NVIDIA's $2000/Year GPUs and Becoming a Defense Contractor

NVIDIA's $2000/Year GPUs and Becoming a Defense Contractor — Gamers Nexus

Gamers Nexus

AI-assisted summary of NVIDIA's $2000/Year GPUs and Becoming a Defense Contractor — original source. Check the original for full context.

Source published:

Key ideas

  • NVIDIA is shifting toward an enterprise software licensing model, aiming to charge approximately $2,000 per GPU annually for agent software stacks on top of existing hardware sales. (0:00)
  • The company reported $96.2 billion in quarterly revenue—$89 billion from data centers alone—generating over $1 billion per day compared to just $5 billion across all of fiscal year 2016. (10:22)
  • NVIDIA has partnered with defense firms Anduril Industries and Armada to build deployable DGX Pod modular data centers for military contested environments and forward warfighter edge processing. (16:19)
  • Following a DOJ indictment alleging $300 million in smuggled high-end GPUs to China via Southeast Asia, NVIDIA shifted its public posture from flat denial to dismissing diversion as an insignificant drop in the bucket. (26:41)

Detailed notes

## Background This video was published on YouTube by Gamers Nexus on November 2, 2026. According to the page description, it covers NVIDIA's evolution at GTC Berlin 2026 into an "infrastructure company" and defense contractor, deploying pod-based data centers to military conflict zones, pursuing enterprise software licensing, leveraging GPU allocations against cloud partners, and managing ongoing GPU smuggling scandals.

## TL;DR Gamers Nexus investigates NVIDIA's radical pivot away from consumer gaming hardware toward recurring enterprise software licensing, defense contracting, and sovereign infrastructure. Host Steve Burke details Jensen Huang's plan to extract roughly $2,000 per GPU annually for agentic AI software while locking in multi-billion-dollar Neocloud customers through aggressive allocation leverage reminiscent of the GeForce Partner Program. The report also scrutinizes NVIDIA's partnerships with defense startups like Anduril and Armada to deploy modular data centers directly into war zones, while contrasting Huang's public charm offensive against mounting DOJ antitrust scrutiny and criminal smuggling indictments.

## Atmosphere & style The video carries an incisive, cynical, and investigative tone, blending Gamers Nexus's trademark hardware consumer advocacy with geopolitical gravity. Steve Burke delivers a weary yet razor-sharp monologue, puncturing corporate PR soundbites with biting sarcasm about 'cute robots' and war-zone computing pods. Watching it feels like a sobering reality check, charting a dramatic, somewhat dystopian arc from the innocent DIY era of zip-tying liquid coolers onto GTX 1080 Tis to confronting a trillion-dollar military-industrial surveillance giant.

## Bulletin - [00:00](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=0s) NVIDIA is shifting toward an enterprise software licensing model, aiming to charge approximately $2,000 per GPU annually for agent software stacks on top of existing hardware sales. - [10:22](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=622s) The company reported $96.2 billion in quarterly revenue—$89 billion from data centers alone—generating over $1 billion per day compared to just $5 billion across all of fiscal year 2016. - [16:19](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=979s) NVIDIA has partnered with defense firms Anduril Industries and Armada to build deployable DGX Pod modular data centers for military contested environments and forward warfighter edge processing. - [26:41](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1601s) Following a DOJ indictment alleging $300 million in smuggled high-end GPUs to China via Southeast Asia, NVIDIA shifted its public posture from flat denial to dismissing diversion as an insignificant drop in the bucket.

## Key takeaways - [00:00](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=0s) Jensen Huang envisions NVIDIA evolving beyond a hardware chip vendor into an indispensable enterprise software and recurring licensing platform akin to Oracle. - [11:42](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=702s) NVIDIA leverages GPU allocation access against AI Neocloud providers, reportedly pressuring them to avoid adopting competing silicon like AMD GPUs or Google TPUs. - [03:44](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=224s) The tactic of using silicon supply limits to compel cloud provider loyalty directly mirrors NVIDIA's historical treatment of add-in-board partners under programs like the GeForce Partner Program. - [04:35](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=275s) NVIDIA is pursuing defense contracts directly, collaborating with the Department of Defense and defense contractors to deploy AI architectures into military theaters. - [16:19](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=979s) The Vera Rubin-based DGX Pod reference architecture enables containerized data centers to be prefabricated and deployed to the field within 14 to 30 days. - [16:19](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=979s) Anduril's Menace system integrates NVIDIA modular architectures to run classified military intelligence in active war zones. - [20:07](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1207s) Venture capital from Peter Thiel's Founders Fund ties together edge infrastructure provider Armada with defense surveillance firms like Palantir and Flock Safety. - [21:38](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1298s) NVIDIA has deprioritized gaming and consumer PCs, relegating retail graphics revenue into a combined 'Edge Computing' financial reporting bucket. - [26:41](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1601s) The DOJ indicted Greg Lei for smuggling more than $300 million worth of restricted NVIDIA A100, H100, RTX 4090, and RTX 5090 hardware into China. - [26:41](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1601s) After SEC filings revealed over 99% of Singapore data center compute billing went to US-based entities while hardware shipped elsewhere, NVIDIA ceased breaking out Singapore revenue. - [31:52](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1912s) Huang advocates for early childhood AI adoption and public optimism on media tours while internally maintaining a notoriously aggressive management culture. - [34:25](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=2065s) Federal regulatory scrutiny is mounting, with DOJ antitrust inquiries probing NVIDIA's customer leverage and investments like Hugging Face, alongside congressional testimony requests.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=0s) The Quiet Part Out Loud: $2,000 Per GPU Per Year

NVIDIA is preparing an aggressive acceleration of its business model toward enterprise software licensing, moving beyond simply selling silicon hardware. Jensen Huang outlines an operational model where enterprise clients will pay ongoing fees of roughly $2,000 per GPU per year to access the software stack required to run autonomous AI agents on their computing clusters. With NVIDIA shipping upwards of 20 million GPUs annually and half of that volume funneling into enterprise data centers, recurring licensing revenue quickly compounds into tens of billions of dollars. As Jensen puts it, this shift could fundamentally alter how the market perceives the company:

> “One of these days, you know, people might not categorize us as a chip company anymore.”

For Gamers Nexus host Steve Burke, this shift represents NVIDIA moving away from its traditional identity as a graphics hardware vendor toward something resembling an enterprise software monopoly like Oracle—locking customers into indefinite licensing payments for hardware they already bought. Despite pulling in over a billion dollars in daily revenue, NVIDIA is searching for fresh avenues to extract margin from its enterprise footprint. While NVIDIA offers enterprise suites like NVIDIA AI Enterprise at $4,500 per GPU annually, this new licensing model targets baseline software execution across all corporate deployments. Alongside this software pivot, NVIDIA is facing mounting scrutiny over global supply lines, ongoing GPU smuggling, and an investigation published by *The New York Times* alleging that Jensen Huang was targeted in a Chinese intelligence plot involving companies like Megaspeed and individuals such as Alice Guo.

```mermaid flowchart TD accTitle: NVIDIA Shift to Enterprise Recurring Licensing Model accDescr: Diagram showing NVIDIA moving from silicon sales to ongoing yearly software licensing. subgraph Hardware["Annual Hardware Volume"] A["NVIDIA Ships ~20M GPUs/Year"] --> B["50% Allocated to Enterprise (~10M GPUs)"] end subgraph SoftwareLicensing["Recurring Monetization Model"] B --> C["Base Hardware Acquisition Cost"] B --> D["Annual Software Support & Agent Licenses (~$2,000/GPU/Year)"] D --> E["Compounding Indefinite Revenue Stream"] end ```

[02:20](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=140s) A Flock for the Living Room at GTC Berlin

In October 2026, Jensen Huang took his executive roadshow through Europe, holding meetings in France, Sweden, the United Kingdom, and Germany before headlining GTC Berlin. During the Berlin keynote, Jensen took an eccentric turn on stage by introducing a family of autonomous robotic ducks equipped with cameras, encouraging the audience to adopt them:

> “It’s going to be so exciting. Everybody should have one. And you have a camera. They have a camera. It's so cute. There's little cameras in the family of micro ducks. Like a flock.”

Beneath the playful demonstration of camera-equipped micro-ducks—which did not even run on NVIDIA silicon—GTC Berlin served as the staging ground for commercial announcements concerning AI safety guardrails, supply chain pacts, and partnerships with specialized neocloud infrastructure providers. However, access to these programs came with distinct caveats, as reports surfaced that cloud providers integrating non-NVIDIA silicon had their GTC access cut or were outright excluded from the ecosystem.

- European tour destinations: France, Sweden, United Kingdom, and Germany. - Stage demonstration: Camera-equipped autonomous robotic duck flock interacting with Jensen. - Policy fallout: Neocloud partners running non-NVIDIA accelerators reportedly faced event exclusions and partner status limitations.

[03:44](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=224s) Allocation As a Lever: From GPP to Cloud Providers

NVIDIA's current commercial tactics mirror its historical dealings with add-in-board (AIB) graphics card partners in the consumer PC gaming space. Longtime hardware journalists recognize the familiar pattern: leveraging tight silicon allocations to compel supply-chain conformity. Where NVIDIA once pressured board partners with preferential allocations to freeze out rival GPUs—a practice epitomized by the controversial GeForce Partner Program (GPP)—the company now applies those same pressure tactics against multi-billion-dollar enterprise cloud firms. By restricting event access, partner tier perks, and chip allotments to data centers that deploy hardware from rivals like AMD, NVIDIA dictates behavior across modern cloud infrastructure.

> “It’s the same strategy that we've talked about in the past, just applied to much more substantial things than gaming hardware alone.”

Steve Burke highlights that reporting on gaming hardware for decades provides unique insight into this playbook. The stakes have grown from retail GPU shelves to sovereign computing and macroeconomic infrastructure. The commercial leverage once used to control consumer desktop brands is now wielded against multinational corporations and geopolitical defense clients who are willing to pay ongoing annual licensing retainers without pushback.

[04:35](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=275s) Defense Contracting and the DSX Pod

NVIDIA has introduced the DSX Pod, an enterprise modular container reference design engineered for Vera Rubin computing clusters. Intended for rapid deployment, a DSX Pod unit can be integrated on-site with local power, fluid cooling, and high-speed networking in just 14 days. To construct and distribute these battlefield-ready datacenters, NVIDIA has partnered with Peter Thiel-backed defense entities, including Palmer Luckey’s Anduril Industries and modular edge-infrastructure firm Armada.

