, [48:40] like what plane did you learn to fly on? [48:42] I learned to fly on a Cessna 152. [48:44] And how old was that aircraft? [48:45] Yeah, it was from the 1970s, I want to say. Yeah. [48:48] What the hell? No. [48:49] I was actually, just ogling the Carbon Cub UL, [48:52] which uses a new Rotax engine. [48:54] So, there's this awesome [48:56] new engine technology out of Germany, [48:59] this company Rotax that makes them there. [49:01] And it hasn't been in the US certified market, [49:04] because it hasn't been worthwhile- [49:05] To get certification. [49:06] To go get the engine certified. [49:09] And now, they're selling it in the LSA regime. [49:14] Light-Sport aircraft. [49:15] Exactly. It is possible there. [49:16] And so, I think we're just going to see [49:17] this wave of interesting new, [49:19] because it's the kind of pent up technologies [49:20] that haven't been certified yet, [49:22] such as Rotax engines, which are better everywhere. [49:24] Have we lost like 98% of the audience? [49:26] By talking about 108 and 135? [49:29] I guess but just to share one thing... [49:30] No, but it's specific. [49:31] One thing, like to try to make it [49:32] even more general interest. [49:33] Like what really inspires me, you know, [49:35] is you think about like Skunk Works, right? [49:38] And like the teams that built a lot. [49:41] It is the Lockheed Martin factory [49:43] that made or kind of group, tiny group, [49:45] that made all the fun new stealth aircraft F-108 [49:48] and stuff like that. [49:49] SR-71 even more famously like the. [49:53] The U-2. Yeah, exactly. [49:54] And they didn't make F-104, they made the F-111. Sorry. [49:56] Yeah. And just to give like a specific example, [49:59] a couple examples. [50:00] One, like the first modern passenger aircraft [50:04] that was ever built, the DC-3, [50:05] I want to get everything right, [50:07] like DC-3 was built originally... [50:10] the entire design cost for that aircraft [50:12] was like $2.5 million or $3 million. [50:14] So, even when you convert into today's dollars, [50:16] that's like $75 million to design a passenger airliner. [50:20] Yeah. [50:21] The 787 costs $32 billion to develop. [50:23] Oh yeah, DC-3 was like napkin sketch [50:26] to finish aircraft I think in two years. [50:27] Yep. [50:28] You know, the 787 took 20-25 years, $32 billion. [50:32] It's literally, become 500 times as expensive [50:35] to build an airplane as it was for our grandparents. [50:38] And I think the crazy thing is you're going to be like, [50:40] oh, well, it's like that 787 [50:40] is a much more complicated airplane. [50:42] Yeah. The iPhone is also much more complicated [50:44] than it was five years ago. [50:45] So, if you think about the causes, [50:47] I think there's a great point to that. [50:48] You know, the 787 certification process [50:50] is so much more tortuous than the the DC-3. [50:53] If you think about why that is, [50:55] I think there is a few obvious things you could point to. [50:59] One is, and the one people would point all the time [51:01] is we have much less tolerance for unsafety than we used to. [51:05] The second is, the certification approaches [51:08] are much more prescriptive. [51:10] And so, they require that you do specific things [51:13] to prove how you're going to meet various tests [51:16] and stuff like that. [51:17] And then maybe you could say that [51:19] just the paper turnaround times have gotten slower [51:21] and you know, there's a lot of back and forth. [51:23] If you could wave a magic wand [51:25] to fix aircraft certification [51:28] either across drones or manned aircraft... [51:31] Again, you're put in charge of the FAA, [51:32] what would you actually do? [51:33] I think the answer is actually really obvious. [51:37] And the good news is I think this is largely [51:39] like already happening, [51:40] which is that we need to move away [51:42] from like prescriptive regulatory requirements [51:45] that are like, "Okay, now show me the, [51:49] life support system for the pilot [51:50] should we lose, [51:52] whatever atmospheric pressure in the cabin." [51:54] It's like it's a drone. [51:55] There is no pilot. [51:56] And then the FAA looks at you like, I don't know, [51:59] like we can't move forward, you know, [52:00] like I think there are a lot of things are just [52:02] you have all these prescriptive rules [52:04] about like how we're going