Oklo Builds 14-GW Reactor Pipeline as AI Data Centers Face Power Crunch

AI’s Power Problem and How To Fix It With Olko Inc — Ai4

Ai4

Exhausted electric grids are forcing tech hyperscalers to abandon short-term utility planning and secure modular, liquid-metal nuclear reactors to power massive AI clusters.

AI-assisted summary of AI’s Power Problem and How To Fix It With Olko Inc — original source. Check the original for full context.

Source published:

Detailed notes

## Background Published on YouTube by Ai4 on August 26, 2025, this presentation features Brian Gitt, Senior Vice President and Head of Business Development at Oklo Inc. The session explores how compact, walk-away safe advanced nuclear power plants can overcome historic cost, construction timeline, and public perception barriers to reliably power AI and cloud computing infrastructure.

## TL;DR Brian Gitt explains how surging data center electricity demand is colliding with rigid grid capacity limits, necessitating dedicated behind-the-meter clean energy solutions. Oklo tackles traditional nuclear hurdles through a Power-as-a-Service model deploying modular 75-megawatt liquid-metal fast reactors that boast passive safety, negligible water consumption, and spent fuel recycling capabilities. To solve immediate speed-to-power bottlenecks, Oklo integrates short-term natural gas bridging assets while progressing toward commercial nuclear deployments and fuel recycling facilities across its 14-gigawatt pipeline.

## Atmosphere & style The talk carries a crisp, pragmatic executive tone that blends sobering infrastructure warnings with high-conviction engineering optimism. Gitt speaks with polished clarity, cutting through speculative hype to ground Oklo's value proposition in thermodynamics, historical operational data from national laboratories, and commercial discipline. The session feels like an authoritative industry briefing, shifting from a structured technical overview into a direct, engaging dialogue during the interactive audience Q&A.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=4s) Data center power demand is on track to triple by 2030, driven by massive projects like an 11-gigawatt behind-the-meter campus in Amarillo that exceeds the entire electric capacity of Nevada. - [01:41](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=101s) Traditional nuclear energy has stagnated for 30 years due to three primary structural barriers: $9B–$10B upfront plant capital costs, 8–10 year construction cycles, and public safety/waste perceptions. - [03:27](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=207s) Oklo eliminates upfront capital expenditure for hyperscalers through a Power-as-a-Service model, selling electricity under 20-year Power Purchase Agreements (PPAs) as an independent power producer. - [03:27](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=207s) Using liquid metal coolant instead of pressurized water eliminates the need to site reactors near major water bodies and enables a rapid 36-month deployment on a compact 3-acre footprint. - [06:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=411s) Deploying 75 MW modular blocks matches data centers' phased 100 MW hall buildouts and provides continuous availability through staggered maintenance rather than monolithic plant outages. - [06:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=411s) Oklo has built a 14 GW commercial customer pipeline, securing commercial agreements with major data center operators including Equinix and Switch. - [08:44](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=524s) A 500 MW modular nuclear installation requires just 21 acres running 24/7 continuous baseload, compared to 2,700 acres for an equivalent intermittent solar installation in Arizona. - [10:34](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=634s) Roughly 70% of Oklo's powerhouse components—including an off-the-shelf Siemens Energy steam turbine—come from standard commercial supply chains, avoiding bespoke nuclear manufacturing costs. - [12:11](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=731s) The reactor operates at ~450°C in an unpressurized stainless steel vessel, enabling thermal coupling with Vertiv absorption chillers to directly cool data center server halls. - [15:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=951s) Built on 30 years of operational data from Idaho National Laboratory's EBR-II, the design features passive walk-away safety relying on natural convection and gravity, shrinking the Emergency Planning Zone to the 3-acre boundary. - [17:42](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1062s) Fast reactors can extract the remaining 90%+ energy trapped in legacy nuclear waste stockpiles, offering enough fuel to power the entire United States for 150 years without new uranium mining. - [20:36](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1236s) To solve near-term speed-to-power bottlenecks, Oklo partners with Liberty Energy to deploy natural gas generation within 12 months, transitioning gas to backup and balancing once nuclear comes online. - [22:13](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1333s) Supported by White House executive orders directing an 18-month NRC licensing timeline, Oklo's first planned plant at Idaho National Laboratory is designed as a fully commercial facility targeting 2028 deployment. - [24:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1491s) Hyper-scale operators are increasingly shifting to captive behind-the-meter microgrids to avoid straining regional utility grids and to prevent public backlash against residential rate increases.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=4s) The Unprecedented Power Appetite of AI Infrastructure

Opening the final session of the day, Brian Gitt, Senior Vice President and Head of Business Development at Oklo Inc., addresses the intersection of advanced nuclear power and the computational demands of artificial intelligence. Oklo designs, builds, and operates advanced nuclear power plants tailored directly to powering data centers and emerging AI workloads. To illustrate the scale of the challenge, Gitt points to an announcement made just a month prior: an 11-gigawatt data center campus planned in Amarillo, Texas.

To put an 11-gigawatt demand into perspective, Gitt emphasizes that this single campus exceeds the entire electric power capacity of the state of Nevada, Las Vegas included, and surpasses the total peak electricity demand of New York City. Crucially, this campus is designed to sit "behind the meter"—dedicated entirely to powering digital infrastructure without drawing from or depending on the broader utility grid. With OpenAI planning mega campuses of up to 5 gigawatts and Meta making similar power-heavy commitments, industry projections indicate that data center power demand is on track to triple by 2030. The central challenge facing the digital economy is finding a way to generate and deploy this massive amount of clean, reliable capacity on an unprecedented schedule.

- **Amarillo Campus Benchmark:** An 11 GW project behind the meter exceeds the entire generation capacity of Nevada and the total peak load of New York City. - **Hyperscaler Ambitions:** OpenAI is planning campuses requiring up to 5 GW of dedicated capacity. - **Demand Trajectory:** Broad industry consensus projects data center electricity demand will triple by 2030.

---

[01:41](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=101s) The Nuclear Stagnation and Its Three Historic Hurdles

Nuclear energy currently supplies 20 percent of all electricity generated in the United States, with 94 commercial reactors operating reliably for the last 70 years. Yet despite this dependable baseline, Gitt explains that the domestic commercial nuclear industry has remained essentially stagnant for three decades. In recent memory, the country has only managed to bring two new reactor units online—referring to the Vogtle expansion in Georgia. The stagnation is not caused by a failure of the physics, but by three persistent, structural barriers that traditional nuclear projects have failed to overcome.

The first barrier is prohibitive upfront capital cost. A standard 1,000-megawatt conventional nuclear plant requires a capital outlay of $9 billion to $10 billion. The second hurdle is the extensive construction timeline: these mega-infrastructure projects regularly take 8 to 10 years to build, tying up billions in unproductive capital before generating a single kilowatt-hour. The third challenge is public perception surrounding safety and nuclear waste. While Gitt notes that commercial nuclear is arguably the safest industrial sector in America—recording only one commercial accident in 70 years with zero injuries or fatalities—lingering fears over catastrophic meltdowns and long-term waste disposal continue to stall large-scale adoption.

```mermaid flowchart TD accTitle: Historic Nuclear Bottlenecks vs Oklo Solutions accDescription: Flowchart mapping traditional nuclear industry barriers to Oklo's modular solutions.

subgraph Traditional["Traditional Nuclear Challenges"] C1["High Upfront Cost: $9B–$10B per plant"] C2["Long Build Times: 8–10 years"] C3["Perception: Safety and radioactive waste"] end

subgraph Oklo["Oklo Modernized Approach"] S1["Power-as-a-Service: Zero customer capex, PPA model"] S2["Rapid Deployment: 36 months total (18 mo. build)"] S3["Inherent Walk-Away Safety & Fuel Recycling"] end

C1 --> S1 C2 --> S2 C3 --> S3 ```

---

[03:27](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=207s) The Oklo Model: Power-as-a-Service, Modular Siting, and Liquid Metal Cooling

To dismantle these three historic barriers, Oklo reimagined nuclear energy across three interrelated axes: business model, sizing, and core technology. Rather than building towering, multi-gigawatt facilities framed by concrete cooling towers that require 800 acres and proximity to an ocean or major river, Oklo builds compact 75-megawatt powerhouses designed to integrate cleanly into industrial parks, suburban environments, or master-planned communities. Crucially, Oklo avoids demanding that technology clients become nuclear plant owners. Recognizing that hyperscalers want power rather than infrastructure balance-sheet debt, Oklo acts as an independent power producer, financing the plants via long-term Power Purchase Agreements (PPAs) over 20-year horizons and selling electricity as a service.

From an engineering standpoint, Oklo slashes construction timelines from a decade down to 36 months—with actual on-site assembly taking roughly 18 months, alongside regulatory licensing and long-lead procurement. Each 75 MW plant occupies a mere three-acre footprint to the security perimeter. Furthermore, Oklo eliminates the need for massive external water bodies by utilizing liquid metal cooling instead of water. Gitt highlights first principles: heat can be extracted from fuel using water, gas (such as helium), or liquid metal. Because metal possesses vastly superior thermal conductivity, it transfers heat far more efficiently without needing pressurized water systems. Additionally, Oklo's advanced fast reactors can recycle used nuclear fuel; over 90 percent of the energy potential stored in the nation's existing spent fuel stockpiles remains untapped and can be converted into clean electricity.

- **Capital Structure:** Power-as-a-Service model eliminates customer upfront capital costs through 20-year PPAs funded by project-level debt and equity. - **Deployment Velocity:** 36 months from project greenlight to power delivery, featuring an 18-month on-site construction period. - **Footprint:** 75 MW per unit on just 3 acres of land up to the fence line. - **Cooling Architecture:** Liquid metal coolant replaces pressurized water loops, eliminating the requirement to site plants near lakes, rivers, or oceans. - **Fuel Recycling:** Capable of running on recycled commercial spent nuclear fuel, unlocking the remaining 90%+ latent energy in existing nuclear waste stockpiles.

---

[06:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=411s) Modular Synchronization, Redundancy, and the 14 GW Commercial Pipeline

A core strategic advantage of Oklo’s small 75-megawatt plants is synchronization with data center construction phases. While headlines routinely tout 500-megawatt or multi-gigawatt campus master plans, developers never construct them all at once; they build in phased increments of roughly 100 megawatts per individual data center hall over successive years. Siting a multi-gigawatt traditional plant creates massive, uneconomic surplus power that sits idle for years. Deploying power in 75 MW modular blocks matches computational buildouts in lockstep, ensuring capital efficiency and avoiding overbuilt idle capacity.

Smaller modular blocks also solve the availability and redundancy dilemmas that plague traditional baseload plants. Large 1,000 MW reactors must occasionally go offline for planned maintenance and refueling outages, creating an enormous supply vacuum that an "always-on" AI cluster cannot tolerate. With an array of modular 75 MW units, operators achieve high availability through staggered maintenance: individual units can be cycled offline and refueled sequentially while the remaining reactors maintain steady, uninterrupted baseload delivery to the computing cluster. This operational logic has fueled strong market demand, with Oklo amassing a 14-gigawatt commercial customer pipeline featuring major data center operators like Equinix and Switch.

- **Phase-Matched Growth:** Power increments of 75 MW match the standard 100 MW phased expansions of modern data center campus halls. - **Continuous Availability:** Staggered refueling and routine maintenance cycling across multiple modular units eliminate the severe power disruptions caused by taking a single 1,000 MW reactor offline. - **Pipeline Scale:** 14 GW of customer pipeline under commercial development, including key agreements with Equinix and Switch.

---

[08:44](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=524s) Energy Density and Environmental Footprint: Nuclear vs. Solar and Water Savings

Gitt grounds the long-term case for nuclear power in the physical reality of energy density: the sheer ratio of physical material required to generate a kilowatt-hour. Nuclear fuel is so concentrated that the entire energy consumption of an average human being across their whole lifetime can be provided by an amount of nuclear material small enough to fit inside a single golf ball. When evaluated through an environmental lens—accounting for raw material extraction, mining, supply-chain logistics, and land transformation—nothing matches the efficiency of nuclear.

To contextualize this spatial footprint, Gitt compares a 500-megawatt installation in Arizona using ideal solar radiation numbers versus an equivalent 500-megawatt modular nuclear array. The nuclear installation requires just 21 acres of land and runs 24/7, whereas a solar installation requires roughly 2,700 acres while only generating electricity during daylight hours. Furthermore, water conservation is critical across arid regions of the American West, such as Nevada and Arizona. Because Oklo’s reactors utilize liquid metal rather than open water loops for primary reactor cooling, operational water consumption is virtually negligible—restricted entirely to closed-loop makeup water for the secondary power conversion steam turbine.

| Metric / Dimension | Oklo Advanced Nuclear Array | Solar Farm Comparison (Arizona Benchmark) | |:--- |:--- |:--- | | **Capacity Target** | 500 MW baseload | 500 MW nameplate | | **Land Footprint** | ~21 acres | ~2,700 acres | | **Operating Profile** | Continuous 24/7/365 baseload | Intermittent (daylight hours only) | | **Primary Cooling** | Liquid metal (no environmental water draw) | N/A (requires panel washing) | | **Water Consumption** | Minimal closed-loop turbine makeup water | Dependent on local cleaning regimens | | **Fuel Density** | Lifetime human energy fits in a golf ball | Low energy density; requires extensive materials |

[10:34](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=634s) Water Efficiency and Off-the-Shelf Balance of Plant

At its fundamental thermodynamic level, an Oklo powerhouse operates just like any traditional thermal plant: it serves as a concentrated heat source that boils water into steam to spin a turbine. Because of this closed architecture, the only water consumed by the facility is minimal makeup water for the steam turbine loop. That low consumption puts it on par with solar power—a comparison that often surprises people until they realize that large-scale solar installations require massive volumes of water to routinely wash away blowing dust and dirt. When contrasted with conventional gigawatt-scale nuclear reactors or natural gas combined-cycle plants that draw enormous quantities of cooling water, Oklo's water requirements are virtually negligible.

A core pillar of Oklo's cost-reduction strategy is ensuring that roughly 70% of every component inside a powerhouse comes from standard, existing industrial supply chains without requiring specialized nuclear ratings or certifications. The entire balance of plant—the hardware that converts thermal energy into usable electricity—is anchored by an off-the-shelf steam turbine and generator package from Siemens Energy. Siemens has deployed around 5,000 of these proven industrial units across the globe. Oklo simply tunes the hardware to match its steam conditions. By confining nuclear-specific standards strictly to the fuel assemblies and the immediate surrounding structures, Oklo avoids the crushing capital costs typical of bespoke nuclear builds.

- **Non-nuclear components:** Roughly 70% of the powerhouse uses standard commercial supply chains. - **Siemens Energy partnership:** Utilizes a standard industrial steam turbine and generator design with an established global operating footprint of ~5,000 units. - **Water footprint:** Only requires minor makeup water for the turbine loop, comparable to solar panel washing and far lower than conventional thermal power stations.

---

[12:11](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=731s) Modular Scalability, 450°C Heat, and Data Center Cooling

Oklo designs its powerhouses like digital infrastructure, allowing customers to add capacity incrementally like Lego blocks rather than committing upfront to an indivisible multi-gigawatt station. A customer can start with an initial 75 MW unit, expand to 150 MW, scale up to 300 MW or 1,000 MW, and ultimately cluster up to 5,000 MW on a single site. Oklo is already in active discussions with hyperscalers about massive, multi-gigawatt campuses built along these modular lines.

The reactor operates at a high temperature of roughly 450°C, producing valuable thermodynamic efficiencies and high-grade heat that extends far beyond direct electricity generation. In collaboration with digital infrastructure provider Vertiv, Oklo is pairing its powerhouses with absorption chillers, allowing high-temperature steam to directly cool data center server halls in addition to powering them. Structurally, the core sits within an unpressurized stainless steel vessel—roughly 40 feet tall and 20 feet in diameter—fabricated from conventional 304 or 316 stainless steel. Tanks of this exact profile and metallurgy are routinely manufactured at scale for the petrochemical and pharmaceutical sectors, eliminating the need for exotic alloys or ultra-heavy pressure vessel fabrication.