```mermaid flowchart LR accTitle: Modular Edge Datacenter DSX Pod Supply Architecture accDescr: Flow from NVIDIA's Vera Rubin design to deployment on military fronts via defense contractors. subgraph Architecture["Reference Platform"] A["NVIDIA DSX Pod<br>(Vera Rubin Modular Spec)"] end subgraph Contractors["Defense Partners"] A --> B["Anduril Industries<br>(Palmer Luckey)"] A --> C["Armada<br>(Peter Thiel-backed)"] end subgraph TargetZones["Deployment Fronts"] B --> D["Warfighter Tactical Processing"] C --> E["US Navy Combat Ships & Desert Conflict Zones"] end ```

Anduril markets these self-contained compute systems as a way of delivering real-time processing straight to the forward warfighter, while Armada focuses deployments on combat ships and desert staging environments across the Middle East. Jensen Huang has become open about pursuing state defense expenditure, framing corporate alignment with military initiatives as patriotic contribution.

> “And so we want America to be the wealthiest country so that we can fund the mightiest military. And I think that that is our way of contributing to national security.”

When discussing foreign theaters and regional conflict, Jensen expressed confidence in long-term commercial integration, remarking: “I believe at the end of the war, Middle East will be more stable than before. And so if we were there, if we're considering it before, we should absolutely be considering it after. And so I'm 100% in on that.”

- **DSX Pod Deployment Window:** 14-day turn-up time for modular data pods with power, cooling, and network hookups. - **Hardware Architecture:** Engineered around Vera Rubin-generation enterprise racks. - **Defense Partners:** Anduril Industries (battlefield tactical AI systems) and Armada (maritime and forward operating edge centers).

[06:23](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=383s) Physical Media Production at Gamers Nexus

To counterbalance subscription lock-in and corporate cloud control, Gamers Nexus highlighted its push for physical media ownership, manufactured in-house at their headquarters. The studio produces physical disc runs for Blu-ray and audio CD formats with zero DRM restrictions, ensuring owners can back them up to private local platforms like Plex or Jellyfin. Production involves on-site burning with multi-drive disc duplicators, high-speed USB writing, and custom cases 3D-printed using New Jersey-sourced filament.

- **AI GPU Black Market Documentary Pack:** A two-disc Blu-ray release covering the investigation into covert enterprise GPU smuggling networks, accompanied by a custom magazine booklet, hand-transcribed closed captions, and an optional USB drive insert. The reporting was an official finalist for an excellence in journalism award. - **Special Reports Blu-ray Collection:** Curated deep-dive technical investigations, engineering interviews, factory tour features, and the multi-part EVGA closure retrospective. - **Cyber Dystopia Music CD:** An electronic and chip-tune soundtrack composed by Aaron Pauley for GN's investigative documentaries—including the Sega police raid report—packaged in a 3D-printed vaporwave case with optional digital audio USB storage. - **Merchandise Restocks:** Parody "Microslop" apparel restocked across all standard sizes at `store.gamersnexus.net`.

[08:06](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=486s) Redefining the Company: Infrastructure, Sovereignty, and Compute as GDP

NVIDIA’s leadership no longer views the company as an ordinary component supplier or graphics card developer. While Jensen occasionally references NVIDIA’s gaming beginnings to ground the company’s history, he now presents NVIDIA as the indispensable backbone of modern industrial civilization:

> “Nvidia today is really a full stack end-to-end complete AI factory platform.”

Under this messaging, GPUs are pitched not just as commercial compute hardware, but as sovereign assets that determine the baseline economic potential of nation-states. Jensen routinely tells international heads of state and enterprise executives that compute capacity directly governs domestic economic output:

> “So the first thought is that compute equals revenues. Now the big idea of course compute equals GDP that we also know. Compute equals a country’s GDP. I am certain compute equals revenues. I’m certain also that compute equals GDP.”

To keep growing its enterprise revenues beyond its record quarterly numbers, NVIDIA is pairing this macroeconomic rhetoric with ongoing support licensing fees, defensive edge compute packages, and persistent hardware allocations. This transformation unfolds against a backdrop of covert GPU smuggling operations, corporate surveillance allegations, and direct military technology partnerships.

[10:22](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=622s) NVIDIA Today vs. the GTX 10 Era NVIDIA

Covering GTC is completely different from how it used to be, reflecting an astonishing evolution where the NVIDIA of today is entirely unrecognizable compared to the company of a decade ago. In its fiscal year 2016, around the high-water mark of NVIDIA's gaming dominance with the GeForce GTX 1080 and GTX 1080 Ti era, the company made roughly $5 billion in total annual revenue, with $1.4 billion pulled in during the fourth quarter. While NVIDIA had begun talking about machine learning and deep learning back then, gaming was firmly the priority. By contrast, in a single recent quarter, NVIDIA reported $96.2 billion in revenue, with $89 billion generated exclusively by its data center segment.

Because a fiscal quarter spans roughly 90 days, NVIDIA now generates over $1 billion every single day. Measuring that against the peak of its 10-series gaming era, NVIDIA now grosses more money in just two days than it used to make across an entire three-month quarter. As Steve Burke remarks on the financial explosion: “That's insane. But that's where we are.” Armed with this massive momentum, CEO Jensen Huang has spent the year traveling relentlessly to pitch governments, massive corporations, and defense contractors. Acting like a traveling merchant with wares for anyone who has the coin, Huang touched down in China, South Korea, the UK, Sweden, France, Taiwan, Germany, Japan, Kazakhstan, and numerous other regions before arriving in Europe for GTC Berlin.

- **FY 2016 total revenue:** ~$5 billion (~$1.4 billion in Q4). - **Recent single quarter revenue:** $96.2 billion total, with $89 billion driven strictly by data center sales. - **Run-rate milestone:** NVIDIA generates >$1 billion in revenue per day, earning more in two days than an entire 2016 quarter.

---

[11:42](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=702s) Allocation As a Lever Against Neoclouds

Behind the European partner showcases at GTC Berlin, NVIDIA was executing familiar, aggressive business plays. Research group SemiAnalysis revealed that NVIDIA barred several "Neoclouds" from GTC Berlin. Neoclouds operate as GPU-as-a-service providers, renting out GPU compute power specifically for AI processing. NVIDIA effectively spawned this industry, having poured billions into backing startups like CoreWeave, Nscale, Crusoe, Irys, Nebius, Lambda, and Firmus. In Jensen Huang’s own words: “In the case of clouds, if we didn't support CoreWeave to exist, these Neoclouds, these AI clouds wouldn't exist.”

```mermaid flowchart TD accTitle: NVIDIA Allocation Leverage Loop accDescr: Diagram showing how NVIDIA uses GPU supply constraints to prevent Neoclouds from diversifying into rival hardware.

subgraph Dependency ["Ecosystem Dependency"] A["NVIDIA GPU Allocation"] -->|"Powers 100% of Fleet"| B["Neocloud Providers<br/>(CoreWeave, Nscale)"] end

subgraph Retaliation ["Threat of Diversification"] B -.->|"Attempts to derisk with"| C["Alternative Hardware<br/>(AMD GPUs, Google TPUs)"] C -.->|"Triggers supply retaliation"| D["NVIDIA Leverage<br/>(Reduced Allocations, Event Bans)"] D -->|"Forces compliance"| A end ```

The friction emerged when these Neocloud providers attempted to derisk their operations by adopting non-NVIDIA silicon, such as AMD GPUs and Google TPUs. NVIDIA responded by flexing its market power through an "allocation play"—a pressure tactic Gamers Nexus previously documented in the consumer hardware space, such as EVGA’s split from NVIDIA and laptop OEMs dropping AMD models to avoid losing NVIDIA mobile GPU allocations. SemiAnalysis reported that Neocloud executives feel subjected to high-pressure tactics to remain exclusively on NVIDIA, forcing some to evaluate competitor chips in secret out of fear of retribution.

Public SEC filings substantiate the severity of this single-source vulnerability. Nscale identified NVIDIA's supply willingness as a major operational risk factor that could cripple its commitments. CoreWeave—referred to by Burke as "NVIDIA sugar baby CoreWeave"—operates an infrastructure consisting solely of NVIDIA hardware. In its quarterly filings, CoreWeave warned investors that its complete dependency on limited suppliers introduces risks like diminished control over manufacturing costs and potentially being forced into binding purchase commitments above market rates.

> “Executives feel NVIDIA has been using high-pressure tactics to keep them NVIDIA only. As such, Neocloud executives are starting to consider offering TPUs or AMD GPUs quietly to avoid getting pressured by NVIDIA and to avoid the feeling that NVIDIA will retaliate.” — *SemiAnalysis, cited by Steve Burke*

---

[16:19](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=979s) NVIDIA Pivoting to Defense Contracting and War-Zone Pods

Beyond consumer and cloud hardware, NVIDIA has pivoted directly toward military contracting. In May, NVIDIA teamed up with the U.S. Department of Defense and seven other companies to deploy AI models on defense systems, and Jensen Huang has publicly voiced geopolitical support for military action, stating his belief that the Middle East will emerge more stable in the aftermath. At GTC Berlin, this military direction manifested concretely in the announcement of the DGX Pod, a containerized, modular reference platform built around the Vera Rubin AI architecture that compresses deployment timelines from six months down to as little as 14 to 30 days.