to. [52:05] Need more performance based service. [52:06] You need performance based. [52:07] So basically, everything needs to transition to a world [52:10] where like the FAA is setting a statistical level of safety. [52:14] And from there, like any company can, you know, [52:20] basically self-certify against that level of safety. [52:24] And you can imagine different ways of like [52:25] what kind of data you might have to provide [52:27] to prove that you're meeting that level of safety, [52:30] but like there are a lot of different ways [52:31] to build something that's safe. [52:33] And what you want is like engineering innovation, [52:36] you want people trying new things. [52:37] If you try something new and it's an amazing innovation, [52:39] it's suddenly like blowing it out of the water, [52:41] like with a prescriptive absolute regulatory framework, [52:46] you lose all of that. Yeah. [52:47] Because it's basically, like the FAA [52:49] is just looking to make sure every subsystem [52:51] that's like the way we've been [52:52] designing airplanes for 100 years— [52:54] The subsystem by subsystem stuff [52:55] is crazy rather than performance based. [52:56] It's really painful. Yeah. [52:57] So anyway, I know this might be like [52:59] a little like in whatever, you know, shop talk or whatever, [53:02] but I think that it's really important. [53:05] And I also think you can probably see similar versions [53:08] of this battle playing out in many different parts [53:10] of like the US government, you know? [53:11] When you look at the way we build like nuclear power plants, [53:14] you know, we created, what is it called? [53:16] The nuclear, is it the— [53:17] Nuclear Regulatory Commission. [53:18] Nuclear Regulatory Commission [53:19] started in I think like 1978. [53:21] The commission has started to ensure [53:23] that power plants are safe, [53:25] and since then we've built zero power plants, I think. [53:28] Or is it like one? [53:29] I think it's zero. [53:30] Yeah. Yeah. Yeah. [53:31] It is definitely the rate, It's a bit painful. [53:33] The rate slowed down. [53:34] The commission achieved its goal though. [53:35] Yeah. Yeah. [53:36] Because we built zero, and that's like painful, you know, [53:39] as an American it's like, ah. [53:42] You're describing failure [53:43] to engage with trade off space— [53:45] Yeah. [53:47] In an intellectually honest way [53:48] where there's a trade off between safety and cost [53:50] and you just have to actually, address those questions. [53:53] And there's a trade off in speed and say these things. [53:55] And the cool thing in aviation is that [53:58] like there's now 50 years of pent-up technology [54:01] that is totally not been put to work at all. [54:04] It's actually very similar. [54:05] I think it's why SpaceX is like this, you know— [54:07] Totally. Enormous success. [54:08] Like, because they were like, oh yeah, [54:09] we'll just actually, use modern microcontrollers. [54:12] Yes, yes, yes. Or modern semiconductors. [54:14] Whereas like, you know, ULA was building on a semiconductor, [54:19] you know, architecture I think was like 30 years old, [54:21] because they just didn't want to certify [54:22] a new flight computer. [54:24] And so, you've launched in Texas now? [54:27] Just Texas in the United States? [54:28] Well, we launched originally in Bentonville [54:30] and now we're expanding in Dallas. [54:31] Okay. So, you're expanding in Dallas [54:33] and delivering for people like Walmart and Chipotle. [54:36] And where else should people be ordering from in Dallas? [54:39] I mean, you can just go to the Zipline app. [54:41] Okay. So yeah. [54:42] And we have all the different storefronts on there. [54:43] You go to Zipline, [54:44] download the Zipline app. Those are the big ones. [54:45] Yep. Those are the biggest. [54:47] And I mean, you know, to give a sense, [54:48] right now we're launching— [54:49] Give me your explosive growth stats. [54:50] I mean it's kind of crazy. [54:51] It's really only the last three months, [54:52] but we're now launching a Walmart Supercenter every week. [54:56] You know, those stores are ramping extraordinarily fast. [54:59] I mean, I don't know exactly how much I'm allowed to share. [55:01] It's probably more for Walmart to share. [55:03] Give some more magnitude. [55:04] I mean, order of magnitude, you know, [55:07] the service is becoming quite large [55:09] relative to