```mermaid flowchart TD accTitle: Oklo Powerhouse Heat and Power Cycle subgraph Core ["Reactor Vessel (Non-Pressurized)"] Fuel["Fission Fuel Core"] --> LiquidMetal["Liquid Metal Pool"] end

subgraph Thermal ["Thermal Exchange & Steam Loop"] LiquidMetal --> HeatEx["Heat Exchangers (450°C)"] HeatEx --> SteamGen["High-Temperature Steam"] end

subgraph Offtake ["Electricity, Cooling & Heat Rejection"] SteamGen --> Turbine["Siemens Energy Steam Turbine\n(Electricity Generation)"] SteamGen --> Chillers["Vertiv Absorption Chillers\n(Data Center Cooling)"] Turbine --> Condensers["Air-Cooled Condensers\n(Waterless Heat Sink)"] end ```

- **Operating temperature:** ~450°C, enabling high thermodynamic efficiency, direct industrial heat supply, and thermal chilling. - **Vertiv partnership:** Deploying high-temperature steam directly to drive absorption chillers for data center cooling. - **Vessel profile:** Non-pressurized 304/316 stainless steel tank (~40 feet high, ~20 feet in diameter), matching standard industrial chemical process vessels. - **Scalable capacity steps:** 75 MW increments up to multi-gigawatt (5,000 MW) master-planned campuses.

---

[13:58](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=838s) Non-Pressurized Liquid Metal Design and Defense-Grade Reliability

Inside the main stainless steel vessel, the nuclear fuel sits at the bottom submerged in a pool of liquid metal. As fission begins, heat transfers directly from the fuel into the liquid metal bath, which circulates upward through internal heat exchangers to produce steam. Because the entire heat-dissipation system is paired with air-cooled condensers, large fans discharge the plant's excess heat directly into the ambient air, completely eliminating the need for an external cooling water source like a river, lake, or ocean.

This architecture delivers the ultra-high reliability demanded by mission-critical facilities. While the operating US commercial nuclear fleet already leads all generation types with an average capacity factor around 92%, Oklo's modular approach enables sufficient onsite redundancy to deliver effectively 100% uptime. This high degree of resiliency caught the attention of national security leaders, leading the United States Department of Defense to issue a notice of intent to award Oklo a contract to build a powerhouse for an Air Force base in Alaska.

- **Cooling loop:** Air-cooled condensers eliminate environmental thermal discharge and external cooling water requirements. - **Baseline uptime:** Conventional nuclear leads with ~92% capacity factor; modular multi-unit redundancy enables 100% effective availability. - **Military deployment:** Received a notice of intent to award from the US Department of Defense to power an Air Force base in Alaska.

---

[15:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=951s) Four Decades of Proven Heritage and Walk-Away Passive Safety

Oklo's core reactor technology is not an unproven laboratory science experiment; it is the direct commercialization of fast-reactor technology pioneered by Argonne National Laboratory and Idaho National Laboratory. Worldwide, roughly 25 liquid metal fast reactors have been constructed, accumulating over 400 reactor-years of operational history. Oklo’s specific design directly builds upon Experimental Breeder Reactor II (EBR-II), which operated reliably at Idaho National Laboratory for 30 years between 1964 and 1994. The vast trove of empirical operational and safety data generated by EBR-II is being submitted directly to the Nuclear Regulatory Commission (NRC) for Oklo’s commercial license application, eliminating the time and expense of constructing preliminary demonstration prototypes.

During historic testing at Idaho National Laboratory, engineers repeatedly subjected the reactor to simulated severe accident conditions—deliberately attempting to induce core damage by shutting off all cooling flow without scramming the plant. In every test, the reactor safely cooled and regulated itself. The system relies entirely on passive physics—natural air circulation, thermal expansion of materials, and gravity—without needing human operators, mechanical backup pumps, or emergency electrical power. Because the physics prevent radioactive release even in the event of severe natural disasters like floods, fires, or hurricanes, the regulatory Emergency Planning Zone (EPZ) is confined strictly to the powerhouse's 3-acre site boundary, in contrast to the 10-mile radius mandated around legacy gigawatt-scale plants.

- **EBR-II operational benchmark:** Operated successfully for 30 years (1964–1994) at Idaho National Laboratory. - **Track record:** Backed by >400 aggregate reactor-years of liquid metal reactor operation across ~25 facilities globally. - **Passive safety mechanisms:** Meltdown-proof physics driven by thermal expansion, natural convection airflow, and gravity. - **Emergency Planning Zone (EPZ):** Confined to the 3-acre property line, down from the traditional 10-mile evacuation zone.

---

[17:42](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1062s) Fuel Recycling and Commercial Deployment Milestones

The United States currently sits on an extraordinary domestic energy reserve in the form of commercial "used" nuclear fuel. Because traditional light-water reactors extract less than 10% of the usable energy contained in uranium fuel assemblies, more than 90% remains trapped inside the discharged material. By deploying fast reactors capable of recycling this inventory, the United States could power the entire nation for 150 years without mining a single additional ounce of uranium out of the ground. Recent federal policy initiatives have shifted heavily toward facilitating commercial fuel recycling and unlocking existing government stockpiles of high-assay material to accelerate advanced reactor rollouts.

Oklo is targeting commercial operation of its first commercial plant in 2028. The company is submitting its license application to the NRC in the fall, initiating a sequence of parallel workstreams designed to compress overall delivery timelines. Rather than waiting for final regulatory signoff before breaking ground, Oklo is running early site development, civil preparation, and long-lead equipment procurement alongside the licensing review, paving the way for widespread commercial scaling throughout the early 2030s.

- **Fuel energy retention:** Over 90% of original energy remains in conventional used nuclear fuel. - **US power potential:** Existing stockpiles of used fuel could supply total US electricity needs for ~150 years without new uranium mining. - **Deployment timeline:** Submitting NRC license application in the fall, targeting first commercial reactor deployment in 2028. - **Parallel execution:** Running long-lead component procurement and site development concurrently with NRC regulatory review to scale in the early 2030s.

---

[19:21](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1161s) Q&A: Hyperscaler Urgency and the Grid Capacity Crunch

During the open Q&A session, an attendee asked how Oklo bridges the tension between advanced nuclear lead times—such as licensing uncertainty, commercial readiness, and multi-year construction schedules—and the immediate, mission-critical appetite of data center operators whose power demands are exploding right now. The questioner noted the apparent paradox of asking hyperscalers to sign long-term Power Purchase Agreements (PPAs) while facing near-term delivery pressures.

Brian Gitt responded by emphasizing that speed to market is indeed the single most acute bottleneck confronting the data center industry today. When he first pitched small modular reactors to hyperscalers three years prior, their universal reaction was dismissive: "You guys are too far out for us. Come back in a few years when you're close." However, that dynamic shifted dramatically about 18 months ago. Hyperscale operators suddenly realized that regional electric grids across the United States were hitting hard transmission and generation ceilings, exhausting all available spare capacity. That stark realization forced data center infrastructure planners to discard short-term thinking and proactively lock in multi-gigawatt energy solutions years in advance.

[20:36](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1236s) The AI Power Crunch, Longer Planning Horizons, and Natural Gas Bridging

Brian Gitt explains how the rapid expansion of digital infrastructure has forced data center operators into a "huge wakeup call." Traditionally, planners only looked out two to three years because electric utilities typically held surplus capacity and could readily fulfill interconnection contracts. Today, utilities across the country have essentially run out of available power, rendering two-to-three-year horizons obsolete. Compared to just 18 months earlier, data center operators are actively engaged in long-term commercial discussions and power planning.

```mermaid flowchart LR accTitle: Power Transition Model accDescription: Flow showing rapid deployment of natural gas leading into long-term baseload nuclear power. subgraph NearTerm["Phase 1: Speed to Power (<12 Mos)"] G["Liberty Energy Natural Gas"] --> DC["Energize Data Centers Early"] end subgraph LongTerm["Phase 2: Nuclear Baseload Transition"] N["Oklo Nuclear Powerhouses"] --> BL["Primary Baseload Power (24/7)"] G -.-> LF["Gas Reallocated to Load Following & Backup"] end ```

To solve the immediate "speed to power" issue, Oklo has partnered with natural gas providers like Liberty Energy. Liberty can deploy gas generation within 12 months depending on the required scale, allowing data center buildings to energize immediately. As Oklo's nuclear powerhouses come online, the primary load shifts over to the nuclear plants, while the natural gas assets transition into load following, grid balancing, and backup services—ensuring the gas equipment avoids becoming stranded assets while maintaining a long, viable commercial life.

- **Legacy Planning**: 2–3 year horizons with simple utility off-take agreements. - **Gas Bridge Timeline**: Gas generation deployed within 12 months (via Liberty Energy partnership). - **Long-term Asset Mix**: Nuclear handles primary baseload; natural gas handles load balancing and peaking support.

---

[22:13](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1333s) Licensing Accelerators and the Idaho Commercial Plant

During the Q&A session, an audience member asks whether Oklo’s planned installation at the Idaho National Laboratory (INL) is a prototype, pilot, or a fully functioning commercial plant. Gitt clarifies that the Idaho facility is a fully commercial plant rather than a pilot or proof of concept. The underlying sodium-cooled fast reactor technology already has a 30-year operational track record from the Experimental Breeder Reactor II (EBR-II) at INL, allowing Oklo to draw directly from decades of empirical test data for its Nuclear Regulatory Commission (NRC) licensing applications.

Gitt notes that the regulatory landscape is shifting rapidly. Four recent executive orders from the White House directed the NRC to review advanced reactor license applications within an 18-month timeline. Competitor TerraPower recently had its application docketed under an 18-month review schedule, establishing a new licensing baseline. By running engineering, site preparations, and manufacturing in parallel with this 18-month review, Oklo aims to maintain an aggressive overall deployment schedule.

- **Idaho Facility Status**: Fully commercial power plant, not a pilot project. - **Historical Data**: Backed by 30 years of operational and safety data from EBR-II testing. - **Regulatory Benchmark**: NRC review target of 18 months per White House executive orders (with TerraPower serving as a recent precedent).

---

[23:42](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1422s) Nuclear Fuel Recycling and Government Down-Blending

Responding to an audience inquiry regarding how Oklo plans to utilize nuclear fuel stored at existing reactor sites, Gitt reveals that their first commercial plant will directly leverage recycled fuel derived from the historic 30-year demonstration at INL. Furthermore, Oklo conducted a nationwide site evaluation and plans to announce the location of a dedicated, full-scale commercial fuel recycling facility later in the year. This facility will convert spent fuel from the existing commercial light-water reactor fleet into usable fuel for advanced fast reactors.

In parallel with used fuel recycling, Oklo is pursuing fuel procurement through government down-blending programs. The United States government maintains substantial stockpiles of highly enriched material that can be down-blended into High-Assay Low-Enriched Uranium (HALEU) to power advanced reactor designs.

- **First Plant Fuel Source**: Recycled fuel from historical fast reactor operations. - **Dedicated Recycling Facility**: Location chosen after a nationwide search; announcement slated for later in the year. - **Down-Blending Pipeline**: Accessing federal stockpiles of legacy nuclear material blended down to reactor-grade fuel.

---

[24:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1491s) Grid Impacts: Behind-the-Meter Microgrids vs. "Good Grid Citizens"

Asked how captive data center power solutions will affect the broader electrical grid over the long term, Gitt predicts a hybrid model between "behind the meter" microgrids and reformed "in front of the meter" utility arrangements. To prevent localized public backlash and power bill inflation, hyper-scale facilities (such as the 11-gigawatt project mentioned earlier in the talk) will increasingly deploy captive microgrids, essentially bringing their own generation off-grid.