```mermaid flowchart LR accTitle: DGX Pod Deployment Pipeline accDescr: Flow diagram showing how NVIDIA and defense partners convert Vera Rubin architecture into war-zone data centers.

subgraph Factory ["Prefabs & Validation"] A["NVIDIA Vera Rubin<br/>Reference Design"] --> B["Prefabrication & Wiring<br/>(Armada / Anduril)"] end

subgraph Field ["Deployment"] B --> C["DGX Pod Containers<br/>(Menace System)"] C --> D["Tactical Edge / War Zones<br/>(Local Power & Cooling Hookup)"] end ```

While Huang framed the DGX Pod on stage as an innocent revolution in data center supply chains, the fine print of the announcements confirmed that these modular units are explicitly intended for battlefield theaters. NVIDIA partnered with primary defense contractors, notably Anduril Industries and Armada. Anduril, a major DoD contractor manufacturing cruise missiles, submarines, and autonomous defense drones, revealed that its "Menace" deployable edge pods use NVIDIA's DGX modular architecture to run classified defense and intelligence AI in contested environments—which Burke plainly translates to "data centers for war zones." Simultaneously, startup Armada partnered with Johnson Controls to build deployable "SI" (System Intelligence) pods based on the Rubin platform, fully plumbed, cabled, and validated prior to leaving the factory floor.

> “The speed by which we can turn on, go from installation, arrival of computing to production of tokens will go from many months—we think something along the lines of four to five months—down to about a month. Okay. So, this is a complete transformation of the world's supply chain and this will change the way that people install and build data centers.” — *Jensen Huang*

- **Anduril Menace pods:** Modular, edge-ready data centers built to execute classified defense intelligence tasks in "contested environments" (active war zones). - **Armada and Johnson Controls:** Manufacturing factory-prefabricated, drop-in Rubin-based containers branded as "SI factories" (System Intelligence). - **Rapid provisioning:** Reduces deployment from up to half a year to roughly two to four weeks, arriving pre-plumbed and pre-wired for immediate connection to on-site power and cooling.

[20:07](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1207s) Pod-Based Data Centers and Defense Ties

NVIDIA's expansion into ruggedized, transportable infrastructure extends beyond traditional enterprise facilities directly into military defense contexts and harsh industrial environments. Steve Burke jokes that if NVIDIA’s AI GPUs—sarcastically correcting himself to say "SI" so as not to be labeled an enemy of the state—aren't deployed to active war zones, they will end up on oil rigs, quipping, “Drill, baby drill.” He details Armada’s massive Leviathan data center pod, which serves as the largest form factor in their Gallant lineup. Per Armada's specifications, a single Leviathan unit packs up to 576 NVIDIA GB300 NVL72 GPUs within an estimated 1.77-megawatt IT envelope. Standard Leviathan site deployments are tuned for 5 to 20 megawatts but can scale up to 50 megawatts, packing an immense density of compute into shipping containers.

```mermaid flowchart TD accTitle: Peter Thiel's Founders Fund Ecosystem accDescr: Diagram showing Founders Fund investments connecting surveillance, defense, and NVIDIA hardware deployments. subgraph Capital["Venture Funding"] FF["Founders Fund<br>(Peter Thiel)"] end

subgraph Entities["Funded Companies & Partners"] ARM["Armada<br>(Leviathan Data Centers)"] PAL["Palantir Technologies"] FLOCK["Flock Safety"] PER["Persona<br>(ID Verification)"] end

subgraph Hardware["Compute Infrastructure"] NV["NVIDIA<br>(GB300 NVL72 GPUs)"] end

FF --> ARM FF --> PAL FF --> FLOCK FF --> PER ARM -->|"deploys up to 576 GPUs per unit"| NV NV -.->|"publicly partners / 'hearts'"| PAL ```

Beyond hardware partnerships, Armada shares deep institutional ties through venture capital backing. Both Armada and key surveillance players have received funding from Peter Thiel's Founders Fund. That same firm bankrolled Persona, the identity verification service originally tapped by Discord. Thiel’s capital frequently concentrates in military and surveillance tech, which increasingly intersects with online ID verification mandates. These commercial pushes align with bipartisan government legislation, including New York's SAFE for Kids / Child Data Protection Act (NY KOSA), which Gamers Nexus previously covered on their GNCA channel. Thiel also co-founded data-mining contractor Palantir—an organization NVIDIA openly embraces, with Steve noting they displayed an unironic "NVIDIA hearts Palantir" graphic. Thiel's fund has also provided startup capital to automated license-plate surveillance network Flock Safety, solidifying NVIDIA's deep entanglement with the military-industrial surveillance apparatus.

- Armada Leviathan features up to 576 NVIDIA GB300 NVL72 GPUs in a single modular container. - Leviathan deployments run on an IT envelope baseline of 1.77 MW, scaling from 5–20 MW up to 50 MW. - Peter Thiel’s Founders Fund has backed Armada, Persona, Flock Safety, and Palantir Technologies. - Both Democratic and Republican lawmakers continue pushing ID verification initiatives like New York's KOSA.

---

[21:38](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1298s) Computers for Agents and RTX Spark

Even while courting defense ministries at GTC Berlin, NVIDIA reserved a minor, perfunctory nod for consumer-facing hardware through its RTX Spark platform. Burke sarcastically points out that RTX happens to double as the stock ticker for defense behemoth RTX Corporation (formerly Raytheon), half-joking that it is a conspiracy theory for another day. During the keynote, NVIDIA highlighted expensive mobile and workstation hardware geared toward European markets, showing off systems that retail for upwards of $7,000. In recorded keynote footage, Jensen Huang pitched the hardware as an evolutionary step beyond traditional computing instruments:

> “We started to think what can we do to re-imagine the personal computer for this new era where the personal computer is not just your tool your personal computer tool but your personal agent or your personal assistant and the way you interact with this computer would be essentially a companion somebody who helps you do work.”

Huang touted high-performance companion workstations capable of running 1 petaflop to handle CAD, modeling, and engineering suites like CATIA, Autodesk, Adobe, Cadence, and Synopsys, pointing users toward Windows-powered DGX Stations. Regarding traditional PCs, Steve notes Jensen barely spared a thought: “And at least Jensen remembered desktops this time sort of... That was literally the extent of his memory that they exist.”

Alongside the expensive hardware, NVIDIA promoted "Underdog," a desktop agent marketed to process AI locally and privately on RTX Spark devices. The RTX Spark initiative, originally billed alongside Microsoft as a fundamental reinvention of the PC, has since been taken on an international keynote tour. However, Gamers Nexus argues this sparse smattering of announcements underlines how far NVIDIA has pivoted away from ordinary end users. Financial reports reflect this abandonment: NVIDIA now obscures gaming and consumer hardware numbers by folding them into a catch-all "Edge Computing" reporting segment alongside robotics, automotive electronics, and AI-RAN telecommunications base stations.

- RTX Spark workstations target massive compute thresholds, reaching up to 1 petaflop for enterprise suites like Synopsys and CATIA. - RTX Spark laptops highlighted at the European presentation command price tags as steep as $7,000. - NVIDIA introduced "Underdog," a local, on-device AI agent for RTX Spark systems. - Consumer and GeForce gaming revenue is no longer reported cleanly, buried instead inside Edge Computing alongside robotics and automotive divisions.

---

[23:57](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1437s) The GPU Smuggling Denial Narrative

Transitioning from European stages to Asian markets, Gamers Nexus dissects NVIDIA’s rapidly shifting PR posture on illicit GPU smuggling. Historically, NVIDIA flatly dismissed international chip diversion reports as unevidenced fiction, with executives once branding the reporting as outright "fake news." US sanctions restrict top-tier accelerators from being legally exported to adversarial nations like China and Russia. In reality, these restricted GPUs consistently seep across borders: Gamers Nexus documented the pipeline firsthand in an investigative book (bundled on Blu-ray and physical flash drives at store.gamersnexus.net), and NVIDIA silicon frequently surfaces inside recovered Russian battlefield reconnaissance drones. Under President Donald Trump, an arrangement was floated where select cut-down accelerators like the H200 could ship to China in exchange for a 25% revenue cut paid back to the US Treasury, yet China has instituted reciprocal roadblocks to favor local champions like Huawei.

Faced with mounting scrutiny, Jensen Huang persistently attempted to laugh off smuggling claims during public appearances in 2025:

> “Over the years, people have speculated about diversion. We've chased down every single concern and we've repeatedly tested and sampled data centers around the world and found no diversion. Nvidia's chips are not little tiny potato chips. These are two tons. Each one of our GPUs is a rack that's two tons, two tons large or the even the smaller versions of our racks are hundreds of pounds. And so it's not easy to smuggle.”

Huang routinely scoffed at the press, remarking that if people claim GPUs are being smuggled, he "really would love to see it." Despite these dismissals, independent investigations by Gamers Nexus, Reuters, and the Financial Times continued to document extensive black-market distribution networks.

---

[26:41](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1601s) Arrests, Paper Trails, and "Drops in the Bucket"

The corporate defense collapsed into open contradiction when federal authorities stepped in. On October 1st, the Department of Justice arrested Greg Lei in California for orchestrating a massive illegal operation that smuggled more than $300 million worth of high-end enterprise hardware into China. The indictment charged Lei with conspiracy to violate the Export Control Reform Act, outbound smuggling, and money laundering. Between 2023 and 2024, Lei routed export-controlled compute hardware—including NVIDIA A100, H100, RTX 4090, and RTX 5090 systems procured from domestic US system builders—through intermediary shell hubs in Malaysia and Singapore before forwarding them into mainland China. In Taiwan, local police separately arrested an active NVIDIA employee connected to an illicit scheme smuggling intact server units.

```mermaid flowchart LR accTitle: Greg Lei Illicit GPU Smuggling Route accDescr: Illicit supply chain routing enterprise GPUs from US manufacturers through Southeast Asia into China. subgraph US["United States"] MFR["US System Integrators"] LEI["Greg Lei & Co-Conspirators"] end

subgraph Hubs["Transshipment Hubs"] SEA["Malaysia & Singapore<br>(Invoicing & Re-export)"] end

subgraph Destination["Sanctioned Market"] PRC["Mainland China<br>(Restricted End Users)"] end

MFR -->|"Procures A100, H100,<br>RTX 4090 / 5090"| LEI LEI -->|"Outbound Smuggling<br>(2023-2024)"| SEA SEA -->|"Sanctions Evasion"| PRC ```

When confronted by Bloomberg about these criminal indictments, NVIDIA abandoned its claim that diversions were impossible, instead downplaying the multi-million-dollar illicit pipeline as trivial:

> “Governments and media have found insignificant diversion of controlled US products, a drop in the bucket compared to the ocean of domestic compute China already has. US prosecutors allege less than 1/half of 1% of Nvidia products have been diverted.”