the way that people receive. [55:12] That is not relative. [55:14] How many drones do you're flying in the Dallas area? [55:16] Yeah, so right now [55:17] we are integrated with about 10 Walmarts [55:21] and we're launching one a week. [55:22] And each Walmart will have somewhere [55:25] between like eight and 20 Zips flying from it. [55:28] So, it gives you like. [55:29] So, you have hundreds of drones. [55:31] Yeah. [55:32] Now, delivering in the Dallas area. [55:33] Serving, you know, [55:34] and I mean the crazy thing is the service. [55:35] And each drone can do hundreds of deliveries a day? [55:39] Like I think, 10 each. [55:40] Yeah, high tens is the way I think about it. [55:42] So, you have the capacity for a thousand-ish, [55:44] thousands of deliveries a day. [55:45] Yep. That's the way I think about it, [55:46] and that's growing rapidly. [55:47] We're actually expecting to get [55:48] to 10,000 deliveries a day in Dallas. [55:50] Like that's actually, the company's goal— [55:51] Great. [55:52] Moving as fast as we can toward that. [55:54] Looking at the statistics, the service has been growing [55:56] 25% to 30% week over week for the last three months. [56:00] And the interesting thing— That's cool. [56:01] The last couple things. [56:02] I mean, really it's only the last three months that like, [56:04] because we only launched in Dallas [56:05] on April 1st or April 4th I think. [56:08] Yeah. [56:09] So, seeing it grow like that, [56:10] the crazy thing is that about a month ago [56:12] we were starting to have all these capacity issues, [56:14] because it's like we, you know, [56:15] we can't do as many deliveries. [56:16] It's growing really fast. [56:18] So, we decided to turn off all the marketing. [56:20] We turned off the in-app notifications, [56:22] we turned off the field service marketing. [56:24] Basically, all the demand generation marketing [56:26] we're not doing, [56:27] and the service, just like, [56:29] you know, I mean we hit new records. [56:30] Because it's just a much better product. [56:31] Yeah. Saturday we hit a new record, [56:33] like all-time daily delivery record. [56:35] And then on like two days ago, [56:37] and then yesterday we blew that record [56:39] out of the water by 20%. It's crazy. [56:42] So, I think that, you know, a couple takeaways, [56:45] like one is that it is amazing how quickly science fiction [56:48] becomes entitlement. Normal. [56:50] Like, you know. [56:51] We see this with Waymo in San Francisco. [56:53] It's exactly like Waymo [56:54] where people get in and they're like, [56:54] "Oh my God, autonomous car, insane." [56:57] And then three minutes later [56:57] they're like scrolling on Instagram, you know? [56:59] And I mean, when you go to Dallas and you just. [57:01] What's the version of that in Dallas? [57:02] I think the version in Dallas is that like, you know, [57:04] for the first day that we launch a new Walmart, [57:06] you'll see people kind of [57:07] like on the hoods of their cars [57:09] just like watching the service operate. [57:11] You know, like people are kind of taking note [57:12] and then three days later people [57:14] are just totally going about their normal lives. [57:15] Yeah. Do not care. [57:17] You've got this like, installation of autonomous aircraft. [57:19] They're just coming and going nonstop at crazy volume. [57:22] Yeah. [57:23] You know, I think the other example is like, [57:25] I mean I visited a 78-year-old grandma [57:27] who's ordered 340 times from Zipline over the last year. [57:31] She's been like the country— [57:32] I was going to ask are you seeing induced demand [57:34] that had like— This induced demand. [57:35] Where people are ordering more stuff? [57:37] It's crazy. [57:38] Like she's been, I think the county clerk [57:39] for 40 years straight. [57:41] And you know, she's just showing me, she's like, [57:42] "Oh yeah, like let's place an order right now." [57:43] And she's like around on her iPhone, you know, [57:46] putting everything into her basket. [57:47] She's like, double click the side of the phone, [57:49] face ID, Apple Pay. [57:50] She's like, "It's on its way. [57:51] It'll be here at eight minutes, you know? [57:52] She's 78? She's 78 [57:53] and she's totally living in the future. [57:55] And you know, or you talk to like moms who are like, [57:58] oh yeah, this, you know, [57:59] buys me back like three or four hours a week. [58:01] I