```mermaid flowchart TD accTitle: Future Data Center Power Integration Modes accDescription: Comparison between Behind-the-Meter Microgrids and In-Front-of-the-Meter Utility Frameworks. DC["Data Center Power Strategy"]

subgraph BTM["Behind-the-Meter (BTM)"] DC --> B1["Dedicated Microgrids (e.g., 11 GW Campus)"] B1 --> B2["Overbuilt Capacity for Town Emergency Backup"] end

subgraph FTM["In-Front-of-the-Meter (FTM)"] DC --> F1["Reformed Utility Tariffs"] F1 --> F2["Shields Residential & Small Business Rates"] end ```

Simultaneously, Gitt highlights how data centers can act as "good grid citizens." Rather than solely contributing to local parks or community funds, operators can intentionally overbuild their generation capacity to supply surplus power to municipal services or offer emergency grid resiliency during regional blackouts. On the regulated utility side, new tariff structures and commercial constructs are being designed so that massive infrastructure upgrades are directly funded by the data centers rather than being passed along to residential homeowners or small businesses.

- **Public Rate Protection**: "It's politically infeasible" for utilities to pass large data center infrastructure costs onto regular rate payers without causing public pushback. - **Grid Citizen Opportunities**: Overbuilding on-site power to supply emergency reserves and municipal power during outages. - **Dual Structural Future**: Parallel development of behind-the-meter captive microgrids and redesigned front-of-the-meter commercial tariffs.

## Notable quotes > The arc of human progress has always moved away from less concentrated sources of energy toward more concentrated ones—from wood to coal, to oil and gas, to nuclear power. — Brian Gitt

> Small plants are walk-away safe and cool themselves without relying on people, pumps, or mechanical systems. They can also recycle used fuel and convert nuclear waste into clean energy. — Brian Gitt

> [19:21](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1161s) You guys are too far out for us. Come back in a few years when you're close. — Hyperscalers (quoted by Brian Gitt)

> [24:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1491s) It's politically infeasible — Brian Gitt

## Mentioned Ai4, Amarillo, Texas, Argonne National Laboratory, Brian Gitt, Equinix, Experimental Breeder Reactor II (EBR-II), Idaho National Laboratory (INL), Liberty Energy, Meta, Nuclear Regulatory Commission (NRC), Oklo Inc., OpenAI, Siemens Energy, Switch, TerraPower, United States Department of Defense, Vertiv, Vogtle expansion

## Video description Brian Gitt, SVP, Head of Business Development, Oklo Inc

Abstracts

The arc of human progress has always moved away from less concentrated sources of energy toward more concentrated ones—from wood to coal, to oil and gas, to nuclear power. Future progress moves in the same direction: 24/7 clean nuclear power that unlocks the benefits of AI and cloud computing. Historically, there have been three big challenges to nuclear: 1. High upfront costs: ~$10+ billion cost of construction 2. Long build time: 7-10 years to deploy 3. Public perception: concerns about safety and nuclear waste Small nuclear power plants can overcome each of these challenges. They’re simpler and less expensive to build. Factory prefabrication combined with minimal civil works enables plants to deploy in 3 years. Small plants are walk-away safe and cool themselves without relying on people, pumps, or mechanical systems. They can also recycle used fuel and convert nuclear waste into clean energy.

[Source](https://www.youtube.com/watch?v=o_SM4BJkf2w)

中文简报

AI数据中心面临电力短缺，Oklo推进14吉瓦反应堆项目储备 不堪重负的电网正迫使科技超大规模企业放弃短期公用事业规划，转向锁定模块化液态金属核反应堆，为庞大的AI集群提供动力。 Brian Gitt解释了数据中心激增的电力需求如何与僵化的电网容量限制发生冲突，从而催生了对专用“表后”（behind-the-meter）清洁能源解决方案的需求。Oklo通过“电力即服务”（Power-as-a-Service）模式化解传统核能面临的阻碍，部署具备非能动安全、极低水耗及乏燃料回收能力的75兆瓦模块化液态金属快堆。为解决当前急迫的通电速度瓶颈，Oklo整合了短期天然气过渡资产，同时在其14吉瓦的项目管线中持续推进商业核电部署与燃料回收设施建设。 受得克萨斯州阿马里洛 11 吉瓦表后微网等巨型项目驱动，数据中心电力需求预计将在 2030 年增长两倍，促使科技巨头迫切寻求清洁可靠的基荷电力。 (0:04) 传统核电长期受困于高达 90 亿至 100 亿美元的前期成本以及 8 至 10 年的漫长建设周期；Oklo 采用“电力即服务”与 20 年购电协议（PPA）模式消除客户资本支出，并将交付周期压缩至 36 个月。 (1:41) Oklo 采用液态金属冷却技术替代高压水循环，单台 75 兆瓦模块化机组仅占地 3 英亩且摆脱了对大型水体的依赖，同时具备利用现有乏核燃料发电的闭环能力。 (3:27) 公司已积累 14 吉瓦商业客户储备管线并与 Equinix 和 Switch 达成协议，同时携手 Liberty Energy 部署 12 个月内可落地的天然气桥接发电，目标在 2028 年实现首座商业核电厂投产。 (6:51)

Reading summary

Brian Gitt, Senior Vice President at Oklo Inc., argues that surging artificial intelligence workloads are exhausting regional electrical grid capacity, necessitating a shift toward modular fast-fission nuclear reactors that deliver dedicated, water-efficient baseload power under long-term service contracts. AI Infrastructure Growth and Regional Grid Ceilings (0:04) Digital infrastructure expansion is forcing data center developers to confront severe physical limitations on regional power grids. With industry projections forecasting a tripling of data center electricity consumption by 2030, planned commitments have reached unprecedented scales, including OpenAI's proposed 5-gigawatt campuses and an 11-gigawatt project planned in Amarillo, Texas. Gitt notes that this single Amarillo development exceeds the total electric generation capacity of the state of Nevada and surpasses the peak electricity demand of New York City. Because regional utilities have largely exhausted surplus generation and transmission margins, hyperscale operators can no longer rely on traditional two-to-three-year utility interconnection schedules. Consequently, developers are pursuing dedicated behind-the-meter microgrids to generate power on-site without drawing from the broader grid. Gitt observes that captive microgrids shield local residential ratepayers from the expense of major utility infrastructure upgrades, while creating opportunities to supply emergency backup power to municipal services during regional blackouts. Modular Deployment and the Power-as-a-Service Model (3:27) To bypass the historic structural barriers that have stagnated domestic commercial nuclear construction for three decades—specifically capital costs of $9 billion to $10 billion per 1,000-megawatt plant and 8-to-10-year construction timelines—Oklo developed a modular 75-megawatt powerhouse design. Gitt highlights that hyperscalers typically expand campuses in phased 100-megawatt halls rather than building multi-gigawatt facilities all at once. Modular units allow power generation to scale in lockstep with compute capacity, preventing massive capital outlays on underutilized, idle generation. Oklo structures its business around a Power-as-a-Service model, financing plants directly and selling electricity through 20-year Power Purchase Agreements (PPAs) so technology clients avoid adding infrastructure debt to their balance sheets. Operating multiple 75-megawatt reactors also resolves the availability risks inherent in large 1,000-megawatt plants: individual modular units can be sequentially cycled offline for refueling and routine maintenance while remaining units maintain uninterrupted baseload delivery. According to Gitt, this framework underpins a 14-gigawatt commercial customer pipeline that includes data center operators Equinix and Switch. Cooling Architecture and Standardized Balance of Plant (10:34) Oklo replaces conventional pressurized-water cooling loops with liquid metal, which possesses higher thermal conductivity and circulates at atmospheric pressure inside conventional 304 or 316 stainless steel vessels. This unpressurized design, paired with air-cooled condensers, eliminates the requirement to site plants near oceans, rivers, or lakes. Gitt states that operational water consumption is confined to closed-loop turbine makeup water, making its footprint comparable to the water required to wash solar panels and far lower than conventional thermal power stations. To control capital costs and avoid specialized nuclear supply chains, roughly 70 percent of the powerhouse uses standard commercial hardware. The thermal-to-electric conversion system uses an off-the-shelf industrial steam turbine and generator package from Siemens Energy, an architecture deployed across approximately 5,000 global installations. Operating at roughly 450°C, the system also produces high-grade thermal energy; Oklo has partnered with Vertiv to route steam directly through absorption chillers, enabling the installation to cool data center server halls while simultaneously providing electricity. Regulatory Milestones, Fuel Recycling, and Gas Bridging (15:51) Oklo’s reactor technology commercializes fast-reactor systems developed at Idaho National Laboratory, drawing directly from 30 years of operational and safety data from Experimental Breeder Reactor II (EBR-II). During historical safety tests, EBR-II demonstrated passive shutdown mechanisms: thermal expansion, natural convection, and gravity cooled the reactor without human intervention or auxiliary power. This passive design confines the regulatory Emergency Planning Zone to the 3-acre site boundary. Oklo plans to submit its commercial license application to the Nuclear Regulatory Commission (NRC) under a targeted 18-month review timeline, aiming for initial commercial operations in Idaho by 2028 while conducting long-lead procurement in parallel. The commercial plant plans to use recycled fuel from historical demonstrations, supplemented by federal programs that down-blend enriched government stockpiles into reactor fuel. Gitt highlights that tapping the remaining 90-plus percent of energy stored in existing domestic spent fuel stockpiles could theoretically power the United States for 150 years without new mining. To address near-term deficits before reactors come online, Oklo has partnered with Liberty Energy to deploy natural gas generation within 12 months, intending to transition those gas assets into backup and load-following roles once the nuclear units deliver baseload power. Oklo’s deployment timeline depends on achieving an accelerated 18-month NRC licensing review, executing concurrent equipment procurement, and establishing a dedicated commercial-scale fuel recycling facility. While modular fast reactors present a high-density, water-efficient baseload solution, hyperscalers face an immediate reliance on temporary natural gas bridging to manage rapid AI workload expansion over the next several years.