Burke highlights the absurdity of dismissing $300 million in smuggled military-grade compute as mere pocket change, noting that “only NVIDIA could attribute to $300 million worth of anything” as insignificant.

Further regulatory focus zeroed in on NVIDIA’s revenue pipeline through Singapore, which served as a major billing transit point. In an August 2025 SEC filing, NVIDIA stated:

> “Customers use Singapore to centralize invoicing, while our products are almost always shipped elsewhere, over 99% of controlled data center compute revenue billed to Singapore was for orders from US-based customers for the second quarter and first half of fiscal year 2026.”

Shortly after filing this disclosure, NVIDIA stopped reporting Singapore revenues as a distinct line item to deflect public scrutiny. By late October, reports from *The Information* and *Tom's Hardware* revealed that an Israeli cybersecurity firm discovered export-controlled NVIDIA B300 accelerators active within a Chinese state-owned telecom provider. NVIDIA responded by pivoting its language once again, brushing off the breach:

> “With tens of millions of GPUs in service in approved countries, this report again demonstrates that smuggling is a non-factor blocked by governments on both sides of the globe.”

Steve notes that NVIDIA subtly swapped its old dismissive buzzword "non-starter" for "non-factor," attempting to spin ongoing federal indictments, arrests of Super Micro co-founders, and their own rogue personnel as meaningless statistical noise.

- DOJ indicted Greg Lei for smuggling over $300 million in AI servers into mainland China via Malaysia and Singapore. - Smuggled hardware cited in the Lei indictment spans enterprise A100 and H100 servers alongside high-end consumer RTX 4090 and RTX 5090 cards. - Taiwanese law enforcement arrested an NVIDIA employee implicated in smuggling full server assemblies. - NVIDIA stopped breaking out Singapore revenue after SEC filings showed over 99% of Singapore-billed high-compute orders were billed to US entities while shipped elsewhere. - An Israeli firm uncovered illicit B300 hardware operating inside a state-owned Chinese telecom network.

[30:11](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1811s) Recategorizing NVIDIA: From Chips to Surveillance and Defense

Steve Burke reflects on how NVIDIA's identity is fundamentally shifting away from hardware. While NVIDIA will remain a chipmaker, Jensen Huang's statements hint at an Oracle-style future dominated by recurring software and licensing revenue. Gamers have been sidelined entirely; Burke jokes that he is unsure how NVIDIA's gaming division employees still have jobs, unless the company simply forgot they worked there. Instead, every corporate initiative centers on accelerating artificial intelligence adoption, with Huang steering the company toward surveillance systems and military contracting.

Recalling a tour of NVIDIA headquarters from around 2012 to 2014, Burke remembers seeing early demonstrations that now appear chilling in hindsight. Staff proudly showcased video walls featuring airport camera footage processed by NVIDIA hardware running automated facial recognition. At the time, Burke admits he did not appreciate the implications of the technology and merely wanted to see gaming hardware. Looking back, NVIDIA had already laid the groundwork to integrate its processors into mass surveillance and defense infrastructure over a decade ago.

> “You know one of these days you know people might not categorize us as a chip company anymore.” — Jensen Huang

[31:52](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1912s) The AI Charm Offensive vs. Internal Realities

To soften public perception, Huang has embarked on a worldwide charm tour, appearing on variety television shows, dining with locals in night markets, and presenting "cute robots" onstage to evangelize an optimistic AI future. Huang actively advocates for integrating artificial intelligence into early education, framing exposure for children as empowering rather than dangerous. Burke views this as a deliberate strategy to hook younger generations early so that algorithmic integration becomes normalized. In Burke's estimation, anyone questioning AI or digital identity verification down the line will end up viewed as "the old man yelling at the data center-shaped cloud."

Behind this warm public persona, however, lies an aggressive internal corporate culture. Former NVIDIA employees consistently describe Huang in private conversations as cutthroat and brutal. In one alleged all-hands meeting following a failed product launch, Huang interrogated the room, demanding to know who was responsible until individual employees could be singled out. This sharp-elbowed posture matches NVIDIA’s broader commercial history, such as the punitive 2018 GeForce Partner Program (GPP) used against vendors distributing AMD graphics cards, its evasive responses regarding GPU smuggling, and its active efforts to secure sales with military defense contractors.

> “What I would advise all children is this: Engage AI. Use it. Use it to empower yourself. Use it to lift yourself.” — Jensen Huang

[34:25](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=2065s) Regulatory Pushback and Antitrust Scrutiny

On policy, Huang adamantly opposes proactive government guardrails, arguing that lawmakers should wait for proven, physical harm rather than drafting rules around theoretical dangers. Despite this resistance, federal regulators are intensifying their scrutiny of NVIDIA's market dominance and commercial tactics. On October 28, Reuters reported that the U.S. Department of Justice opened inquiries into NVIDIA's business practices, focusing on whether the company exerts anticompetitive pressure on clients and questioning its investments in AI model hub Hugging Face.