can just like be with my kids [58:03] or like take care of the family [58:04] rather than be like out in a car [58:05] trying to shop or get something. [58:07] So, I think a lot of families we're seeing now [58:09] ordering like three to four times a week on average, [58:11] where they'll do like a big grocery run once a week [58:14] and then they can do three to four. [58:16] Sometimes they talk about them as like top up visits [58:19] where it's like, if you don't know [58:20] what you're making for dinner, [58:21] or are you like something unexpected, [58:22] or you have people over, [58:23] you just get everything you need delivered autonomously. [58:25] Why can we not have this in California? [58:28] Why are you in Texas and not California? [58:30] Because like, I would've thought [58:31] from a regulatory point of view, [58:32] this federal preemption of the airspace, [58:34] and so, I would've thought all places are equal. [58:36] Why is Texas a more hospitable environment [58:38] for this than California? [58:40] You could ask the same question about Rwanda. [58:41] I mean I think that, you know, Zipline has probably been— [58:43] Rwanda had a different regulatory framework, [58:44] like your drones are banned in the US [58:45] whereas drones aren't banned in California. [58:47] No. However, you know, [58:51] a lot of what we do is still about a partnership [58:53] with like the community and the government. [58:55] I see, you need a permit from, [58:56] for the base or whatever. Still need a permit [58:57] to build like, yep, exactly. [58:59] We're still permitting [59:00] the ground infrastructure for example. [59:01] It's a bit like, you know, [59:02] Tesla building supercharger infrastructure in the US. [59:05] Who do I need to write a letter to? [59:06] Is it by mayor? [59:07] We're working on it. [59:08] Is it my local representative. [59:09] Yeah, we're working on it. [59:10] But to be honest, like as someone who lives in the Bay area, [59:13] and so, like 100% feels your pain. [59:15] It's like why is this not like serving my house right now? [59:17] Like the vast majority of technology [59:19] has first focused on serving like, you know, whatever, [59:22] the elite cities on the coast of the United States. [59:24] I actually think it's okay [59:26] that for once like technology is kind of going— [59:28] Because this works better in a low-density area. [59:29] It's going the opposite. [59:30] Well, it's actually, I mean Dallas is very high density. [59:33] Like, I mean it's not Manhattan, [59:35] but it's probably denser than San Francisco. [59:37] I wouldn't say it's as much about density. [59:39] It is a lot about like, you know, [59:41] we typically, want to launch somewhere [59:43] where people are freaking psyched to have us. [59:45] And where there isn't, you know, [59:47] I do think there's a little bit of a flavor. [59:48] I mean you had Kyle on here, right? [59:50] And like seeing what Cruise went through in San Francisco [59:52] with the conning of the cars. [59:53] Our war on electric scooters here in San Francisco. [59:57] Hey, I'm not judging. [59:58] But we have a new mayor, Daniel Lurie. [60:00] I know. He's doing awesome. [60:01] He's doing great. [60:02] I think that, look, I just think that [60:04] sometimes cities might, you know, [60:06] if they just assume that they're always [60:07] going to have the best technology [60:08] and that like they don't have to, you know, [60:10] do anything about, or I don't know. [60:12] I think that there are a lot of cities and states out there [60:15] that like are 100% like aggressively courting, you know, [60:20] infrastructure, new solutions, they want high-paying jobs, [60:24] and they're like unapologetically [60:27] excited about new technology. [60:28] So, you're going to go to the cities that want you first, [60:30] and then you can get around to the cities [60:32] that are more ambivalent. [60:34] I think what's going to happen is that at some point, [60:36] you know, again, there are like certain cities [60:39] that are just extremely forward-leaning. [60:42] Like an example, I brought one of our board members [60:45] from San Francisco, you know, [60:48] leads one of the biggest funds in the world [60:50] here in San Francisco to Dallas [60:52] and he was like looking at what we were doing [60:54] and his, I mean, I could see the smoke [60:56] coming out of his ears. [60:57] He was like, "Oh