精炼阅读摘要

Oklo公司业务拓展高级副总裁Brian Gitt指出，人工智能算力激增正使美国电网逼近容量上限，而传统核电的高昂造价与漫长工期无法满足即时需求；Oklo主张通过小型模块化快堆、标准工业供应链与电力即服务模式，为数据中心提供快速部署的离网高密度清洁基荷。 算力电力缺口与传统核电瓶颈 (0:04) Gitt表示，行业共识预测数据中心电力需求到2030年将增长两倍。例如德克萨斯州阿马里洛规划的11吉瓦数据中心园区，其单体规模已超过内华达州全州发电容量及纽约市夏季用电高峰；同时OpenAI也在规划高达5吉瓦的超大型园区。由于区域公用电网面临输电和发电容量枯竭，超大规模运营商普遍转向“表后”（behind the meter）独立供电方案。然而，尽管美国现役94座商用反应堆供应了全美20%的电力且保持了70年的安全运行纪录，但国内传统核电行业已停滞近30年。 Gitt指出传统核电存在三大结构性障碍：单座1000兆瓦常规核电站前期投资高达90亿至100亿美元；建设周期长达8至10年，在产生收益前沉淀巨额资本；公众对堆芯熔融与放射性废料处置的长期安全顾虑，依然阻碍着传统大型核电的规模化推广。 小型模块化架构与电力即服务 (3:27) 为消除前期资本开支壁垒，Oklo采用“电力即服务”（Power-as-a-Service）商业模式，作为独立发电商承担建设与运营成本，客户只需签署为期20年的购电协议（PPA），无需将重资产核设施计入自身资产负债表。针对数据中心按机房模块分期扩建（通常每期约100兆瓦）的特点，Oklo将单机容量设计为75兆瓦，使电力供应与算力扩容节奏精确匹配，避免大型核电站初期数吉瓦产能闲置导致的资金浪费，并可通过多机组交替轮换维护确保持续不间断供电。 在交付速度与空间占用上，Oklo将传统核电近十年的建设周期大幅压缩至36个月（其中现场施工约18个月），单台75兆瓦机组至安防围栏占地仅需3英亩。截至演讲时，Oklo已储备达14吉瓦的商业客户合作意向，涵盖Equinix与Switch等数据中心运营商。 液态金属冷却与成熟工业级供应链 (10:34) 在反应堆工程上，Oklo采用液态金属冷却剂与空冷冷凝器取代传统高压水冷回路，彻底摆脱了电站对河流或海洋等外部水源的依赖，厂区耗水量仅限于二次蒸汽回路的极少量补给水。在土地效率对比中，500兆瓦的Oklo模块化阵列仅需约21英亩土地即可实现全年全天候基荷发电，而亚利桑那州同等标称功率的光伏电站则需约2700英亩且发电具有间歇性。 为降低系统建造成本，Oklo反应堆约70%的组件采购自成熟的通用工业供应链，堆外发电系统直接采用西门子能源全球已部署约5000台的标准蒸汽轮发电机组，堆容器亦采用石化与制药行业通用规格的非承压不锈钢罐体。堆芯工作温度约450°C，产生的高温蒸汽还可配合Vertiv吸收式制冷机直接为机房供冷。该设计继承了爱达荷国家实验室二号实验快堆（EBR-II）长达30年的运行数据，依托热膨胀、自然对流和重力实现固有被动安全，将其应急规划区（EPZ）范围严格收缩在3英亩厂区红线之内。 乏燃料循环、过渡供电与电网共生 (17:42) 针对核燃料来源，Gitt指出传统轻水堆仅释放了铀燃料不到10%的能量，现有商用乏燃料蕴藏90%以上可用潜能。Oklo快堆能够回收利用这类核废料发电，现有库存理论上足以满足美国全国约150年的用电需求；公司首座商业电站将直接使用历史快堆乏燃料，并计划在年内公布专用商用燃料回收设施选址，同时利用联邦政府库存的高浓度低浓缩铀（HALEU）稀释料补给燃料。 面对当前数据中心数月内的用电紧迫性与核电交付周期的落差，Oklo联合Liberty Energy提供天然气发电过渡方案，可在12个月内为数据中心供电；待核电机组并网后，天然气机组转为调峰和应急备用，避免资产闲置。此外，对于公用电网承受力问题，自建表后微电网既能避免将电网升级成本转嫁给居民用电账户，又能在区域电网发生故障时反向提供应急备用电力，实现良性共生。 Oklo通过技术降维与商业重构为AI算力基础设施提供了高密度的能源替代路径，其首座商业电站计划于2028年投入运营，并受到白宫推动NRC执行18个月许可审查周期的政策支持。然而，该路线图的最终兑现仍取决于商用许可证审批进度、自主燃料回收产业链的落地周期，以及超大型表后微电网与区域公用电网监管体制的长远博弈。