Simultaneously, Senator Elizabeth Warren has repeatedly petitioned for Huang to testify before congressional committees. The Department of Justice is also examining whether NVIDIA attempted to circumvent antitrust scrutiny regarding its arrangements with Groq. In the face of mounting political pressure, NVIDIA's financial growth remains staggering: leadership projected a 70% revenue increase for the upcoming fiscal year, alongside quarterly revenue guidance of $108 billion (with a 2% margin of error).

```mermaid flowchart TD accTitle: Federal Regulatory Pressure on NVIDIA accDescr: Outline of antitrust and congressional investigations targeting NVIDIA's business practices and acquisitions.

subgraph DOJ["DOJ Antitrust Inquiries"] direction TB A["Hugging Face Deal & Investments"] --> D["Monopoly & Distribution Scrutiny"] B["Customer Sales & Leverage Tactics"] --> D C["Groq Antitrust Inquiries"] --> D end

subgraph Congress["Legislative Action"] E["Senator Elizabeth Warren Petitions"] --> F["Calls for Jensen Huang Congressional Testimony"] end

DOJ --> G["Broader Federal Scrutiny"] Congress --> G ```

> “Right now the technology is in its formative stages, and the advice that I would have is to regulate practical and actual harm, and not regulate theoretical and hypothetical harm.” — Jensen Huang

- Reuters confirmed the Department of Justice is reviewing whether NVIDIA's acquisitions allow it to lock down fundamental AI software distribution points. - Senator Elizabeth Warren has pushed repeatedly for Jensen Huang to appear before congressional committees to address NVIDIA's market power. - NVIDIA projected revenue of $108 billion (±2%) for the subsequent financial quarter, reporting earnings on November 17.

[36:38](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=2198s) Conclusion: The Infrastructure for the Planet

Burke closes by contrasting modern enterprise NVIDIA against the PC gaming enthusiast landscape of 2016. A decade ago, Gamers Nexus made content taking $700 GeForce GTX 1080 Ti cards with basic factory coolers, tearing them down, and modding them with closed-loop liquid coolers and zip-tied fans over the VRMs. Today, NVIDIA has repositioned itself to build and control what Huang terms the "infrastructure for the planet," leaving consumer gaming behind in pursuit of enterprise computing, cloud rentals, and defense systems.

To support independent technical investigative journalism—such as forthcoming pieces on data centers, smart TV tracking, and corporate surveillance—Burke directs viewers to the Gamers Nexus store and Patreon. He highlights the physical Blu-ray and USB documentary packages, which include an exclusive printed booklet insert shipping between December and January.

> “But as you know, NVIDIA is also the infrastructure for the planet.” — Jensen Huang

## Notable quotes > [08:06](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=486s) I am certain compute equals revenues. I’m certain also that compute equals GDP. — Jensen Huang

> [04:35](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=275s) And so we want America to be the wealthiest country so that we can fund the mightiest military. And I think that that is our way of contributing to national security. — Jensen Huang

> [34:25](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=2065s) Right now the technology is in its formative stages, and the advice that I would have is to regulate practical and actual harm, and not regulate theoretical and hypothetical harm. — Jensen Huang

> [00:00](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=0s) One of these days, you know, people might not categorize us as a chip company anymore. — Jensen Huang

> [11:42](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=702s) In the case of clouds, if we didn't support CoreWeave to exist, these Neoclouds, these AI clouds wouldn't exist. — Jensen Huang

> [31:52](https://www.youtube.com/watch?v=KbWm2MHL2i0&t=1912s) What I would advise all children is this: Engage AI. Use it. Use it to empower yourself. Use it to lift yourself. — Jensen Huang

## Mentioned NVIDIA, Jensen Huang, Steve Burke, Gamers Nexus, GeForce Partner Program, DSX Pod, DGX Pod, Vera Rubin, Anduril Industries, Armada, Peter Thiel, Palmer Luckey, Founders Fund, Palantir Technologies, CoreWeave, Nscale, Crusoe, Nebius, Lambda, Firmus, SemiAnalysis, AMD, Google, Greg Lei, US Department of Justice, Elizabeth Warren, Hugging Face, Groq, RTX Spark, Flock Safety, Persona, Super Micro

## Video description SUPPORT US: Get the GN Disc Packs https://store.gamersnexus.net/?category=Discs+%26+Posters or our freshly restocked Microslop shirts! https://store.gamersnexus.net/products/microslop-parody-t-shirt NVIDIA's GTC has changed over the years. GTC Berlin 2026 sees NVIDIA launching pod-based data centers with companies that deploy them to military conflict zones, a focus on computers for "agents," and separate news about GPU smuggling. We used to cover gaming technology, emergent scientific computing technology, and advancements in computer graphics for video games. In this one, we talk about NVIDIA's move toward becoming something that we think starts to look like a defense contractor, its shift in focus to being an "infrastructure company" (per Jensen Huang himself), and allegations of using GPU allocation as leverage.

The best way to support our work is through our store: https://store.gamersnexus.net/

Like our content? Please consider becoming our Patron to support us: http://www.patreon.com/gamersnexus

TIMESTAMPS

00:00 - The Quiet Part Out Loud 02:20 - A Flock for the Living Room 04:35 - Defense Contracting 06:23 - SUPPORT US 08:06 - NVIDIA Says It's an Infrastructure Company 10:27 - NVIDIA Today vs the GTX 10 Era NVIDIA 11:42 - Allocation As a Lever 16:19 - NVIDIA Pivoting to Defense Contracting 21:38 - Computers for Agents 23:57 - GPU Smuggling 30:24 - Conclusion: Recategorizing NVIDIA

** Please like, comment, and subscribe for more! **

Links to Amazon and Newegg are typically monetized on our channel (affiliate links) and may return a commission of sales to us from the retailer. This is unrelated to the product manufacturer. Any advertisements or sponsorships are disclosed within the video ("this video is brought to you by") and above the fold in the description. We do not ever produce paid content or "sponsored content" (meaning that the content is our idea and is not funded externally aside from whatever ad placement is in the beginning) and we do not ever charge manufacturers for coverage.

Follow us in these locations for more gaming and hardware updates:

t: http://www.twitter.com/gamersnexus f: http://www.facebook.com/gamersnexus w: http://www.gamersnexus.net/

Our policies, processes, and ethics statements relating to review samples, advertising, travel, errors, and more are transparently and publicly available on this page: https://gamers.nexus/ethics-statements

Steve Burke: Host, Writing Mike Gaglione: Editing Ben Benson: Writing

[Source](https://www.youtube.com/watch?v=KbWm2MHL2i0)

中文简报

英伟达年费2000美元的GPU与转型国防承包商 Gamers Nexus深入调查了英伟达从消费级游戏硬件向企业软件持续订阅授权、国防承包及主权基础设施的彻底转型。主持人Steve Burke详细披露了黄仁勋的战略规划：通过智能体AI软件向每块GPU每年收取约2000美元，并利用类似GeForce伙伴计划的激进配额杠杆锁定价值数十亿美元的Neocloud客户。该报道还审视了英伟达与Anduril和Armada等国防初创企业的合作，计划将模块化数据中心直接部署到战区，同时将黄仁勋在公开场合的魅力攻势与美国司法部日益加剧的反垄断审查及走私刑事起诉进行了鲜明对比。 英伟达正在转向企业软件许可模式，计划在现有硬件销售的基础上，对智能体软件栈按每块GPU每年收取约2000美元的费用。 (0:00) 该公司报告季度营收达962亿美元，其中仅数据中心就贡献了890亿美元，相当于每天进账超10亿美元，而2016财年全年的营收仅为50亿美元。 (10:22) 英伟达已与国防企业Anduril Industries和Armada展开合作，打造可部署的DGX Pod模块化数据中心，用于军事对抗环境和前线作战边缘计算。 (16:19) 在美国司法部就价值3亿美元的高端GPU经由东南亚走私至中国提起起诉后，英伟达的公开表态已从断然否认转向将违规流转淡化为微不足道的“沧海一粟”。 (26:41)

Source quotes

[8:06] Jensen Huang: I am certain compute equals revenues. I’m certain also that compute equals GDP. [4:35] Jensen Huang: And so we want America to be the wealthiest country so that we can fund the mightiest military. And I think that that is our way of contributing to national security. [34:25] Jensen Huang: Right now the technology is in its formative stages, and the advice that I would have is to regulate practical and actual harm, and not regulate theoretical and hypothetical harm. [0:00] Jensen Huang: One of these days, you know, people might not categorize us as a chip company anymore. [11:42] Jensen Huang: In the case of clouds, if we didn't support CoreWeave to exist, these Neoclouds, these AI clouds wouldn't exist. [31:52] Jensen Huang: What I would advise all children is this: Engage AI. Use it. Use it to empower yourself. Use it to lift yourself.

Original transcript

[0:00] One business model we've been working on [0:01] for some time and will be accelerated uh [0:04] in the near future is software [0:06] licensing. [0:07] >> Software licensing will be will become [0:09] increasingly a major part of our [0:11] company's business. Um the way that we [0:13] license will be uh some amount of some [0:16] for example like two it's about $2,000 [0:19] per GPU per year to run all the software [0:21] that Nvidia provides that runs agents on [0:24] these AI batteries. [0:25] >> And so software license for enterprise [0:27] business will be very very large. And [0:30] so, you know, if if our software license [0:32] is, you know, just roughly say for for [0:34] for simple math, a couple of thousand [0:36] dollars per GPU per year and we have [0:38] millions of GPUs, [0:39] >> right? [0:40] >> We have we we ship, you know, 20 million [0:42] GPUs, say a year. I'm growing and so [0:45] half of it goes into world's enterprise. [0:47] It adds up. One of these days, you know, [0:49] people might not categorize us as a chip [0:51] company anymore. I have to torture you [0:53] through some education first and then [0:56] the entertainment is coming. Nvidia had [0:58] already disowned gaming to say that it's [0:59] no longer a graphics company, but it's [1:01] looking to shift its classification even [1:03] further to somewhere between defense [1:05] contractor and basically Oracle with [1:09] software licensing for hardware forever. [1:12] And alongside all this, they're working [1:14] with companies that deploy data centers [1:16] literally to battlefields [1:19] uh and warfronts, which is maybe not [1:22] great, but at least it's not a national [1:25] security risk or something. Quote, [1:27] "Nvidia chief seen as target of Chinese [1:29] spy plot." End quote. Okay, maybe maybe [1:32] a little bit of little bit of national [1:34] security risk that we try to be humble [1:36] here, but between the rental GPUs and [1:39] paying for them forever and the GPU [1:41] smuggling and the questions around Alice [1:44] Juan and Meg Speed, I I hate to say it, [1:47] but I just have to. We told you so. [1:50] Other than the expiring allegations from [1:53] the new New York Times investigation, [1:55] apparently making more than a billion [1:57] dollars per day in revenue isn't enough [1:59] for Nvidia. So, Nvidia needs more ways [2:01] to extract more money from its [2:03] customers, and that's licensing for [2:05] using its GPUs. Nvidia hasn't expanded [2:07] upon Jensen's software licensing [2:09] comments, but we assume they're not [2:11] about a cut for Nvidia's AI enterprise [2:14] software that costs $4,500 per year per [2:16] GPU. [2:17] >> I love getting rich. You know, who [2:19] doesn't love making money? [2:20] >> In October, Jensen continued his 2026 [2:23] world tour by sweeping through Europe, [2:25] visiting France, Sweden, the United [2:27] Kingdom, and Germany. Jensen's European [2:28] trip culminated at GTC in Berlin where [2:31] he opened for a family of ducks. [2:36] Hey guys. [2:40] Oh yeah. Okay. Come on. Here. Come over [2:43] here. [2:47] Take a nap. [2:49] Take a