my God, you guys are just doing things." [61:01] You know, we went to a site [61:03] where we were building charging infrastructure [61:05] and we had literally, broken ground three days previously. [61:07] There was a huge hole in the ground [61:09] and he was like, "so, when is this going to be online?" [61:11] And I was like, "It's going tobe online [61:12] in 15 days before the board meeting." [61:14] We had a board meeting in 17 days. [61:15] And he was like, "That's insane. [61:17] It is not going to be ready for the board meeting. [61:18] What are you talking about?" [61:19] And it was actually ready like [61:20] five days before the board meeting. [61:22] Basically, different kinds of cities [61:24] have different requirements [61:25] and like there are places in the US [61:26] where you can get things done super, super fast. [61:28] Yeah. [61:33] We are here in south San Francisco just down the road [61:35] from your aircraft manufacturing facility. [61:37] Tell me about that. [61:38] You know, how do you build them? [61:40] How many are you building? [61:41] How many will you build? [61:42] You know, the most important thing [61:43] is the speed of engineering iteration. [61:46] I think you'll appreciate this. [61:47] You know, that's something I think that's been [61:49] a very well understood concept in software companies. [61:52] Not so well in hardware. [61:53] Way harder to do in hardware. [61:55] And so, the way to do that is full vertical integration [61:58] and then having engineering [62:00] practically on top of manufacturing [62:02] so that like as you're rolling [62:04] hardware changes through onto the line. [62:06] You have like the engineers who design those changes [62:08] working with manufacturing engineers [62:09] and then working with the people on the line itself [62:11] to like figure out how to make that work [62:13] and get it into production. [62:15] And so, even though it seems like a bit of a weird thing [62:17] to say, like, "Yeah, we're producing all these aircraft [62:19] here in South San Francisco." [62:21] The value of us being fully integrated in that way, [62:24] what it does for the speed at which the plane can iterate [62:27] and evolve the product is way more valuable than like, [62:30] "Oh, it's more expensive to hire someone [62:32] in South San Francisco than it is in Nevada," [62:35] or something like that. [62:37] So, it's an easy trade. [62:38] So yeah, this facility... [62:39] and we just expanded to 160,000 square feet, [62:42] it'll be capable of building [62:43] around 50,000 aircraft a year at scale, [62:47] which should get us to like near national scale [62:49] in the United States. [62:51] 50,000 aircraft a year. [62:53] And how many aircraft will like [62:54] a medium sized city require to support it? [62:57] I think it's like around 1,000. [62:59] It depends on the city. [63:00] Cities come in a lot of different shapes and sizes. [63:02] Yeah. Okay. Yeah. [63:03] So, 50,000 a year is really a lot [63:06] Compared, you know, again, [63:07] it depends what you're comparing to. [63:08] If you're comparing to iPhones, not really, [63:09] but if you're comparing to like airplanes, [63:11] then like, yeah, really a lot. [63:12] Like compared to the Cessna— [63:14] Yeah, exactly. You learned to fly in. [63:15] Yeah. Okay. [63:17] And so, you'll be able to, [63:18] in your current production facility in South San Francisco [63:20] produce 50,000 drones a year? [63:21] Yeah. Three shifts. 24/7 operations. [63:24] That's great. [63:24] We already operate our test sites 24/7. [63:26] Huh. What do you do with the test sites? [63:28] What do you test? [63:30] Yeah, I think, you know, [63:31] when people come visit the test sites, [63:33] I think that's one of the things about Zipline [63:34] that is most mind-blowing to people. [63:36] And I think it's just, it doesn't really look [63:38] like anything people have seen before. [63:42] You know, people often ask us, they're like, [63:43] "Wow, 120 million commercial autonomous miles [63:45] and zero safety incidents, why don't other people do that?" [63:49] I mean it turns out that, you know, [63:51] designing something that's safe is hard. [63:54] Reusable rocket, that's a good idea. [63:55] Yeah, exactly. [63:57] It's like, yeah, everybody should just do it. [63:59] It turns out some of these things are hard to do. [64:01] And I think the way you do it is [64:03] it's