Detailed notes — zh

## Background 本内容由 Ai4 于 2025 年 8 月 26 日发布在 YouTube 平台。页面简介显示，演讲者为 Oklo 业务发展高级副总裁 Brian Gitt，主题聚焦于如何利用紧凑、清洁且全天候稳定运行的小型先进核电站，攻克传统核电高成本、长周期与安全顾虑等瓶颈，从而化解 AI 和云计算爆发带来的电力危机。 ## TL;DR 面对德克萨斯州 11 GW 级超级数据中心园区及 2030 年全美算力用电激增三倍的严峻挑战，Oklo 提出了 75 MW 液态金属冷却小型模块化反应堆（SMR）解决方案。该方案采用“电力即服务（Power as a Service）”模式自建自营，通过 20 年 PPA 免除客户百亿美元的前期 CAPEX，并以 3 年极速部署、模块化步进扩容、零水耗堆芯及回收 90% 乏燃料等技术优势打破传统核电瓶颈。同时，Oklo 联合 Liberty Energy 推出 12 个月内天然气快速供电并平滑过渡至零碳核能的组合策略，结合表后微电网与电网共生机制，为 AI 算力集群提供确定性基荷保障。 ## Atmosphere & style 整场演讲充满前瞻性与扎实的工业落地底气，Brian Gitt 以沉着自信、数据驱动的第一性原理视角切入，将硬核核工程与炙手可热的 AI 算力基建巧妙串联。叙事节奏紧凑有力，从用电海啸的巨大危机感自然过渡到成熟工程解法带来的笃定，现场问答环节更展现出直面行业交期痛点与监管现实的坦诚务实风格。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=4s) AI 算力扩张带来空前电力需求，单体数据中心园区用电规模已高达 11 GW（超过纽约市全峰值负荷），全美数据中心电力需求预计在 2030 年前激增至现有水平的三倍。 - [01:38](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=98s) 传统大型核电面临前期资本高（单堆 90-100 亿美元）、建设工期长（8-10 年）以及公众安全/废料顾虑三大核心痛点，导致全美核电扩张停滞近 30 年。 - [03:22](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=202s) Oklo 采用“电力即服务”（Power as a Service）商业模式自建自持电厂，通过 20 年期 PPA 协议实现项目融资，免除客户高额的前期 CAPEX 负担。 - [03:22](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=202s) Oklo 采用 75 MW 小型模块化反应堆（SMR），仅占地 3 英亩，部署周期缩短至 36 个月以内，且彻底摒弃水冷、采用液态金属冷却实现零水耗堆芯。 - [06:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=411s) 75 MW 的离散化小堆与数据中心单栋约 100 MW 的分批上电节奏高度同步，并支持梯队式轮换停机维护，杜绝单点停机断电，实现高冗余 24/7 运行。 - [08:38](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=518s) 核能拥有极高能量密度，500 MW 集群仅占地约 21 英亩，远低于同等容量地面光伏电站所需的约 2,700 英亩，且全生命周期物料和占地足迹极小。 - [10:34](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=634s) Oklo 电厂约 70% 零部件来自常规工业成熟供应链（如与西门子能源合作汽轮发电机组），常压不锈钢罐体大幅降低了非标核级定制的高昂成本。 - [12:40](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=760s) 反应堆在 450°C 高温下运行，副产高温蒸汽不仅提高发电效率，还可驱动 Vertiv 吸收式制冷机组为数据中心直接降温或工业供热。 - [15:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=951s) 采用传承自爱达荷国家实验室（EBR-II 运行 30 年）的成熟液态金属堆设计，具备物理法则保障的固有被动安全（Walk-away Safe），应急规划区仅限于 3 英亩厂界之内。 - [17:42](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1062s) 全美现有核废料（乏燃料）中仍蕴含超过 90% 未释放能量，循环再处理后足以支撑美国未来 150 年用电，Oklo 正在筹建首座商业化核燃料回收工厂并推进高浓铀稀释利用。 - [20:36](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1236s) 为缓解数据中心即刻通电的燃眉之急，Oklo 联合 Liberty Energy 推出 12 个月内快速部署天然气发电的过渡方案，待核电落成后天然气转为调峰和备用电源。 - [24:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1491s) 数据中心未来将呈现“表后（Behind-the-Meter）独立微电网”与“表前电网协同”的共生格局，在满足自身算力的同时避免转嫁居民电费并可作为社区应急备用电源。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=4s) AI 时代的用电海啸：从 11 吉瓦园区到三倍用电增长 会议进入尾声，Oklo 业务发展高级副总裁 Brian Gitt 登台发表了压轴演讲。他开门见山地指出，Oklo 的核心业务是为现代数据中心和新型 AI 工作负载设计、建造并运营先进核电站。为了让在场听众直观感受当前算力扩张对能源带来的空前压力，Brian Gitt 分享了演讲前一个月在德克萨斯州阿马里洛（Amarillo, Texas）刚公布的一个 11 GW（吉瓦）超大规模数据中心园区项目。 为了建立清晰的体量概念，他给出了两个极具冲击力的参照：整个内华达州（包括拉斯维加斯在内）的所有发电装机容量总和尚不足 11 GW；全美最大都市纽约市在用电最高峰时的总负荷同样低于 11 GW。而阿马里洛这处规划在“表后”（behind-the-meter）的单一数据中心园区，其电力需求就达到了这一惊人水平。这正是整个行业未来的发展轨迹——OpenAI 正在筹划高达 5 GW 的超级园区，Meta 也在不断抛出庞大的扩建计划。多项行业权威报告一致预测，到 2030 年前后，全美数据中心的电力需求将激增至现有的三倍。面对这一严峻缺口，如何以极快速度开发出稳定、清洁的超大容量电力，成为 AI 产业发展的首要命题。 - 德克萨斯州阿马里洛 11 GW 数据中心园区：单一园区用电规模超过内华达州全州发电装机，亦超过纽约市全峰值负荷。 - 巨头扩张态势：OpenAI 正在推进最高 5 GW 级的超级计算园区，Meta 等科技巨头算力需求同步暴涨。 - 行业负荷预测：数据中心电力总需求将在未来数年内快速翻倍，预计在 2030 年前达到当前水平的三倍。 --- ### [01:38](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=98s) 传统核电的历史基石与制约行业三十年的三大瓶颈 Brian Gitt 回顾了核能在美国能源结构中的历史地位。目前全美共有 94 座在运商用核反应堆，为美国贡献了约 20% 的电力供应。在过去 70 年中，这些机组持续稳定运转，提供了巨大的基荷支撑。然而，整个核电行业在过去长达 30 年的时间里陷入了近乎停滞的泥潭——数十年来，全美仅在佐治亚州（Georgia，即 Vogtle 核电站）新投产了两台机组。 传统核电之所以长期无法规模化扩张，根源在于行业面临的三大核心痛点： 1. **前期资金门槛过高**：以近期投运的 1,000 MW 大型核电站为例，单座造价高达 90 亿至 100 亿美元，属于典型的超大型重资产基建项目。 2. **建设周期漫长**：从立项到投产往往耗时 8 到 10 年，长周期的资本占用极大增加了投资风险。 3. **公众对安全与核废料的担忧**：尽管 Brian Gitt 强调核电是全美最安全的工业之一——在过去 70 年商业运营中仅发生过一次无人员伤亡的事故——但公众对堆芯事故及乏燃料处理的疑虑依然构成巨大的社会阻力。 ```mermaid flowchart TD accTitle: 传统核电的三大核心瓶颈 subgraph TraditionalProblems["传统大型核电面临的瓶颈 (1,000 MW 级)"] A["超高初始投资<br/>(单堆成本 $90-100 亿美元)"] B["超长建设周期<br/>(工期长达 8-10 年)"] C["公众安全与废料顾虑<br/>(水冷事故隐患与乏燃料堆积)"] end subgraph IndustryResult["行业后果"] D["产业停滞 30 年<br/>(数十年来全美仅投产 2 台新机组)"] end A --> D B --> D C --> D ``` - 传统大型核电参数：单机容量约 1,000 MW，占地约 800 英亩，单座成本 90-100 亿美元，建设周期 8-10 年。 - 传统冷却依赖：必须临近大型地表水源（湖泊、河流、水库或海洋）以获取巨量冷却水。 - 行业安全事实：商用核电运营 70 年来仅发生过一起事故，且未造成任何人员伤亡，但公众心智中的安全与废料包袱依然沉重。 --- ### [03:22](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=202s) Oklo 的破局之道：技术重构、商业模式创新与小型模块化 为彻底打破传统核电的桎梏，Oklo 围绕“技术创新、商业模式重构、机组尺寸小型化”三大维度推出了全新解决方案。Oklo 建造的并非带有庞大混凝土冷却塔的传统超级工程，其紧凑外观与环境兼容性甚至可以无缝嵌入郊区环境或整体规划社区。 在商业模式上，数据中心运营商绝不可能为电力基础设施垫付 100 亿美元的前期资本。因此，Oklo 采取了“电力即服务”（Power as a Service）模式，以新一代独立发电商（IPP）的身份自建、自持并运营电厂，用户仅需按千瓦时（kWh）购买电力。Oklo 与超大规模云厂商（Hyperscalers）或大型托管服务商签署为期 20 年的长期购电协议（PPA），并基于该长期合约通过债权与股权组合完成项目融资，免去了客户的高额初始 CAPEX。 在工程与技术方面，Oklo 采用 75 MW 的小型模块化反应堆（SMR）： - **极小占地面积**：单座 75 MW 机组包含所有围栏及配套设施仅占地 3 英亩，可像搭积木一样并排扩建。 - **大幅缩短周期**：全周期部署时间可缩减至 36 个月以内（其中实际现场建造工期仅约 18 个月，其余为主设备长周期采购与审批许可时间）。 - **液态金属冷却**：彻底摒弃水冷却方案，采用热传导效率最高的液态金属介质，完全不依赖自然水源，可在干旱地区灵活选址。 - **乏燃料循环再利用**：全美现有核电站储存的乏燃料堆中，超过 90% 的残留能量可被 Oklo 反应堆回收提纯并转化为新燃料。 ```mermaid flowchart LR accTitle: Oklo 商业模式与技术交付架构 subgraph CommercialModel["商业模式 (Power as a Service)"] direction TB PPA["20 年期 PPA 长期购电协议"] Finance["项目融资 (债务 + 股权)"] PPA --> Finance end subgraph ReactorTech["技术优势 (75 MW 模块化微堆)"] direction TB Site["极小占地 (3 英亩/台)"] Speed["极速部署 (36 个月全周期)"] Cooling["液态金属冷却 (零水耗堆芯)"] Recycle["90% 乏燃料回收利用"] Site ~~~ Speed ~~~ Cooling ~~~ Recycle end Finance --> ReactorTech ReactorTech --> Customer["为 AI/数据中心提供 24/7 确定性电力"] ``` --- ### [06:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=411s) 小堆的扩展性优势：建设节奏同步与高冗余度保障 针对当前动辄 500 MW、1 GW 甚至更高容量的数据中心集群，Brian Gitt 指出，数据中心在现实中绝非一次性建成的，而是以单栋机房（约 100 MW）为单位，分阶段在数年内逐步落成。传统大电站一次性并网 1,000 MW 会导致巨大的初期电力闲置，使资本效率极低。而 Oklo 采用 75 MW 的离散化小堆分批并网，使发电侧的扩容节奏与数据中心的算力交付完全“同步”（Synchronization），消除了过度建设带来的经济浪费。 除了容量同步，小堆化带来的另一核心优势是极高的系统级冗余度（Redundancy）。传统大型电站（如 500 MW 或 1,000 MW 堆）在停机更换燃料或进行大修维护时，会造成大面积电力中断，对算力负荷而言是不可承受的；而采用多台 75 MW 机组组成的集群，运营方可以轮流对单台机组进行离线换料与检修，其余机组无缝接替并保持满载，实现真正意义上的 24/7 不间断运行。 - 算力建设同步性：按 75 MW 步长分期投产，完美契合数据中心单栋约 100 MW 的分批上电需求。 - 维护与换料冗余：小机组支持梯队式轮换维护，杜绝单点停机导致整个园区断电。 - 商业储备管线：Oklo 目前已建立了高达 14 GW 的项目储备管线，并已与 Equinix、Switch 等头部数据中心运营商推进商业合作。 --- ### [08:38](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=518s) 第一性原理下的能源密度：材料利用、占地与水资源对比 从第一性原理（First Principles）审视能源形式，核能的本质优势在于无与伦比的“能量密度”（Energy Density）。Brian Gitt 列举了一个形象的比喻：一个人一生中所需消耗的全部能源，如果折算为核燃料材料，其体积仅仅相当于一颗普通的高尔夫球。从全生命周期环境评估来看，如此极端的能量富集度意味着核能在前端开采量、工程材料消耗、运输物流以及用地足迹上远低于其他任何清洁能源。 在土地与水资源占用方面，Brian Gitt 给出了具体的量化对比： - **土地占用对比**：以光照条件优异的亚利桑那州（Arizona）为例，若建设 500 MW 的发电设施，Oklo 核电机组集群仅需占用约 21 英亩土地；相比之下，同等容量的地面光伏电站需占地约 2,700 英亩，且光伏只能在白天发电，无法提供全天候连续基荷。 - **水资源消耗对比**：在美国西部等干旱地区（如拉斯维加斯周边），水资源极为稀缺。Oklo 堆芯完全采用液态金属传热，全厂唯一消耗水的部分仅为二次回路蒸汽轮机所需的极微量工质补给水（makeup water），在极大程度上减轻了区域水资源承载压力。 | 对比维度 | 传统大型核电站 (1,000 MW) | 集中式地面光伏 (500 MW，亚利桑那州) | Oklo 小型模块化堆 (75 MW 模块集群) | |:--- |:--- |:--- |:--- | | **占地面积** | 约 800 英亩 | 约 2,700 英亩 | 约 21 英亩 (以 500 MW 集群计) | | **全周期建设工期** | 8 - 10 年 | 2 - 4 年 (含外送配套) | 约 36 个月 (现场施工约 18 个月) | | **初期投资模式** | 需客户自筹 ~$100 亿美元 | CAPEX 自建或公用事业接入 | 电力即服务 (PPA)，客户零前期重资产 | | **供电特性** | 24/7 基荷 (单堆停机影响大) | 间歇性 (仅白昼发电) | 24/7 基荷 (多机组轮流换料，高冗余) | | **水资源依赖** | 极度依赖大型天然水体冷却 | 清洗组件微量用水 | 仅需汽轮机极微量补给水 (堆芯零水耗) | ### [10:34](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=634s) 极低耗水量与成熟供应链的降本优势 核电在本质上就是一种热源，通过将水加热沸腾产生蒸汽，进而驱动汽轮机发电，这与常规火电厂的运行机制完全一致。Oklo 的电厂设计不需要庞大的冷却水消耗，唯一用水仅为供给蒸汽轮机的补给水。这种耗水量极低，几乎与太阳能发电持平。许多人可能会疑惑太阳能面板为何需要水，事实上，由于空气中大量的灰尘和污垢会附着在面板表面，需要定期耗费大量水资源进行清洗；Oklo 的用水量就与这种清洗耗水基本相当，远低于传统大型核电站或天然气联合循环电厂等高耗水热电设施。 ```mermaid flowchart LR accTitle: Oklo 电厂热能与发电系统架构 subgraph Core["反应堆核心系统 (常压容器)"] Fuel["核燃料裂变产热"] --> LiquidMetal["液态金属冷却剂 (450°C)"] LiquidMetal --> HeatExchanger["热交换器"] end subgraph PowerGen["常规动力岛 (Siemens Energy 供应链)"] HeatExchanger --> SteamGen["蒸汽发生器"] SteamGen --> SteamTurbine["西门子蒸汽轮机与发电机"] SteamTurbine --> Electricity["电力输出 (75MW+)"] end subgraph HeatSink["非水冷散热与副产应用"] SteamGen -.->|"高温蒸汽"| AbsorptionChiller["Vertiv 吸收式制冷机组\n(数据中心冷却 / 工业供热)"] SteamTurbine --> AirCondenser["风冷冷凝器\n(大型风扇无水散热)"] end ``` Oklo 降低建设成本的关键策略在于最大化利用现有工业供应链：一座被称为“Oklo Powerhouse”的核电站中，约 70% 的零部件都来自常规成熟的供应链，无需专门进行严苛的核级认证。例如，负责将热能转化为电力的整套常规岛系统（Balance of Plant），包括蒸汽轮机与发电机，均与西门子能源（Siemens Energy）合作。西门子能源生产的这类成熟现货产品在全球已累计销售约 5,000 台，经过适当调校即可直接投入使用。全厂仅有核燃料及直接包裹核燃料的少数部件属于专用核级设备。在场地建设上，这种设计就像乐高积木一样具备高度模块化和可扩展性，单机功率可从 75 兆瓦（MW）起步，平滑扩展至 150 MW、300 MW、1,000 MW 乃至 5,000 MW，完美契合超大规模数字基础设施的扩容节奏。 --- ### [12:40](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=760s) 高温运行、多联产应用与常压容器设计 Oklo 反应堆的运行温度高达 450 摄氏度。