nap. [2:52] You need to take a nap. [2:55] You guys are so cute. Can you do a [2:57] somersault? [3:00] It's going to be so exciting. Everybody [3:02] should have one. And you have a camera. [3:05] A They have a camera. It's so cute. [3:08] There's little cameras in the family of [3:11] micro ducks. Like a flock. [3:15] Oh Besides telling everyone they [3:18] need a robotic duck that has cameras and [3:21] that doesn't have Nvidia chips in them, [3:23] Nvidia announced updates related to AI [3:25] safety, supply chain agreements, [3:27] Neocloud partnerships, and more, [3:29] presumably for the Neoclouds who [3:31] received an invite and weren't banned [3:33] from attending GTC due to using chips [3:35] that weren't Nvidia in their cloud [3:37] infrastructure. Nvidia played back its [3:39] greatest hits from its past of gaming, [3:41] like holding allocation hostage, as [3:44] we've often reported on when AIB [3:46] partners of Yore would talk about the [3:48] ways that they felt Nvidia was using its [3:51] allocation, its availability of GPU [3:53] silicon to guide them to behave in a [3:56] particular way. In this way, now that [3:59] this is happening at a global scale with [4:02] much larger companies that have huge [4:04] economic impact for countries around the [4:06] world, the history of being gaming [4:09] hardware reporters, [4:11] I'm not happy about this. Uniquely [4:14] positions us to see what Nvidia is [4:16] doing. It's the same strategy that we've [4:18] talked about in the past, just applied [4:21] to much more substantial things than [4:23] gaming hardware alone. In the past [4:25] weeks, the company has been accused of [4:26] restricting access to events for [4:28] companies that use devices sold by AMD [4:31] or by other companies, which sounds [4:33] awfully familiar to the old GeForce [4:35] partner program. And Nvidia has even [4:37] courted sales with its new favorite [4:39] customers, not you, not me, defense [4:42] contractors. And that would include [4:45] agencies who are happy to pay thousands [4:47] of dollars per GPU per year for ongoing [4:49] software support. Seeing as that money [4:52] probably indirectly came from Jensen [4:54] anyway in the gigantic circle, it's not [4:57] so much for them. Nvidia introduced the [4:59] DSX pod, a reference design for building [5:01] Vera Rubin based systems in modular [5:03] containers. DSX pod units can be [5:06] deployed in 14 days and connected to a [5:09] site's power, cooling, and networking [5:11] infrastructure. Nvidia's DSX pod [5:13] partners include military suppliers and [5:15] Peter Tealbacked companies Anderil and [5:18] Armada. Anderil, founded by former [5:21] Oculus Rift maker Palmer Lucky, and [5:23] Armada make modular data centers for [5:25] defense agencies, and Nvidia wants in on [5:28] the government checks. Andreil touts its [5:30] modular data centers as bringing quote [5:33] processing to the war fighter end quote. [5:35] And Armada says that its data centers [5:37] are meant for places like US Navy combat [5:39] ships and deserts in the Middle East. [5:42] Nvidia's pivot towards becoming a [5:43] military contractor isn't surprising [5:46] given Jensen's affection of military [5:48] money, [5:48] >> the wealthier our nation, [5:51] the more we can fund the mightiest [5:54] military on the planet. [5:55] >> I'm happy that that uh our military is [5:58] going to use AI technology for defense. [6:01] And so we want America to be the [6:02] wealthiest country so that we can fund [6:04] the mightiest military. And I I think [6:06] that that is our way of contributing to [6:08] national security. and given his seeming [6:10] support for the war in Iran. And [6:12] >> I believe at the end of the war, Middle [6:14] East will be more stable than before. [6:16] And so if we were there, if we're [6:18] considering it before, we should [6:20] absolutely be considering it after. And [6:21] so I'm 100% in on that. [6:23] >> We brought you this video with [6:24] store.gamersac.net [6:26] and our push to own your own media. [6:28] We're making Blu-ray and music CD disc [6:30] packs fully inhouse. There's no DRM, so [6:32] you can rip them to your Plex or [6:34] Jellyfin servers if you want. And we're [6:35] 3D printing the disc cases fully in our [6:37] own office. Burning the Blu-rays and CDs [6:39] at G&HQ with a disc duplicator [6:46] and writing the USB drives with our [6:48] movies on them using a duplicator that [6:49] we bought. You can get a copy of our not [6:51] quite award-winning documentary on the [6:53] AIGPU Black Market, although it was a [6:55] finalist for an excellence in reporting [6:57] journalism award. So far closer than [6:59] we'd have ever expected. The black [7:00] market movie comes with two Blu-ray [7:02] discs, including hand transanscribed [7:04] human-crafted closed captions, a [7:07] human-made mini magazine booklet with [7:09] it, and optional USB drives. Our special [7:11] reports Blu-ray pack includes [7:13] educational deep dives like our staff [7:15] picks of the factory tour series [7:17] available with an autographed and a USB [7:19] insert option alongside our EVGA finale [7:21] report and educational engineering and [7:24] technical interviews as a disc pack. [7:25] Finally, we're currently printing our [7:27] cyber dystopia music CD cases for the [7:30] album that we hired Aaron Pauly to make [7:32] for us. Mixing a cyberpunk dystopian [7:34] themed electronic track with a fade to [7:36] chip tune NES and Sega era music. We [7:39] used these tracks for several of our [7:41] documentaries like our Sega Police raid [7:43] investigation. And now you can get them [7:45] on an audio CD with a printed vapor wave [7:47] smiley case that's made here in our [7:50] office using filament made in New [7:52] Jersey. This also has an optional USB [7:54] insert for audio files. Or of course, [7:56] you can grab our Microsoft t-shirts, [7:58] which we've just restocked now finally [8:00] in all sizes. Head over to [8:02] store.gamersex.net to help us out [8:04] directly. Thanks for your support. Now, [8:06] back to it. So, Nvidia is far from the [8:07] gaming first company that it once was. [8:09] again putting us and others in our space [8:12] who've reported on them in a unique and [8:14] weird position as compared to mainstream [8:16] media where what they're doing now might [8:18] be more appropriate to cover there. But [8:21] it came from gaming. And the weird thing [8:24] is like Jensen said, not only is it not [8:27] a gaming company anymore, it might not [8:29] even be a chip first company in the [8:31] future with its look at software [8:33] licensing. Jensen does however hearken [8:36] back to Nvidia's gaming roots to at [8:38] least show where the company came from. [8:40] >> And we started out building a graphics [8:41] chip for gamers and I'm still proud of [8:44] that. [8:44] >> Before talking about how Nvidia is [8:46] building infrastructure for the entire [8:48] planet. [8:48] >> Nvidia is also the infrastructure for [8:49] the planet. We have such a global reach [8:52] and and there's buildout demand all [8:55] around the world. But Nvidia today is [8:58] really a full stack endto-end complete [9:03] AI factory platform. [9:05] >> After all, at this point, Jensen wants [9:07] people to believe that Nvidia's GPUs are [9:08] so important that they are equivalent to [9:12] GDP for a nation. [9:13] >> So the first thought is that compute [9:17] equals revenues. Now the big idea of [9:20] course compute equals GDP that we also [9:23] know. Compute equals a country's GDP. I [9:28] am certain compute equals revenues. I'm [9:30] certain also that compute equals GDP. [9:33] >> He's on a global sales tour to convince [9:35] all the major economies that whatever it [9:38] is he's talking about. He's right. The [9:41] new key is licensing software and [9:43] support per GPU per year so they can [9:46] generate money at Infinite item for [9:49] devices that they sell, especially [9:51] supply constrained. This allows them to [9:53] continue boosting the revenue beyond the [9:55] already unfathomable figures that [9:57] they're at. And in addition to the [9:59] rentals, we are also going to be [10:01] covering the [10:03] new report on Mega Speed and the [10:06] allegations relating to spying. Uh some [10:10] news on smuggling because it didn't stop [10:13] when we released our documentary. Uh and [10:16] also we'll be talking about the defense [10:19] contracting interest from Nvidia. [10:22] It's not not the same as it used to be [10:25] covering GTC, that's for sure. So, today [10:28] we're going to talk about the continued [10:29] announcements from Nvidia and evolution [10:31] of the company. Nvidia today is [10:32] unrecognizable to the Nvidia of 10 years [10:34] ago. In its fiscal year of 2016, Nvidia [10:37] made about $5 billion in revenue total. [10:40] With $1.4 billion in revenue coming in [10:42] the fourth quarter, Nvidia still put [10:44] gaming first back then. That was roughly [10:47] the 1080 Ti era, the 1080 era. It was a [10:50] high point for Nvidia gaming, although [10:52] it had already begun to talk more about [10:54] the growth in AI, mostly referred to [10:57] then as deep learning and machine [10:59] learning. In the prior quarter this [11:01] year, Nvidia made $96.2 billion in [11:04] revenue with $89 billion coming from its [11:07] data center business. Since a quarter is [11:09] about 90 days, Nvidia generated more [11:11] than $1 billion in revenue per day. So, [11:15] to link it back to the highs of Nvidia's [11:18] gaming dominance in that 10 series, in [11:21] the span of 10 years, Nvidia now grosses [11:24] more money in 2 days than it used to [11:29] make in an entire quarter. [11:33] That's insane. But that's where we are. [11:36] And I think even if you inflation adjust [11:38] that, it's probably not going to change [11:41] the figures a whole lot. So, Nvidia CEO [11:43] Jensen Juan traveled constantly this [11:46] year to convince governments and large [11:48] companies and defense contractors that [11:50] he's got something important. And [11:53] Nvidia's got wares if you have coin. He [11:56] visited China, South Korea, the United [11:57] Kingdom, Sweden, France, Taiwan, [11:59] Germany, Japan, Kazakhstan, and probably [12:01] more places that we're forgetting he [12:02] visited because we're not his travel [12:04] agents. In October, the leather jacketed [12:06] globe trotder visited Europe for GTC [12:07] Berlin. Nvidia unveiled a smorges board [12:10] of news at the event, much of it focused [12:12] on European partners. However, before [12:14] GTC began, Nvidia played back the gaming [12:17] hits. Research group Semi analysis [12:19] reported that Nvidia had refused access [12:21] to multiple Neoclouds at GTC Berlin. [12:23] Neoclouds rent out GPUs for AI [12:26] processing and computing power like a [12:28] GPU as a service for AI. Nvidia has [12:31] backed many Neoclouds with billions of [12:33] dollars in investment money including [12:35] Cororeweave, NScale, Crusoe, Irene, [12:38] Nbius, Lambda, and Fermas. Nvidia [12:41] created the Neocloud industry. Just ask [12:44] Jensen. [12:45] >> In the case of clouds, if we didn't [12:47] support Cororeweave to exist, these [12:50] Neoclouds, these AI clouds wouldn't [12:52] exist. If we didn't help Cororeweave [12:56] exist, they would not exist. If we [12:58] didn't support Nscale, they wouldn't be [13:01] where they are today. If we didn't [13:02] support NBS, they wouldn't be where they [13:04] are today. Now they are they're doing [13:07] fantastically. Now it seems like Nvidia [13:09] will flex its market power when those [13:11] Neoclouds try to branch out from Nvidia [13:13] and derisk themselves by using other GPU [13:17] providers. This is an allocation play [13:20] and it's the same play that we've [13:22] covered for over a decade now with the [13:24] most famous of the coverage being EVJ's [13:27] departure from Nvidia as a partner. The [13:30] play is that with the limited allocation [13:32] of GPU silicon, anyone who uses or sells [13:35] or makes money off of that silicon will [13:37] scramble to get it and they will [13:38] sacrifice other partnerships if they [13:40] have to because the Nvidia one is what [13:42] gets them the most money. I covered this [13:44] a lot with laptops for example where [13:47] some companies in the past years had [13:49] told us that they weren't making AMD [13:51] laptops for a particular generation [13:53] because if they did their allocation [13:55] they claimed from Nvidia would be [13:56] reduced and Nvidia sells so many laptops [14:00] that they were willing to cut AMD out [14:02] from their production to secure that [14:04] supply. That's the allocation play. [14:07] That's what's being alleged is being [14:08] used here now with Neoclouds. According [14:11] to Semi analysis, Nvidia banned multiple [14:13] Neoclouds from attending GTC Berlin [14:16] after they added AMD GPUs, TPUs, and [14:18] other non- Nvidia chips to their cloud [14:20] offerings. Semi-analysis wrote on [14:22] Twitter that quote, "Executives feel [14:24] Nvidia has been using high-pressure [14:26] tactics to keep them Nvidia only. As [14:28] such, Neocloud executives are starting [14:30] to consider offering TPUs or AMD GPUs [14:33] quietly to avoid getting pressured by [14:36] Nvidia and to avoid the feeling that [14:38] Nvidia will retaliate." End quote. And [14:40] if you want to learn more about this [14:41] type of play, you can watch our EVGA [14:44] Nvidia series where we broke news on [14:47] EVJ's departure from Nvidia as a [14:49] supplier and talked about a number of [14:52] unique details to our report that we [14:54] were able to obtain for it. So this [14:55] isn't a new play for Nvidia. It's just [14:57] at a much greater scale now with more [14:58] money involved and at some level [15:01] government interest involved. Like the [15:03] gaming AIB partners, these Neocloud [15:04] companies rely on Nvidia for allocation [15:06] from Nvidia to purchase their GPUs. For [15:08] example, NScale primarily relies on [15:10] Nvidia GPUs for its infrastructure. In a [15:12] financial filing, NScale said that [15:13] Nvidia's allocation priorities their [15:15] quote and quote willingness to supply [15:17] GPUs and quote represent a risk factor [15:20] that could impact ncal's ability to [15:22] fulfill its commitments. Similarly, [15:24] Nvidia sugar baby Cororeweave says all [15:25] the GPUs used in its infrastructure are [15:27] Nvidia GPUs. By using only Nvidia GPUs, [15:30] Coreeave is subject to whatever Nvidia [15:32] wants to do. In a quarterly earnings [15:34] report, Cororeweave said its reliance on [15:36] a limited number of suppliers increases [15:38] risks such as quote consolidation in the [15:40] industry, reduced control over [15:42] production costs end quote and the [15:44] potential for quote binding price or [15:46] purchase commitments with our suppliers [15:47] at higher than market rates end quote. [15:50] Now, unfortunately for Cororeweave, [15:52] maybe fortunately for future regulators, [15:54] Cororeweave owes its existence to [15:56] Nvidia. [15:57] >> If we didn't help Cororeweave exist, [15:59] they would not exist. That doesn't sound [16:01] like a future antitrust issue. These [16:03] Neoclouds though, they're at Nvidia's [16:05] mercy, just like the gaming companies [16:07] that we've covered forever now. They're [16:09] all competing with each other for GPU [16:11] allocation and for pricing, except the [16:15] competing for pricing might actually be [16:17] bidding higher to make sure they get [16:19] some. We're not going to talk about [16:20] every one of Nvidia's GTC Berlin [16:22] announcements. In short, Nvidia rounded [16:24] up a grab bag of news for many of its [16:26] favorite topics like robotics, [16:28] autonomous vehicles, AI safety, and AI [16:30] agents. What we are going to talk about [16:32] is Nvidia's further steps to becoming a [16:34] defense contractor. We think Nvidia has [16:37] embraced the military recently. In May, [16:39] Nvidia partnered with the Department of [16:40] Defense along with seven other companies [16:42] to deploy its AI models on the agency's [16:44] systems. Jensen also explicitly voiced [16:47] his support for the Iran war. I believe [16:50] that there's a reason we went to war and [16:52] I believe at the end of the war Middle [16:54] East will be more stable than before. [16:56] >> At GTC Berlin, Nvidia introduced the DSX [16:59] pod which Nvidia calls a quote reference [17:01] design for complete Ver Rubin AI [17:03] factories built by partners and modular [17:05] containers integrated and validated [17:07] before delivery. End quote. Nvidia said [17:09] the units reduced deployment time from 6 [17:12] months to as little as 14 days. The [17:14] units arrived ready to be connected to a [17:17] local site's power, cooling, networking, [17:19] and surrounding infrastructure. In the [17:20] keynote, Jensen said that these [17:22] containerized data centers would [17:23] transform the supply chain. [17:25] >> The speed by which we can turn on go [17:27] from installation, arrival of computing [17:30] to production of tokens will go from [17:33] many months, [17:35] we think something along the lines of [17:37] four to five months down to about a [17:38] month. Okay. So, this is a complete [17:41] transformation of the world's supply [17:43] chain and this will change the way that [17:45] people install and build data centers. [17:48] >> While Jensen didn't explicitly say this [17:50] in the keynote, it does seem like many [17:53] of these data centers are destined for [17:55] conflict areas. At the bottom of its [17:57] press release, Nvidia specified multiple [17:59] military contractors, Anderil and [18:02] Armada, as its partners for these new [18:04] modular data centers. Andreil is a major [18:07] contractor for the US Department of [18:09] Defense and the company makes billions [18:11] from supplying the government with [18:12] drones, missiles, and surveillance [18:14] systems among other military products. [18:16] This year, Andreil landed contracts with [18:19] the US Navy to increase submarine [18:21] production and make cruise missiles for [18:22] the Department of Defense. Andreil also [18:25] brings data centers to the battlefield. [18:27] In a LinkedIn post, Anderil VP Tom Keane [18:30] said that the company is quote [18:31] designing, building, and scaling Menace [18:34] deployable data centers that support [18:35] classified work for defense and [18:37] intelligence customers to Nvidia's DSX [18:40] modular effort. End quote. Kon also said [18:42] that Andreel's designs enable defense [18:44] customers to use quote the latest AI [18:46] hardware with their classified data [18:48] deployed in the most challenging [18:50] environments at the tactical edge. End [18:51] quote. Anderil's Menace page says Menace [18:54] provides quote deployable edge [18:56] infrastructure end quote for quote [18:58] contested environments end quote in [19:00] other words to just distill deployable [19:03] edge infrastructure for contested [19:04] environments data centers for war zones. [19:08] That seems to be what they're talking [19:09] about. Armada makes similar modular data [19:11] centers and has a lengthy relationship [19:14] with Nvidia. The Nvidia Inception [19:16] program for startups welcomed Armada in [19:18] 2024. In a release about supporting [19:20] Nvidia's DSX pod, Armada said that it [19:23] and Johnson Controls will quote [19:25] engineer, manufacture, integrate, and [19:28] deploy prefabricated SI. It's not AI, [19:31] it's SI. That's the other thing. We've [19:34] changed the name officially today. [19:36] >> We've changed the name, Steve, [19:38] officially to SI. Okay. Sai like Sai. SI [19:43] instead of AI [19:44] >> factories built on the Nvidia Ver Rubin [19:46] platform with cabling, wiring, piping, [19:48] and functional validation completed [19:50] before a single unit leaves the factory. [19:52] End quote. Armada also specified that it [19:54] will use the DSX pod reference design [19:57] for its Gallion Forge factories. [19:59] Armada's Gallant product page says that [20:01] its modular data centers are intended [20:02] for the energy, manufacturing, mining, [20:05] and emergency response industries in [20:07] addition to defense. So, if Nvidia's AI [20:10] GPUs, sorry, SI wouldn't want to be an [20:13] enemy of the state aren't in war zones, [20:17] maybe they'll be in oil rigs. [20:19] >> Drill, baby drill. [20:19] >> Armada's Leviathan data center is the [20:21] largest form factor of Gallant. [20:23] According to the product page, Armada [20:25] uses up to 576 Nvidia GB300 NVL72 GPUs [20:30] per unit with they say a 1.77 megawatt [20:33] IT envelope. Full Leviathan deployments [20:35] are optimized for 5 megawatts to 20 [20:37] megawws, but they can scale to 50 [20:39] megawws. That's a lot of GPUs for a [20:42] container. And Armada share something [20:44] else in common. Both companies have [20:46] taken investment money from Peter Teal's [20:48] founders fund, which also funded the [20:50] original ID verification service that [20:52] Discord wanted to use, Persona. Teal's [20:55] funds have often been used for military [20:57] and for surveillance companies. [20:59] Increasingly, that has included ID [21:01] verification solutions for the internet [21:04] and for websites or web services. The US [21:06] government, including efforts from both [21:08] the Democrats and the Republicans, has [21:11] been pushing legislation to increase [21:13] identity verification use on the web, [21:16] including NY Kosa in New York, which we [21:19] covered separately on our GNCA channel. [21:21] Teal is also one of the co-founders of [21:23] surveillance enabler Palanteer which [21:25] Nvidia Harts we didn't photoshop that [21:28] in. They actually said Nvidia Harts [21:30] Palanteer and his VC fund has provided [21:32] startup money for flock surveillance. So [21:35] Nvidia is in bed with the military [21:37] industrial complex and surveillance [21:39] companies. But while at GTC Berlin and [21:42] while it caters towards defense [21:44] agencies, the company had at least a [21:46] cursory mention of RTX Spark in the [21:49] Berlin keynote. Fittingly, RTX, also the [21:52] stock ticker for the Rathon military [21:56] defense, you know, contractor. Uh, but [21:58] that's a conspiracy theory for another [22:00] time. Nvidia rolled out laptops, [22:03] desktops, and a workstation with RTX [22:06] Spark again for the European audience [22:08] this time to talk about the expensive [22:12] devices that it's making like the $7,000 [22:14] laptops. We started to think what can we [22:16] do to re-imagine the personal computer [22:20] for this new era where the personal [22:22] computer is not just your tool your [22:24] personal computer tool but your personal [22:28] agent or your personal assistant and the [22:31] way you interact with this computer [22:34] would be essentially a companion [22:37] somebody who helps you do work. This is [22:41] one pedlop [22:43] pedlop. If you would like to have a [22:44] companion workstation that runs Katilla [22:47] and everything do the and everything [22:49] Autodesk and you know Adobe and