all just like, you know, none of this is fancy. [64:05] It's just practical problem solving. [64:06] And so for us it was like having three major layers of, [64:11] you know, testing and validation. [64:12] So, the first layer is like software testing. [64:14] We do a huge amount of testing, you know, [64:17] at 10 or 100 times real-time speed [64:20] to like test all the different software systems [64:23] that we are constantly iterating. [64:25] And there's, you know, 10 times [64:26] as much software as there is hardware. [64:27] So, this is like the airplane avoidance [64:30] you were describing? Yep. [64:31] It's like you input a signal of like, [64:32] oh, you're near an airplane [64:33] and check that it avoids the airplane [64:35] in a different fashion? Yeah. [64:35] Autonomy, computer vision, you know, [64:39] guidance navigation controls, communications architecture, [64:42] all the backend cloud services like these can all be, [64:45] you know, you can do a huge amount of software testing. [64:47] Because I think the other major way to think about... [64:50] What is the problem we're trying to solve? [64:51] Do you know how long it takes Boeing [64:52] to do a software update? [64:55] A lot of time, I'm guessing. [64:56] You want to guess? [64:57] No. You don't want to guess? [64:58] I'm too horrified. Come on, guess. [65:00] Like from from when to when? [65:03] For example, from when they realize [65:04] that their planes are falling out the air [65:05] and killing people— I see. Sure. [65:06] To when they can then have the software update out. [65:09] And so, the airplane's back in the air. [65:10] I mean order of a year? Three years. [65:12] Oh, that long? Yeah. [65:13] Again, this is part of the problem [65:15] of like how the FAA regulates software [65:16] with those kinds of aircraft. [65:18] But, you know, Zipline right now [65:19] is doing a full new version. [65:22] Like we're releasing new software versions [65:25] to the entire global fleet every 30 days. [65:27] And so, validating that, you know, [65:29] basically, it might be, you know, 50 to 100 bug fixes. [65:34] Maybe, you know, 20 to 50 performance improvements [65:37] and then like five to 20 [65:39] major feature additions or upgrades. [65:42] And that gets validated through the software testing layer. [65:45] Then it gets validated through HITL, [65:47] which hardware in the loop testing. [65:49] It's basically like all of the avionics of the aircraft [65:51] assembled on shelves that are kind of like [65:54] the brain of the aircraft connected to the matrix. [65:57] They're all operating in simulation. [65:58] We can detect like 95% [66:00] of what's going to go wrong on an aircraft, [66:02] we can detect in HITL. [66:03] Before we then go to the final layer, [66:05] which is the test sites. [66:06] Test sites have hundreds of aircraft operating 24/7. [66:10] So, they never stop. [66:11] Even, you know, Saturday at like 3:00 am [66:14] you could go to the test site [66:15] and you'll see like the aircraft are just there [66:17] and there are humans overseeing them. [66:18] You're describing your build manufacturer [66:22] testing infrastructure in the United States. [66:25] One of the top discussed topics [66:28] right now is American dynamism, [66:31] the reemergence of American manufacturing. [66:34] And when people talk in DC [66:36] about the importance of American manufacturing [66:38] and the criticality of the supply chain [66:41] in the United States, you know, [66:44] the example they often point to as a supply chain [66:48] that we have outsourced is obviously drones. [66:50] Drones where there is not kind of as serious— [66:55] I thought you're going to say semiconductors. [66:56] Oh, well, semiconductors— [66:57] Semiconductors is like even more scary— [66:58] Semiconductors get talked about a lot. [66:59] Yeah, yeah, for sure. [67:01] But you know, there's the headlines [67:02] about Intel and everything, [67:03] but people also talk a lot about drones [67:04] and just how much more scaled up [67:07] the Chinese drone industry is than the US drone industry. [67:10] People also talk about the fact that it's not competitive [67:13] to manufacture in the United States anymore, [67:15] because supply chain, I don't know, [67:16] did you read the "Apple in China" book? [67:17] It was like a pretty good— [67:18] I've heard of it. I haven't read it yet. [67:19] It's great. And it's like