高温运行不仅大幅提升了发电热效率，还带来了丰富的综合热能利用场景。除了发电之外，高温副产蒸汽可直接用于工业供热，甚至可用于数据中心的冷却系统。Oklo 最近与数字基础设施及冷却方案巨头 Vertiv 达成合作，直接利用反应堆产生的高温蒸汽驱动吸收式制冷机组（Absorption Chillers）为数据中心降温，实现电力供应与制冷协同输出。 与传统核电的高压反应堆不同，Oklo 反应堆采用常压设计。整个反应堆容器由一个大型不锈钢罐体构成，类似于石化或制药行业广泛使用的标准储存罐，尺寸大约为高 40 英尺、直径 20 英尺。该罐体完全处于非受压状态，材质为业内极为通用的 304 或 316 不锈钢，材料和制造工艺成熟规范，无需进行昂贵且复杂的非标定制。 --- ### [13:58](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=838s) 液态金属冷却、风冷散热与超高运行可靠性 反应堆的工作流程非常直观：核燃料置于大型不锈钢罐体底部，随着裂变反应展开持续释放热量；整个罐内注满液态金属冷却剂，液态金属吸收热量并上升流经热交换器，将热量传递至二回路生成蒸汽，最终驱动西门子能源的汽轮机发电。在余热导出方面，系统采用了风冷冷凝器（Air-Cooled Condensers），依靠大型风扇进行空气冷却，完全摆脱了对外部冷却水源的依赖。 在可靠性方面，核电长期以来一直是全美运行最为可靠的基础负荷电源，平均容量因子高达 92% 左右，除定期换料和日常维护外几乎全年不间断满发。得益于模块化的小型化设计，系统可以通过多机组冗余配置实现接近 100% 的有效可用率（Uptime）。这种极高的可靠性与弹性吸引了美国国防部的高度关注，Oklo 已收到向美国阿拉斯加州某空军基地部署微型反应堆的授标意向通知书（Notice of Intent to Award）。 - **运行容量因子**：传统核电机组约 92%，Oklo 凭借模块冗余可实现极高可用性 - **冷却方式**：全空气冷却冷凝，无须冷却塔或自然水源 - **国防应用**：获得阿拉斯加空军基地项目意向支持 --- ### [15:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=951s) 400 余年技术积累与固有被动安全机制 Oklo 采用的堆型并非未经测试的全新实验性概念，而是源自美国爱达荷国家实验室（INL）和阿贡国家实验室（ANL）数十年的技术沉淀。全球范围内已建造并运行过约 25 座同类设计堆型，技术运行历史累计超过 400 堆年（Reactor-Years）。Oklo 最直接商业化的技术蓝本在爱达荷国家实验室成功连续运行了 30 年（1964 年至 1994 年，即实验增殖堆二号 EBR-II）。该反应堆过去三十年积累的海量试验与运行数据，全部可直接用于支持 Oklo 提交给美国核能管理委员会（NRC）的许可证申请，因此公司无须耗资从头建造验证堆或中试工厂。 在历史上的极限安全测试中，科研人员曾尝试人为切断所有冷却系统、模拟最严苛的堆芯熔毁事故，而该堆型凭借物理法则实现了自我冷却与功率自调节（Walk-away Safe）。反应堆在发生故障时无需人工干预，无需备用应急电源启动水泵，也无需任何机械运动部件介入，仅依靠重力、热膨胀以及自然空气对流等自然法则即可导出余热。即便遭遇飓风、火灾、洪水等最严重的外部自然灾害导致电厂紧急停运，也绝对不会发生超出厂区范围的放射性泄漏。传统大型核电站的应急规划区（EPZ）通常覆盖方圆 10 英里，而 Oklo 电厂的应急规划区仅限于其占地约 3 英亩的电厂边界之内。 --- ### [17:42](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1062s) 乏燃料循环再利用与 2028 年商业化路线图 乏燃料（核废料）回收蕴藏着巨大的清洁能源潜力。目前商业核反应堆卸出的乏燃料中，有超过 90% 的能量仍未被释放。如果对美国现存的全部乏燃料进行再处理和回收利用，在无需新开采一盎司天然铀矿石的前提下，所蕴含的能量就足以支撑全美未来 150 年的用电需求。近期美国在政策层面已经出现积极转向，重点推进乏燃料循环利用并开放政府库存原料的使用权限。 按照 Oklo 的商业化推进时间表，公司计划于当年秋季向 NRC 提交联合许可申请，并在审评期间同步开展长周期设备采购、前期土地勘测与场地开发等并行工作。Oklo 规划在 2028 年正式投用首座商用微型核电站，并预计在 2030 年代初全面实现规模化部署。 ```mermaid flowchart LR accTitle: Oklo 商业化时间节点与并行推进路径 NRC["提交 NRC 许可申请\n(当年秋季)"] --> Parallel["多线并行筹备\n(长周期设备采购 / 场地前期开发)"] Parallel --> FirstPlant["首座商用电厂并网\n(2028 年)"] FirstPlant --> CommercialScale["全面规模化量产部署\n(2030 年代初)"] ``` --- ### [19:21](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1161s) 现场问答：数据中心的迫切电荒与算力巨头的态度转变 在演讲后的现场提问环节，有观众指出当前核能关注度极高，但针对 Oklo 约需数年的建设周期、NRC 监管许可的不确定性以及商业购电协议（PPA）的落地风险，询问作为业务拓展负责人，该如何平衡数据中心即刻爆发的电力需求与核能较长交付周期之间的矛盾。 Brian Gitt 回应称，上市交付速度（Speed to market）确实是当前数据中心与云计算行业面临的最大痛点，基础设施扩建的速度已难以跟上算力增长的步伐。他坦言，大约三年前他与各大超大规模数据中心服务商（Hyperscalers）洽谈时，对方普遍反应冷淡，直言："你们的投产时间太遥远了，等过几年接近落地时再来找我们吧。"然而，局势在过去的约 18 个月内发生了戏剧性的逆转。这些科技巨头猛然意识到全美各地的电网新增容量正被迅速耗尽，不得不将电力规划视线从短期转向更为长远的可靠清洁基荷方案。 ### [20:36](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1236s) 电力供需失衡下的过渡策略：天然气先行与核能承接 算力基础设施的规划范式正在发生翻天覆地的变化。过去，数据中心开发商的规划周期通常只有两到三年，因为传统公用事业电网总有充裕的备用电力，客户只需直接签订供电合同即可。然而，如今全美大部分公用事业公司的电力配额已消耗殆尽，这给整个行业敲响了沉重的警钟。仅仅展望两三年的旧模式已彻底失效，迫使各方不得不展开更长远的商业谈判与战略布局，规划深度远超18个月前的水平。 为了解决数据中心对“供电速度”（Speed to Power）的燃眉之急，Oklo 采取了与天然气发电开发商深度绑定的过渡战略。例如，Oklo 近期宣布与 Liberty Energy 达成合作，根据所需规模，天然气发电设备最快可在 12 个月内完成部署，第一时间为新建数据中心通电投运。随着后续 Oklo 的核能发电厂（Powerhouses）逐步落成，大部分基荷用电将平滑切换至清洁核能，而前期的天然气资产则转为调峰、负荷跟踪、电网平衡及备用电源。这种模式确保了天然气资产不会沦为“搁浅资产”（Stranded Assets），在具备长期商业价值的同时，实现了向零碳核能的平稳过渡。 ```mermaid flowchart TD accTitle: Gas to Nuclear Transition Strategy subgraph Phase1["第一阶段：快速供电部署（12个月内）"] A["Liberty Energy 天然气发电"] --> B["数据中心快速通电并投入运营"] end subgraph Phase2["第二阶段：核电落成与负荷转移"] C["Oklo 核能电厂 (Powerhouses) 投产"] --> D["承接数据中心主力基荷供电"] B -.-> D end subgraph Phase3["第三阶段：长期协同运行"] D --> E["24/7 零碳稳定电力"] A --> F["转为负荷跟踪、调峰与备用电源"] end ``` - **规划周期巨变**：数据中心行业彻底告别以往 2 至 3 年的短期电力规划，全面转向更长周期的能源锁定。 - **快速通电能力**：通过与 Liberty Energy 合作，最快 12 个月内即可交付天然气电力以激活数据中心。 - **资产价值保全**：天然气发电机组在核电投产后承担调峰与备用职能，避免资产过早报废。 --- ### [22:13](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1333s) 商业化首堆与监管审批：依托30年成熟数据的极速落地 针对现场观众关于爱达荷国家实验室（INL）项目的提问，演讲者明确澄清：Oklo 即将在爱达荷国家实验室部署的首座反应堆是一座**全功能的商业化发电厂**，绝非概念验证（Proof of Concept）或试点原型装置。这项技术底层的概念验证早在数十年前就已经完成——同类反应堆在该实验室成功连续运行了长达 30 年之久，积累了极其详实的历史运行记录、安全数据与工程测试成果。Oklo 无需重新摸索，而是直接将这些现成且经过充分验证的数据整合进提交给美国核能管理委员会（NRC）的许可证申请中。 谈及项目推进节奏与 3 年部署周期，政策层面的提速为先进核能带来了重大利好。白宫近期签署了四项行政命令，明确指示 NRC 原则上须在 18 个月内完成核电许可证申请的审查。例如行业同行 TerraPower 的申请近期已正式获批立案（Docketed），并进入了标准的 18 个月审查周期。在监管审批推进的同时，大量电厂工程、设备预制与前期准备工作均可同步并行展开，从而大幅压缩整体交付耗时。 - **非试点验证堆**：首座位于爱达荷国家实验室的反应堆即为商业运营电厂，直接继承过往 30 年的反应堆运行数据遗产。 - **审批周期提速**：受白宫行政令推动，NRC 先进反应堆的审查周期正向 18 个月的标准化流程看齐。 - **并行推进机制**：工程建设与预制生产在 18 个月审批期间同步启动，确保核准后迅速并网。 --- ### [23:42](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1422s) 核废料循环利用与高浓缩铀稀释：燃料供应的闭环方案 在回应关于如何利用现存反应堆乏燃料的提问时，演讲者揭示了 Oklo 在核燃料供应链上的双轨布局。对于首座示范商业堆，Oklo 将直接利用过往试验堆运行 30 年所留下的回收燃料。更重要的是，Oklo 正在筹建一座全尺寸的商业化核燃料回收工厂。公司在全美范围内完成了选址考察与多维度评估，预计将于今年晚些时候正式公布该回收工厂的具体选址。该设施将专门接收现有商业核电反应堆产生的乏燃料，通过先进回收工艺将其加工提纯，重新转化为 24/7 不间断运行的清洁电能。 除了自主建设乏燃料回收设施，Oklo 还在积极推进“高浓缩铀稀释”（Down-blending）路线。美国政府目前储备了大量高丰度核材料库存，通过降浓稀释技术，这些材料可转化为先进反应堆所需的特种燃料。Oklo 正在全力推进将这部分国家储备转化为商业堆可用燃料的合作通道，形成多渠道、自给自足的燃料保障闭环。 - **首堆燃料来源**：首座电厂直接继承利用成熟的回收燃料。 - **全国选址建厂**：全美范围选址已完成，年内将公布首座全尺寸乏燃料回收工厂所在地，变商业核废料为电能。 - **政府库存稀释利用**：同步锁定美国政府高浓缩核材料储备的稀释利用渠道，拓宽原材料来源。 --- ### [24:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1491s) “自备电力”与电网共生：数据中心微电网与表前商业模式 针对数据中心大规模“自备电力”（Bring Your Own Power, BYOP）对全美及全球电网影响的深度提问，演讲者指出未来将呈现“表后微电网”与“表前电网协同”并存的混合格局。为了避免巨额用电需求推高当地居民用电成本、引发公众抵触情绪，大型科技公司正普遍倾向于建立“表后”（Behind-the-Meter）独立微电网。例如前文提及的 1.1 GW 级别数据中心，本质上就是一套完全自给自足的微电网系统，不挤占公共电网配额。 另一方面，数据中心还可以转型成为“电网良好公民”（Good Grid Citizens）。传统上数据中心落户当地主要通过出资修建公园或资助社区项目来获取支持；而在未来，数据中心可以通过适度超配发电装机，在满足自身算力的同时反斜电力，为地方政府公共服务提供低成本电力补贴，或在极端天气导致电网大停电时扮演紧急备用电源角色。至于“表前”（In-Front-of-the-Meter）模式，公用事业机构正在与数据中心设计全新的电价关税（Tariff）与商业结构，确保基础设施扩容成本由数据中心全额承担，绝不转嫁给普通家庭或小企业主。从政治现实来看，转嫁电费成本是绝不可能被公众接受的，因此表后自建微电网与创新的表前定价机制将共同构成算力爆发时代的能源基石。 - **表后微电网（Behind-the-Meter）**：数据中心自主供电，与公共配电网物理或逻辑隔离，避免引发区域用电通胀。 - **电网良好公民（Good Grid Citizen）**：通过超额配置发电能力，在电网突发断电或危机时向社区反哺应急电力。 - **表前机制创新（In-Front-of-the-Meter）**：设计新型工业电价与分摊机制，杜绝数据中心建设成本外溢至民生电费账单。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=4s) 整个内华达州（包括拉斯维加斯在内）的所有发电装机容量总和尚不足 11 GW；全美最大都市纽约市在用电最高峰时的总负荷同样低于 11 GW。 — Brian Gitt > [08:38](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=518s) 一个人一生中所需消耗的全部能源，如果折算为核燃料材料，其体积仅仅相当于一颗普通的高尔夫球。 — Brian Gitt > [10:34](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=634s) 一座被称为“Oklo Powerhouse”的核电站中，约 70% 的零部件都来自常规成熟的供应链，无需专门进行严苛的核级认证。 — Brian Gitt > [15:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=951s) 反应堆在发生故障时无需人工干预，无需备用应急电源启动水泵，也无需任何机械运动部件介入，仅依靠重力、热膨胀以及自然空气对流等自然法则即可导出余热。 — Brian Gitt > [17:42](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1062s) 如果对美国现存的全部乏燃料进行再处理和回收利用，在无需新开采一盎司天然铀矿石的前提下，所蕴含的能量就足以支撑全美未来 150 年的用电需求。 — Brian Gitt > [19:21](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1161s) 你们的投产时间太遥远了，等过几年接近落地时再来找我们吧。 — 超大规模数据中心服务商 > [20:36](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1236s) 天然气发电设备最快可在 12 个月内完成部署，第一时间为新建数据中心通电投运。随着后续 Oklo 的核能发电厂逐步落成，大部分基荷用电将平滑切换至清洁核能。 — Brian Gitt > [22:13](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1333s) Oklo 即将在爱达荷国家实验室部署的首座反应堆是一座全功能的商业化发电厂，绝非概念验证或试点原型装置。 — Brian Gitt > [24:51](https://www.youtube.com/watch?v=o_SM4BJkf2w&t=1491s) 确保基础设施扩容成本由数据中心全额承担，绝不转嫁给普通家庭或小企业主。从政治现实来看，转嫁电费成本是绝不可能被公众接受的。 — Brian Gitt ## Mentioned Brian Gitt, Oklo Inc, OpenAI, Meta, Equinix, Switch, Siemens Energy (西门子能源), Vertiv, Liberty Energy, TerraPower, 美国核能管理委员会 (NRC), 爱达荷国家实验室 (INL), 阿贡国家实验室 (ANL), 美国国防部, 实验增殖堆二号 (EBR-II), Vogtle 核电站 ## Video description Brian Gitt, SVP, Head of Business Development, Oklo Inc Abstracts The arc of human progress has always moved away from less concentrated sources of energy toward more concentrated ones—from wood to coal, to oil and gas, to nuclear power. Future progress moves in the same direction: 24/7 clean nuclear power that unlocks the benefits of AI and cloud computing. Historically, there have been three big challenges to nuclear: 1. High upfront costs: ~$10+ billion cost of construction 2. Long build time: 7-10 years to deploy 3. Public perception: concerns about safety and nuclear waste Small nuclear power plants can overcome each of these challenges. They’re simpler and less expensive to build. Factory prefabrication combined with minimal civil works enables plants to deploy in 3 years. Small plants are walk-away safe and cool themselves without relying on people, pumps, or mechanical systems. They can also recycle used fuel and convert nuclear waste into clean energy. [Source](https://www.youtube.com/watch?v=o_SM4BJkf2w)

Source quotes

[0:04] Brian Gitt: 整个内华达州（包括拉斯维加斯在内）的所有发电装机容量总和尚不足 11 GW；全美最大都市纽约市在用电最高峰时的总负荷同样低于 11 GW。 [8:38] Brian Gitt: 一个人一生中所需消耗的全部能源，如果折算为核燃料材料，其体积仅仅相当于一颗普通的高尔夫球。 [10:34] Brian Gitt: 一座被称为“Oklo Powerhouse”的核电站中，约 70% 的零部件都来自常规成熟的供应链，无需专门进行严苛的核级认证。 [15:51] Brian Gitt: 反应堆在发生故障时无需人工干预，无需备用应急电源启动水泵，也无需任何机械运动部件介入，仅依靠重力、热膨胀以及自然空气对流等自然法则即可导出余热。 [17:42] Brian Gitt: 如果对美国现存的全部乏燃料进行再处理和回收利用，在无需新开采一盎司天然铀矿石的前提下，所蕴含的能量就足以支撑全美未来 150 年的用电需求。 [19:21] 超大规模数据中心服务商: 你们的投产时间太遥远了，等过几年接近落地时再来找我们吧。 [20:36] Brian Gitt: 天然气发电设备最快可在 12 个月内完成部署，第一时间为新建数据中心通电投运。随着后续 Oklo 的核能发电厂逐步落成，大部分基荷用电将平滑切换至清洁核能。 [22:13] Brian Gitt: Oklo 即将在爱达荷国家实验室部署的首座反应堆是一座全功能的商业化发电厂，绝非概念验证或试点原型装置。 [24:51] Brian Gitt: 确保基础设施扩容成本由数据中心全额承担，绝不转嫁给普通家庭或小企业主。从政治现实来看，转嫁电费成本是绝不可能被公众接受的。

Source quotes

Brian Gitt: The arc of human progress has always moved away from less concentrated sources of energy toward more concentrated ones—from wood to coal, to oil and gas, to nuclear power. Brian Gitt: Small plants are walk-away safe and cool themselves without relying on people, pumps, or mechanical systems. They can also recycle used fuel and convert nuclear waste into clean energy. [19:21] Hyperscalers (quoted by Brian Gitt): You guys are too far out for us. Come back in a few years when you're close. [24:51] Brian Gitt: It's politically infeasible