a good history of Apple's [67:22] manufacturing in China. Why did this happen? [67:24] Yeah. And they describe kind of the emergence of it, [67:25] but in particular, you know, people sometimes [67:27] thought of China as just low wage. [67:29] Whereas the book correctly describes [67:32] just how much agility there was in adapting supply chains [67:35] and being able to manufacture there. [67:37] And so, you are a company manufacturing [67:40] at scale in the United States. [67:42] What do you think of this topic? [67:43] Yeah. To put a finer point on it. [67:46] In a week, China makes more drones [67:48] than the US makes in a year. [67:50] That is a finer, finer point. [67:51] Yeah. So, you know, it's a little spooky, [67:52] and obviously, you know, you don't have to be like [67:56] a deep student of global events to understand that. [68:00] Like, wow, I mean this technology's clearly [68:02] kind of a like remaking not just industries, [68:04] but remaking warfare, and that's their power. [68:05] And people watch the war Ukraine [68:07] where it is now all drone determined from airspace. [68:11] 100%, like everything is attritable [68:13] and like you wouldn't want humans on the battlefield at all. [68:16] And I think this is like a good news, bad news thing. [68:19] So the bad news is like, yeah, it's true. [68:21] You know, China has like massive manufacturing advantages [68:24] and massive scale in this area [68:26] that the US does not currently have. [68:28] But that is mainly for like, you know, [68:31] plastic quadcopters that cost $1,000. [68:33] Like it's definitely a different class of aircraft. [68:35] When you look at like where this is all going, [68:38] the systems that are really going to matter, [68:39] whether it's like military or logistics, [68:43] they aren't like $1,000 quadcopters, [68:45] they're more complicated technologies. [68:47] They require really strong integration of like, you know, [68:50] software, AI solutions with the hardware itself. [68:53] And again, the hardware is only like [68:55] 10% of the complexity of the overall solution. [68:57] So, in those areas, the US is actually leading, you know? [69:01] The US is leading both in terms of [69:03] I think drone delivery and logistics, [69:05] but also, you know, leading on the software side. [69:08] That's right. Those areas being? [69:10] I would say like more advanced [69:13] hardware, software integrations that are required for like [69:17] really good products for military [69:18] and really good products for logistics. [69:20] Okay. [69:21] So, when we're talking about the 50x, that's like, [69:22] because you have DJI in China [69:24] or like that, you know, [69:25] they're building 10,000 for like drone light shows, [69:28] those kinds of things. [69:29] That is not necessarily the battlefield [69:32] that is going to determine, you know, [69:35] who really like is the front runner [69:37] in the industry that matters the most. [69:39] And so, I think that it's too early to say. [69:42] I think, you know, China has obvious advantages [69:44] and that like they have massive manufacturing scale. [69:47] You mentioned the Apple supply chain book. [69:50] That is a huge reason. [69:51] The fact that Apple outsourced so much of that manufacturing [69:54] and so much of that knowledge [69:55] and expertise is now in Shenzhen, [69:57] like now Shenzhen's ability [69:58] to build phone adjacent products, [70:00] which I think you could definitely describe drones. [70:02] Even cars you could describe as phone adjacent now. [70:05] Sure as, yeah. [70:06] Xiaomi will tell you. Xiaomi, yeah, [70:08] or Tesla could attest, you know? [70:11] It's actually those supply chains [70:12] that make it possible to then like branch out [70:14] and build a lot of these other products way faster. [70:16] So, that's the advantage that China has. [70:18] The advantage of the US has is like a lead in AI, [70:20] a lead in more advanced hardware and software integrations. [70:24] I don't know who's going to win. [70:26] Yeah. The point the book makes is that [70:28] Apple's direct investment in China is much larger [70:31] even in inflation adjusted terms than the Marshall Plan— [70:35] It's insane. In Europe [70:36] where it's just a kind of a huge investment [70:38] in manufacturing capability. [70:40] I was thinking about that on the way here. [70:42] How that's going to look, you know, [70:44] with