Original transcript

[0:04] Good afternoon everyone. So last session [0:07] of the day. My name is Brian Git. I lead [0:09] business development for Oaklo and we [0:12] design, build and operate [0:15] power plants, advanced nuclear power [0:17] plants for data centers and new AI [0:19] workloads. So we're going to talk all [0:20] about how we do that. what are some of [0:23] the biggest challenges and um some of [0:26] the recent events that we're seeing and [0:27] maybe to kick things off about a month [0:30] ago [0:32] 11 gawatt data center campus was [0:35] announced in Amarillo Texas 11 gawatt [0:38] just to level set is all of the power [0:42] capacity in the state of Nevada [0:44] including Las Vegas that's less than 11 [0:47] gawatt that's 11 g New York City at [0:51] total peak load is less than 11 GW. Now, [0:54] this is one data center campus that's [0:58] located behind the meter, meaning it's [1:01] meant just to support a mega digital [1:04] infrastructure data center project. That [1:07] is the trajectory of where we're going. [1:09] Everyone's heard, I'm sure, about [1:11] OpenAI's mega campuses up to 5 gawatts [1:14] that are being developed. We've heard [1:16] Meta make announcements. power demand is [1:19] set to triple within the next several [1:22] years by the certainly by 2030. So [1:24] depending on which report you look at [1:28] however so this is obviously a a clear [1:31] issue and problem. So how are we going [1:32] to meet this demand? How are we going to [1:34] develop all this power capacity? And so [1:37] that's what I want to talk with you [1:38] about today. And historically nuclear [1:41] power just to level set nuclear power [1:43] today provides 20% of all the [1:45] electricity in the United States. So [1:47] there's 94 operating reactors and for [1:49] the last 70 years they've been humming [1:52] along producing a huge chunk of US [1:54] electricity. However, the industry has [1:57] been incredibly stagnant and stuck for [2:00] basically 30 years. We've only turned on [2:03] two units recently in Georgia and in [2:06] decades. So why is that? What is the [2:08] problem? Why has nuclear not been [2:10] growing and why hasn't it scaled? Well, [2:12] there's really three fundamental [2:14] problems that have faced the industry [2:17] and these three problems I'm going to [2:19] talk about how Oaklo is overcoming to [2:21] really deploy these systems in a very [2:24] scalable and expedited manner. So, first [2:26] is high upfront costs. To give you an [2:28] idea, a large 1,000 megawatt power [2:32] nuclear power plant like they recently [2:34] built in Georgia, these cost on order of [2:37] 9 to10 billion dollar. This is mega [2:40] infrastructure project that can take up [2:42] to and to the second problem long build [2:45] time up to 8 to 10 years to build. So [2:49] you got 10 billion worth of investment [2:51] in something that takes 8 to 10 years to [2:53] build. And then although I would argue [2:55] that the nuclear industry is probably [2:58] the safest industry in America in 70 [3:00] years of commercial operation, there's [3:02] been one accident and no one got harmed [3:04] or hurt. Um very few industries can have [3:07] that kind of safety record. But still [3:09] there are safety concerns about [3:12] accidents and waste. And so these are [3:14] the three problems that Oaklo designed [3:17] it to our business model as well as [3:20] systems to overcome. So I just want to [3:22] frame that for context. So how do we do [3:23] it? We do it through three basic areas. [3:27] First is technology. Second is business [3:29] model. Third is size. And I'm going to [3:32] talk about this in the context of a [3:34] traditional plan because I think it's [3:36] helpful to be able to compare. So, first [3:38] off, we're not talking about some [3:40] massive mega infrastructure project with [3:42] a huge concrete cooling tower. We're [3:45] talking about something that could fit [3:46] within a suburb suburban environment or [3:49] even a master plan community. And we [3:51] when we start looking at comparing these [3:53] things, as I mentioned, just massive [3:55] amount of capital and time that goes [3:58] into building traditional plants about a [4:00] thousand megawatt. So, they're really, [4:02] really large. And although they are [4:04] space efficient, they still take up 800 [4:06] acres of land, which is not nothing, [4:09] right? And they typically have to be [4:11] cited near water resources. So they need [4:13] to be near a water reservoir, a lake, [4:15] the ocean, a river, something because [4:17] they use that water to cool the reactor. [4:21] Now, if you contrast, and the last thing [4:23] is although it is a small amount, they [4:26] do generate waste. And right now there [4:28] isn't a long-term solution of what to do [4:31] with that waste. So compare that to [4:34] these newer, more advanced reactor [4:35] designs like what we're building at [4:37] Oaklo. Well, first is about simplicity [4:40] and avoiding that upfront capital [4:43] investment. So most data center [4:45] companies are not going to come up with [4:47] $10 billion in capital to put into power [4:50] infrastructure. They're just not going [4:51] to do that. They typically buy powers as [4:53] a service. So our business model is we [4:56] sell power as a service, a next [4:57] generation utility, so to speak, like an [4:59] independent power producer where we're [5:01] basically selling kilowatt hours. We're [5:03] not asking the data center company to [5:06] come up with all of that capital [5:08] upfront. We project finance these [5:10] against long-term contracts. So, we're [5:12] signing a long-term power purchase [5:14] agreement with that uh hyperscaler or [5:17] collocation company and then they're [5:19] basically paying that over a 20-year [5:21] period and we're financing that through [5:23] a combination of debt and equity um on [5:26] the project. Now, instead of taking 10 [5:28] years to build something like this, we [5:30] could deploy in as little as 36 months. [5:33] Now to build it actually takes just [5:36] about 18 months but when you factor in [5:38] the licensing and longle procurement [5:40] items for certain elements it can take [5:42] longer. It's a tiny land footprint. So [5:45] we build 75 megawatt small power plants [5:48] and you can just put these like fashion [5:51] next to each other on a site and each [5:54] one all the way to the fence line [5:56] including all the infrastructure sits on [5:58] three acres of land. So it's a tiny [6:00] amount of land and there's no water used [6:03] for cooling the reactor. We use a liquid [6:05] metal and that metal basically you think [6:08] about it there's three ways you can [6:09] transfer heat from nuclear fuel to to [6:12] basically get that heat out. One is [6:15] basically water which we talked about [6:17] the traditional plants. Two is a gas [6:19] like a helium gas and then three is [6:22] metal. Well metal just from first [6:24] principles is going to be the most [6:26] effective at transferring heat. So it's [6:28] very very efficient and it avoids the [6:30] the need for water. And then finally we [6:34] can actually recycle used nuclear fuel. [6:37] Over 90% of the energy that is sitting [6:40] around the country right now in these [6:43] spent um used fuel uh stockpiles at the [6:46] existing plants can be recycled and [6:48] turned into new fuel. [6:51] And so another key aspect of this is [6:53] synchronization. So as we everyone in [6:56] the news now is hearing about 500 [6:58] megawatt data centers or thousand [7:00] megawatt data centers or like I said [7:02] even up to 11 gawatt data centers and [7:07] with that they don't build that all at [7:08] once though you're building individual [7:10] buildings one at a time each building is [7:13] about 100 megawatts and those are built [7:15] over successive years in several phases. [7:18] So by building out your power [7:21] infrastructure in similarsized chunks, [7:24] you are synchronizing the buildout so [7:26] you don't overbuild and have power [7:28] infrastructure that's sitting there [7:29] uneconomical for too long, which is [7:32] really critical to making sure you're [7:33] delivering a really cost-effective [7:35] solution. So that's another advantage of [7:38] going smaller. [7:40] In addition to being able to synchronize [7:42] the buildout, another advantage of [7:44] smaller is you can get to higher levels [7:46] of redundancy. Most nuclear power [7:48] plants, they have to go offline when you [7:50] refuel them. And like any power plant, [7:53] you have minor and major major [7:55] maintenance outages. So when you take [7:58] that offline, if you if if it's a,000 [8:00] megawatts or even 500 megawatt plant, [8:02] what are you going to do to replace [8:04] that? I mean, a data center can't wait [8:06] for that power. It needs it immediately. [8:08] But if you have smaller sizes, chunks of [8:10] 75 megawws, now you can just add [8:13] additional redundancy. So when you take [8:15] one off for refueling, you plug the new [8:17] one in and you can cycle through. So [8:19] there's a lot of advantages to building [8:21] smaller. [8:23] We built a pretty significant pipeline [8:25] at this point up to 14 gawatts right now [8:28] and we're working through um those [8:30] commercial agreements with some folks [8:32] such as Equinex and Switch and uh and [8:35] others in the industry. Um as you can [8:38] imagine, there's just a thirst for power [8:40] anywhere they can get it. And I want to [8:44] get into some of the foundational [8:46] aspects of why nuclear power is really a [8:50] superior form of energy. And it really [8:51] gets into this concept of energy density [8:54] in terms of how many how many materials [8:55] does it take to generate a kilowatt hour [8:59] of energy. And you could fit your the [9:02] amount of energy in your entire lifetime [9:04] in terms of materials use into a single [9:07] golf ball. I mean that is how [9:09] concentrated and energy dense this type [9:12] of fuel is. And so when you think about [9:15] that through an environmental lens, when [9:17] you think about the mining that's [9:18] involved and the materials and the land [9:21] use and the transportation of those [9:22] materials, at the end of the day, [9:24] there's nothing that beats nuclear power [9:26] from an environmental lens. And as I [9:28] mentioned, the one concern that people [9:30] have from an environmental standpoint is [9:32] the waste. And it is recyclable. And [9:34] we'll talk a little bit more about that [9:36] later on. [9:39] So when we let's just kind of quantify [9:41] this like even in Arizona and this these [9:43] are solar values in Arizona. So if you [9:45] take some of the best possible um solar [9:48] values and you take a 500 megawatt power [9:51] plant and you plunk it down in Arizona, [9:54] you know, we could basically cover about [9:56] 21 acres of land where you would take up [9:58] about 2700 acres of land with solar. And [10:02] again, the solar is only running during [10:03] the day. It's not running 24/7. So you [10:06] can see the dramatic efficiency that you [10:08] get by using a very highly concentrated [10:12] energy dense fuel. [10:14] Water is another concern everyone. [10:16] Obviously water is scarce especially in [10:18] the American West especially where we [10:20] are here near Las Vegas. And as you can [10:23] see the amount of water required is very [10:25] minimal because the only water we use is [10:28] makeup water for the steam turbine. [10:30] Right? We we operate like a thermal [10:32] power plant. All nuclear is is just a [10:34] heat source. It's a way to basically [10:36] heat up water till it boils. So you can [10:38] create steam and then basically turn a [10:41] turbine just like any thermal power [10:42] plant. So the only water used for us is [10:45] just makeup water for that steam [10:47] turbine. You can see it's basically [10:49] comparable to solar. You might say, [10:51] "Well, wait a minute. What are you using [10:52] water for solar?" You have to clean off [10:54] those panels. There's tons of dust and [10:57] dirt that gets blown onto them. So [10:59] there's tons of water that's used to [11:00] clean them off on a routine basis. So [11:02] it's a it's basically equivalent to [11:04] that. in comparison to kind of these [11:06] much larger water usage thermal plants [11:08] like traditional nuclear or natural gas [11:11] combined cycle. [11:14] One of the key ways that we keep the [11:16] cost down is that most of the components [11:19] 70% of every component that goes into an [11:22] Oaklo powerhouse which is just the name [11:24] of our power plant is basically existing [11:26] supply chain. So there's nothing nuclear [11:29] rated or certified about it. So a good [11:32] example of this is our steam turbine and [11:34] generator. Our whole balance of plant [11:36] basically how we turn that heat into [11:38] electricity. We're work we're partnering [11:40] with Seaman's Energy and Seaman's Energy [11:43] makes this existing product. It's [11:45] basically off the shelf that we tune it [11:48] and they've sold about 5,000 of these [11:50] around the world um over time. So we're [11:52] taking something existing. It has [11:54] nothing to do with nuclear energy. It's [11:57] just basically a steam turbine, right? [11:59] And there's many components that are [12:01] like that in this. It's mostly the [12:03] nuclear fuel and some of the other uh [12:06] components that surround the fuel that [12:08] are really nuclear related. [12:11] So when you look at a site, you can [12:13] think of the again as Lego blocks. You [12:16] can build one and just keep putting them [12:18] next to each other as you scale. So you [12:20] can think about you can start small. You [12:21] start with 75 megawatts, you go to 150, [12:24] you go to you can go to 300 to a,000. [12:26] You could put 5,000 megawatts of power [12:28] on one site and we are having [12:30] conversations with folks about ambitious [12:33] projects like that. So it's very very [12:35] scalable similar to how you build [12:37] digital infrastructure. [12:40] Some just basic aspects of this in terms [12:42] of how it works and what are some of the [12:45] elements. So it's runs really high [12:48] temperature. So 450° C Celsius. This [12:51] makes it very efficient and that means [12:54] there's a lot of heat that we can um [12:57] deploy not just for electricity. We can [12:59] also use that heat for industrial [13:00] purposes and even for data center [13:03] cooling. We're partnering with Verdive [13:05] recently in in an announcement where [13:07] we're actually going to use that high [13:10] temperature steam to drive absorption [13:12] chillers to cool data center. So it can [13:15] be used not only for electricity but [13:16] also for cooling as well as other [13:19] industrial heat applications as well. [13:22] Another key component is this is [13:24] nonpressurized. So if you could imagine [13:26] a large stainless steel tank similar to [13:29] you what you'd see in prochemical or [13:30] pharmaceutical just a huge stainless [13:32] steel tank maybe about you know 40 feet [13:35] tall and about 20 foot in diameter [13:38] around. It's non-pressurized. This this [13:41] kind of tank is made all the time for [13:42] industrial applications. There's nothing [13:45] that is custom or bespoke about it in [13:48] terms of the materials that you're [13:49] using. It's basically stainless steel or [13:52] 316 stainless or 304 depending on the [13:55] part. [13:58] So, as I mentioned before, really at the [14:00] core, what nuclear power is is just a [14:02] way to generate heat to turn to create [14:04] steam to turn a turbine. And you can see [14:06] in this diagram here, you got your fuel [14:09] in the bottom of that big stainless [14:11] steel tank. It's creating the heat as [14:12] the fision reaction starts. As that [14:15] heats up, it heats up. Basically, that [14:17] whole tank is full of liquid metal. And [14:20] as that heat gets transferred to the [14:22] liquid metal, it goes up through some [14:24] heat exchangers and then at that part, [14:26] it's going to generate steam. [14:30] So, there's a few basic components to [14:32] this. You have your reactor vessel, the [14:34] big tank that I described here in the in [14:36] the bottom left, your steam generators. [14:38] Then you have a steam generator, a steam [14:41] turbine, I should say. Uh that's what I [14:42] was referencing with Seaman's energy. [14:44] And then in this is the aircooled [14:46] condensers are basically a way to not [14:48] use water. So this is a way that this is [14:51] the heat sink because you got a lot of [14:53] heat to dissipate. So we can basically [14:55] use large fans in these aircooled [14:57] condensers uh to get rid of excess heat. [15:00] So these are incredibly reliable. The [15:03] the nuclear fleet today in the United [15:05] States is the most reliable thermal [15:06] power source. They operated at like 92% [15:10] capacity factor. That just means they're [15:12] running basically all the time except [15:14] when they're down for some type of [15:15] refueling or maintenance. And because of [15:18] the size of these as I mentioned earlier [15:20] you can design enough redundancy in the [15:23] system to where you can have be have [15:25] 100% uptime basic effectively um on [15:28] these depending what the requirements [15:30] are. So this is why for example the [15:33] department of defense is very interested [15:34] in this technology. We just got a notice [15:37] of intent to award for a military base, [15:39] an air force base um up in Alaska. And [15:43] there it's because of the reliability [15:45] and resiliency of these systems and how [15:48] you can have enough redundancy on them. [15:51] All of this is not new. This goes back a [15:54] long ways. This technology was first [15:56] developed um out of Idaho National [15:58] Laboratory and Argon National [15:59] Laboratory. These are US government labs [16:02] and we are just commercializing [16:04] something that has to date over 400 [16:06] years of operational history. About 25 [16:09] of these um designs have been built [16:11] around the world. The unit that we're [16:14] most closely commercializing ran for 30 [16:16] years from 1964 to 1994 at Idaho [16:20] National Lab. and all of the data, all [16:22] of the testing that was you done on that [16:24] plant, we get to use in our license [16:27] application to the Nuclear Regulatory [16:29] Commission. So there's nothing that is [16:32] really new and novel about this of the [16:34] core technology. This is um we're not [16:36] building test facilities or pilot [16:38] facilities to test this out. That's [16:40] already been done. We're just [16:41] commercializing something that's been [16:43] proven. [16:45] And these safety tests, they actually [16:46] tried to force the fuel to melt to have [16:50] a meltdown event. Basically, they [16:53] self-cool and self-regulate. So, you [16:55] don't need basically humans to do [16:57] anything. You don't need backup power to [16:59] pump anything. You don't need mechanical [17:01] systems or any kind of pumps to operate. [17:04] It basically cools off on its own using [17:06] natural forces, meaning gravity, thermal [17:09] expansion, and air flow will cool the [17:11] reactor. So in the worst case [17:13] catastrophic event, let's say some [17:15] natural disaster, hurricane, fire, [17:17] flood, whatever, obviously it's a bad [17:19] day, the power plant shuts down, but [17:22] there's no risk of radiation release [17:24] beyond the site footprint. So our [17:26] emergency planning zone is the 3 acre [17:29] site boundary. Unlike large plants, [17:32] which are perfectly safe by the way, as [17:33] I mentioned earlier, but they have 10 [17:35] mile emergency planning zones. Ours is [17:38] basically the 3acre footprint of the [17:40] power plant. [17:42] So I mentioned fuel recycling. This is [17:44] just I think an immense resource that [17:47] we're sitting on. We can basically power [17:49] the United States for the next 150 years [17:52] just taking our existing used fuel and [17:54] recycling it because as I mentioned [17:56] earlier over 90% of the energy is still [17:59] in that material. So we it's basically [18:02] we wouldn't have to mine one um ounce of [18:05] uranium out of the ground and you could [18:07] still basically power the US for 150 [18:10] years. That's how much energy is just [18:12] sitting there. And just recently [18:16] policy has really been reset to focus on [18:18] this. And there's a big initiative right [18:20] now to basically facilitate fuel [18:22] recycling of used fuel as well as unlock [18:25] existing government stockpiles of [18:27] material that can be used in this. [18:30] So our time frame for this, we're going [18:32] to be deploying our first commercial [18:34] plant in 2028. We're going to be [18:36] submitting our license application to [18:38] the nuclear regulatory commission this [18:40] fall and you can see from there we're [18:42] going to be doing lots of stuff in [18:44] parallel to that. So doing long lead [18:46] procurement doing initial uh site work [18:50] and in development of the site all can [18:52] happen in parallel to the license [18:54] application and we can you're really [18:56] going to start seeing scale of these [18:58] things in the 2030 early 2030s time [19:01] frames as things really ramp up um [19:03] across the board. So, I know we're just [19:06] about out of time. We have a couple [19:07] minutes. If anyone has any questions, [19:10] happy to answer any and then happy to [19:12] stay around if folks have questions as [19:13] well. Did you have a go? [19:21] >> Uh, first, thanks for this chat. It's a [19:22] fascinating topic and nuclear energy is [19:25] really popular right now. So, it's [19:26] really interesting to get insight from [19:27] the inside source. as a head of business [19:29] development, how do you balance this [19:32] sort of longish timeline, couple years, [19:35] three to three years plus like the [19:37] uncertainty of the licensing, the [19:39] regulatory issues and then whether or [19:42] not the product actually can work um in [19:44] the commercial sense and working with [19:46] these customers and these steps PPAs I [19:49] guess if you that's the acronym you [19:51] used. I'm just kind of curious at like [19:53] how like these mission critical database [19:55] centers, data centers need to be built [19:58] and need to be powered. You mentioned [20:00] that the energy use is going to explode. [20:01] So maybe shouldn't use the word explode, [20:03] but uh but you get Yeah, that's the gist [20:06] of that question. [20:06] >> Yeah. No, it's a great question. I mean, [20:08] speed to market is one of the biggest [20:09] challenges facing the data center [20:11] industry right now. They can't build out [20:13] infrastructure fast enough. And to be [20:15] honest, three years ago when I was [20:17] talking with some of the hyperscalers, [20:18] they're like, "You guys are too far out [20:20] for us. come back in a few years when [20:22] you're close. That was basically what I [20:24] heard and when I was going around [20:26] talking to them and then about 18 months [20:28] ago that changed and what happened was [20:31] they realized that all the power [20:32] capacity is getting eaten up around the [20:34] country and they need to look out much [20:36] further in their planning. You know, [20:38] traditionally they've only really looked [20:39] out two or three years um in terms of [20:41] what it would take to build the digital [20:43] infrastructure because the utilities [20:44] usually had the power. They would just [20:45] go to the utility and contract for the [20:47] power. Well, now most of the power, most [20:49] of the utilities are out of power around [20:51] the country. So, it's a huge wakeup call [20:54] for the industry to say, you know, the [20:56] old practice of just looking out a [20:58] couple years, 2, three years is not [20:59] going to work. We need to look out [21:01] further. And so, now they're very [21:03] engaged in these kind of commercial [21:04] discussions and planning out way beyond [21:07] what they were just 18 months ago. So, [21:09] the the world has changed dramatically [21:11] in that supply demand imbalance. though [21:14] we're partnering with natural gas [21:16] developers to to provide gas generation [21:19] as soon as possible. So we can for [21:21] example through we announced a [21:23] partnership with Liberty Energy um [21:25] recently they can actually deploy within [21:28] 12 months depending on the scale and the [21:30] size that is needed initial gas [21:32] generation. So you can have speed to [21:34] power, start energizing data center [21:36] buildings. Then over time as our [21:38] powerhouses are available, you can start [21:41] moving more of that load to the nuclear [21:43] power plants and then using the gas for [21:46] load following and balancing and backup [21:48] and other services. So those gas assets [21:51] aren't stranded. They're going to have a [21:52] useful business long business case and [21:54] useful life, but most of the energy will [21:57] get trans transitioned towards the [21:59] nuclear assets. So the the timelines [22:02] have shifted quite dramatically recently [22:05] on that front [22:08] question. [22:13] >> Um thanks for the presentation. Um so [22:16] like the example that you showed in that [22:18] I think it's at Idaho National Lab, [22:20] right? Is that like a P or is it going [22:22] to be like a fully functional? [22:23] >> No. Yeah. So the the question is will [22:25] that first plant at Idaho National Lab [22:27] that we're going to deploy that is a [22:28] commercial power plant. It is not a [22:30] proof of concept. It is not a pilot [22:32] facility. The one that ran there for 30 [22:34] years was the proof of concept. So [22:37] there's already 30 years of track record [22:39] and data and testing done on this [22:41] technology. So that's why we know don't [22:43] need to do it. We can just take the data [22:45] that was already tested in general and [22:48] bring that data to the nuclear [22:50] regulatory commission in our [22:52] application. [22:57] Um so like the three years time frame is [23:00] basically like uh using those data and [23:02] then uh using your uh research to set up [23:06] something like that is [23:07] >> yeah and build it everything. So yeah [23:09] we're right now the the White House just [23:12] issued four executive orders and a part [23:14] of those executive orders was directing [23:17] the Nuclear Regulatory Commission to [23:19] basically review all license [23:21] applications within 18 months. So that's [23:23] kind of in in one of our competitors, [23:25] Terap Power, just was submitted recently [23:28] and they were docketed for an 18-month [23:30] review cycle. So that is kind of the new [23:33] kind of standard for the licensing. Now, [23:35] you can do a lot of stuff in parallel [23:37] while something's going through that [23:38] licensing process. So yes, [23:42] >> question, [23:42] >> could you elaborate a little bit on how [23:44] you're going to use the existing fuel [23:48] that's being stored quite literally at [23:50] most of the existing reactors today? [23:53] Yeah. So the question around how are we [23:55] going to leverage all of that existing [23:56] used fuel? We are for our first plant. [23:59] We're actually leveraging recycled fuel [24:01] because we're using fuel that would ran [24:03] in that plant for 30 years. We're using [24:05] some of that fuel in our first one. But [24:07] we're building a fullscale recycling [24:09] facility. We'll be announcing that soon [24:12] um a little bit later this year of where [24:14] that will be. We did a whole nationwide [24:16] search and did a full evaluation of [24:18] various sites. And so we're going to be [24:20] building a facility that will take in [24:22] used fuel from the existing fleet and [24:25] converting that through a recycling [24:27] process into 24/7 clean power. So that's [24:30] how we're going to meet that. In [24:32] addition to the recycling, there's also [24:34] an effort right now to do what's called [24:36] down blending. So the US government is [24:39] sitting on very large stockpiles of [24:40] material that can be down blended and [24:43] used and burned in this in these kind of [24:45] advanced reactors. And so we're also [24:48] pursuing that opportunity as well. [24:51] >> Looks like one back there. [24:55] >> First of all, this is really really [24:56] exciting. Um, but my question is more [25:00] along the lines of the current power [25:03] grid and how and your thoughts on, you [25:06] know, all these data centers are [25:08] starting to bring their own power, [25:09] right? and you're in a good position for [25:11] that, but how do you think that that's [25:12] going to affect the power grid in this [25:15] country and maybe worldwide long term? [25:18] I'm just interested in your thoughts [25:19] there. [25:20] >> Yeah, I think we're going to see a [25:21] hybrid. I think you're going to see data [25:23] centers bringing their own power and [25:25] having their own micro grids so that [25:26] they're not going to basically be [25:29] inflating the cost for the community and [25:31] being seen in a negative light. I think [25:33] that's going to be a real thing. in like [25:35] for example that 11 gawatt data center [25:38] that I mentioned in the beginning that's [25:39] a behind the meter micro grid basically [25:42] that is basically just bringing their [25:44] own power for those data centers but I [25:46] also think grid um centers can be good [25:48] grid citizens and provide additional [25:51] services back to the grid they can [25:52] overbuild capacity I mean right now when [25:54] a data center goes in they'll contribute [25:57] money to build parks or um local [26:00] community projects things like that well [26:01] what if they overbuilt a little power [26:03] infrastructure and provided additional [26:05] power to subsidize local government [26:07] services or whe if there's some kind of [26:10] energy blackout or emergency there would [26:12] be extra power available to power the [26:14] town. I mean there's a lot of creative [26:15] ways that data centers actually can [26:17] contribute value to the community in [26:19] terms of reliability and resilience [26:21] without impacting cost. And then in [26:23] front of the meter, there's new [26:25] commercial constructs being generated [26:27] right now with the way utilities work [26:29] with data centers and the tariffs so [26:31] that they don't get burdened passed [26:33] those costs don't get passed on to your [26:36] average rate, your average um homeowner [26:38] or business owner. So, it's all getting [26:40] figured out and market by market it's [26:42] different. I'm very confident it will [26:44] because it's politically infeasible. [26:45] There's no way the public is going to [26:47] allow massive inflation of their bills [26:50] um because of a data center going in. [26:52] that's just not going to happen. So, [26:54] there's going to be two solutions to [26:55] that. Behind the meter and micro grids, [26:56] bring your own power, as well as these [26:58] new commercial structures for what's [27:00] considered in front of the meter. [27:04] Well, I know we're uh just about out of [27:06] time, but I'm happy to stay around and [27:08] answer some questions if want to stay. [27:10] So, thank you. [27:12] >> Thank you.