the benefit of hindsight. [70:46] Corporations sometimes make like [70:48] the easy short-term decision to maximize profit [70:51] instead of like the more strategic decision of like [70:53] how do we set ourselves up for long-term success? [70:56] And it really does not seem like it would be that hard to, [70:59] you know, ensure like to achieve some of the company goals [71:03] that a company like Apple wanted to achieve, [71:05] but also ensure that you, you know, [71:08] maintained like manufacturing dominance and kept it at home. [71:11] Or at least ensured that you that as a strategic option. [71:14] Like the fact that we're now in a position where [71:17] we don't have the ability [71:18] to build our own smartphones period. [71:21] And where, you know, depending on what happens with Taiwan, [71:24] we not may not have the ability [71:25] to even build modern computers. It's spooky. [71:30] Last question. [71:31] You're one of the success stories [71:34] that people hold up in hard tech [71:38] where you've built a product [71:39] that is a now successful business line. [71:42] You found a good way to start [71:44] that wasn't too capital consumptive, [71:46] were able to get like a real product out there in the field [71:48] and refine it from there. [71:50] And now, you have a product [71:52] that people love in several countries. [71:53] What advice would you have [71:55] for hard tech founders starting out today? [71:57] It certainly seems to me like, you know, [71:59] the whole American dynamism, what do you call it, you know, [72:02] paradigm that we kind of see, [72:04] I think it's making things easier. [72:06] When I moved to the Bay Area in 2013 [72:10] with the goal of kind of building Zipline, [72:12] the most exciting company in the entire, [72:16] in all of Silicon Valley, probably in the entire US, [72:18] was a company doing file sharing. [72:20] You know, it was Dropbox, [72:21] and they had like the best engineers, the best growth story, [72:25] the most ambitious vision. [72:27] And it was, I think at that time like a $5 billion company, [72:30] which is like more money [72:31] than I could possibly imagine, right? [72:32] And that was like the pinnacle of what you could achieve [72:35] in terms of building a startup in Silicon Valley. [72:37] And so, it's interesting to think that in just 12 years, [72:42] like the scope and scale of human ambition [72:44] has radically increased. Yes. [72:48] I mean, whether it's Tesla, or SpaceX, or Stripe, [72:51] or you know, OpenAI, [72:53] or you know, Brian Armstrong was on here, right? [72:56] He is like working on NewLimit. [72:57] They're trying to make humans immortal [73:00] and a lot of these ideas are like working and scaling. [73:03] There are people building nuclear fusion power plants. [73:07] I think that's like kind of probably underappreciated [73:11] just how much change has happened in just 12 years. [73:14] And I think that, you know, I guess to answer the question, [73:17] my biggest advice is that I think, [73:19] especially when it comes to hard tech, [73:20] like the whole challenge is just [73:22] designing something like incredibly simple [73:24] and getting it into the real world quickly [73:26] and learning by doing. [73:27] It's very easy to be in a bunker [73:29] or an R&D lab building cool technology. [73:32] Really hard to like force yourself into the real world [73:34] and then learn by doing. [73:36] I think that, that like Zipline, you know, [73:37] we never thought we were the smartest team. [73:39] We never thought...We were certainly [73:40] not the best funded team, [73:42] but we always had this like deep conviction [73:44] that we were the most practical team, [73:46] and we had this mission that really inspired us. [73:48] We knew exactly what we needed to do, [73:50] it was like deliver to those 21 hospitals. [73:52] And we forced ourselves to launch something [73:54] that to even our own eyes we were quite embarrassed. [73:57] It's like, oh my God, like is that going to... [73:58] I mean, are people going to put up [74:00] with that delivery experience? [74:01] And it turned out that like even that [74:02] very simple version of the product scaled very fast. [74:05] And so, I think that it sounds obvious when you say it, [74:08] and yet, it's extremely rare, [74:10] especially in hard tech companies. [74:12] Keller, thanks for coming by. [74:13] Yeah. Thanks, John.