Cadence [22:53] and Synopsis, you can always get [22:55] yourself a DGX station that runs Windows [22:59] and at least Jensen remembered desktops [23:01] this time sort of. [23:03] >> And of course there's also the desktop [23:05] version. [23:05] >> That was literally the extent of his [23:08] memory that they exist. Nvidia also said [23:09] that Underdog, a new desktop agent for [23:12] running AI locally and privately, [23:14] allegedly, is available for RTX Spark. [23:17] Nvidia announced RTX Spark earlier this [23:19] year, describing it as a quote new [23:21] beginning for personal computers end [23:23] quote that it worked on with Microsoft. [23:25] Since then, Nvidia has been on a keynote [23:28] show to show off RTX Spark devices to [23:30] new audiences. However, Nvidia's der of [23:33] consumer announcements for its GTC [23:35] Berlin event is telling as the company [23:37] moves further and further away from [23:39] selling hardware to actual consumers. [23:41] Nvidia even buries its gaming and [23:43] consumer revenue within the broader edge [23:45] computing category these days, as we've [23:47] reported, meaning that we don't know how [23:48] much revenue NVIDIA even makes from [23:50] selling devices to consumers anymore. [23:52] Nvidia's edge computing category [23:54] includes automotive, robotics, and AI [23:56] ran base stations. Moving away from [23:58] Europe and towards Asia, Nvidia's [24:00] attitude towards AIG GPU smuggling has [24:03] changed a lot. They used to say there is [24:06] no evidence of chip diversion, that it [24:08] was fake news. Literally, that was one [24:11] of the phrases that they used. Now, [24:13] Nvidia wants us to know that it's just a [24:16] just a teeny bit of smuggling. Just a [24:18] little bit of smuggling, but don't pay [24:19] much attention to it. The United States [24:21] placed export controls on Nvidia's most [24:23] powerful AI GPUs to bar Nvidia from [24:25] shipping them to China and other [24:26] countries like Russia. However, we know [24:29] Nvidia's GPUs make it to China because [24:31] we literally wrote a book about it now [24:33] available as an insert with our [24:35] documentary on Blu-ray on [24:37] store.exus.net. [24:39] But we also know that Nvidia's chips [24:40] show up in Russian drones. Nvidia [24:42] shareholder and President Donald Trump [24:44] worked out a deal with Jensen where [24:46] Nvidia can sell some chips like the H200 [24:48] to China, provided Uncle Sam gets a 25% [24:52] cut of Nvidia's revenue in return. China [24:54] has its own rules to limit Nvidia's chip [24:56] imports to the country since China wants [24:59] to foster now its own domestic chip [25:01] manufacturers like Huawei. Smugglers [25:03] still find a way in, of course, to [25:04] circumvent the rules from both countries [25:07] and import Nvidia's AI GPUs into [25:09] mainland China. And again, just to [25:11] continue supporting our own work here, [25:13] we witnessed and documented the [25:15] smuggling of GPUs in our actually [25:18] journalism award finalist, although not [25:21] recipient. It's the almost award-winning [25:23] documentary, uh, the AI GPU black [25:25] market, which again you can get on the [25:27] store with a case that we've entirely 3D [25:30] printed in house with discs that we're [25:32] burning in house on Blu-ray for it. You [25:34] can get the USB drive insert if you want [25:36] and uh, also a booklet insert that we're [25:39] doing. But anyway, that's on [25:40] store.cameac.net [25:41] if you'd like to support us there uh for [25:44] further investigations that we're [25:45] running. Despite our own evidence and [25:47] report, including our own interview with [25:49] a GPU smuggler in the United States and [25:51] reporting from outlets like Reuters and [25:53] the Financial Times, Nvidia has [25:55] continued to deny evidence of chip [25:57] smuggling. Mostly Jensen issued a number [26:00] of these denials in 2025. [26:02] >> Over the years, people have speculated [26:04] about diversion. Um, we've chased down [26:06] every single uh concern and uh we've [26:10] repeatedly uh tested and sampled uh data [26:14] centers around the world and uh found no [26:17] uh diversion. Nvidia's chips are not [26:20] little tiny potato chips. These are two [26:23] tons. Each one of our GPUs is a rack [26:27] that's two tons, two tons large or the [26:29] even the smaller versions of our racks [26:31] are hundreds of pounds. And so it's not [26:34] easy to smuggle. [26:35] >> Every so often somebody says, you know, [26:37] you know, these GPUs are being smuggled. [26:39] I really would love to see it. [26:41] >> But as the walls close in, that tune has [26:43] been changing. On October 1st, the [26:45] Department of Justice arrested Greg Lei [26:47] in California for smuggling more than [26:50] $300 million worth of AI servers into [26:52] China. They allege the government [26:54] charged Lei with conspiracy to violate [26:56] the Export Control Reform Act, outbound [26:59] smuggling, and conspiracy to commit [27:01] money. Lei and his co-conspirators [27:04] purchased and shipped export controlled [27:06] servers to Malaysia and Singapore from [27:08] 2023 to 2024 and then illegally exported [27:12] the servers to China. The DOJ didn't [27:14] name Nvidia in its release, but the [27:16] agency did in the indictment. The [27:18] government said that Lei purchased [27:20] Nvidia's A100s, H100s, RTX490s, and RTX [27:24] 590s from several US manufacturers. An [27:26] Nvidia spokesperson told our BFF [27:29] Bloomberg that quote, "Governments and [27:32] media have found insignificant diversion [27:35] of controlled US products, a drop in the [27:38] bucket compared to the ocean of domestic [27:40] compute China already has." US [27:43] prosecutors allege less than 1/ half of [27:45] 1% of Nvidia products have been [27:47] diverted." End quote. Taiwan has [27:50] likewise arrested an Nvidia employee in [27:52] connection with the smuggling of entire [27:54] servers. So Nvidia's messaging about [27:57] smuggling has changed from no evidence [27:59] to chip diversion to maybe a little bit [28:01] of evidence, but calling it a drop in [28:03] the bucket and it also being $300 [28:06] billion, which they called [28:08] insignificant, [28:09] is something that only Nvidia could [28:12] attribute to $300 million worth of [28:14] anything. If in the story of Greg Lei, [28:17] the defendant is found guilty, the case [28:19] provides evidence for Southeastern Asian [28:21] countries being used by chip smugglers [28:23] to import AIGPUs into China. In early [28:26] 2025, US regulators began to scrutinize [28:29] Nvidia's exports to Singapore after [28:31] local authorities arrested chip [28:32] smugglers and Nvidia reported a [28:34] significant amount of revenue in the [28:36] country. Nvidia used to segment how much [28:38] revenue it made from Singapore. In an [28:41] August 2025 filing, Nvidia said, quote, [28:43] "Customers use Singapore to centralize [28:45] invoicing, while our products are almost [28:47] always shipped elsewhere, over 99% of [28:50] controlled data center compute revenue [28:52] build to Singapore was for orders from [28:54] US-based customers for the second [28:55] quarter and first half of fiscal year [28:57] 2026." End quote. Nvidia subsequently [29:00] removed its revenue attributed to [29:02] Singapore in following quarters. We [29:04] think Nvidia wanted to divert the [29:05] attention away from Singapore. Later in [29:08] October, based on a report from the [29:10] information uncovered by Tom's Hardware, [29:12] an Israeli security company said that it [29:14] identified Nvidia's export controlled [29:16] B300 chips within a Chinaowned telecom [29:19] company. Nvidia once again called the [29:21] evidence of chip diversion a drop in the [29:23] bucket. Quote, with tens of millions of [29:25] GPUs and service in approved countries, [29:27] this report again demonstrates that [29:29] smuggling is a non-factor blocked by [29:32] governments on both sides of the globe. [29:33] End quote. Previously, Nvidia used the [29:35] word non-starter, but Nvidia has a [29:37] unique definition of non-factor. The [29:39] federal government has arrested chip [29:40] smugglers for more than a year now in [29:42] multiple instances, including one of [29:44] Super Micro's co-founders. And Taiwanese [29:46] authorities even charged an Nvidia [29:48] employee. Now, that doesn't sound like a [29:50] non-factor to us or a non-starter as [29:52] they used to say before they weirdly [29:54] changed the language to non-factor. And [29:57] Nvidia has also taken a page from [29:58] Trump's book and used the phrase fake [30:01] news in the past in response to some of [30:02] these reports. So they seem to just try [30:03] and weasle around it every time. Nvidia [30:06] has at this point undergone a complete [30:08] transformation and soon Nvidia seems to [30:11] suggest that it may not even be [30:12] primarily a chip company. Now we think [30:14] they'll stay a chip company of course [30:16] but it is possible they might get a lot [30:18] more revenue or at least a large [30:19] increase in revenue from an Oracle style [30:22] software and licensing push per Jensen. [30:24] >> You know one of these days you know [30:25] people might not categorize us as a chip [30:28] company anymore. It's not news that [30:29] Nvidia hardly cares about or maybe [30:31] doesn't care at all about gamers at this [30:33] point. Uh, but it is getting worse on [30:36] the consumer side in general. Honestly, [30:38] I don't even know how the people that we [30:40] know from the gaming side of Nvidia even [30:42] have jobs anymore other than the fact [30:44] that maybe Nvidia forgot that they [30:45] worked there. I'm not really sure. But [30:47] everything Nvidia does is to accelerate [30:50] AI adoption. Jensen's rhetoric shows [30:52] he's thinking about Nvidia's path going [30:55] forward as continually being with AI and [30:59] maybe the military. And we think Nvidia [31:02] is becoming a defense and surveillance [31:05] contractor. I've told the story before, [31:06] but I remember in maybe it was 2012, [31:09] might have been 2014 when I went to [31:11] Nvidia's campus for the first time and [31:13] someone giving that tour proudly showed [31:15] us. It's looking back it's like even [31:17] more creepy than I think I realized at [31:19] the time. But you walk in, they've got [31:21] this display of displays uh showing I [31:24] think it was like airport footage with [31:26] the cameras doing facial recognition. [31:28] This was early. This would have been [31:30] like maybe 15 years ago and they were [31:33] already showing Nvidia use in facial [31:36] recognition and surveillance systems. I [31:39] didn't connect those dots at that time. [31:41] To me it was boring. It's like all right [31:43] whatever. Show me the gaming stuff. I [31:45] wasn't I wasn't really drawn up enough [31:47] yet to recognize the gravity of what [31:49] they were showing me, but here we are. [31:52] Throughout his global tour, Jensen has [31:54] been on a charm offensive to present a [31:56] fun, optimistic future for AI. He shows [31:58] off these cute robots on stage, as he [32:01] calls them. He eats with people in the [32:03] streets. He appears on variety TV shows. [32:06] Now, he wants people to be optimistic [32:08] about AI and start loving AI at an early [32:10] age. Of course, we are doing this [32:12] because we're optimistic about the [32:14] future. We're doing this because we [32:16] believe in a better future for society, [32:18] for our families, for our children, for [32:19] each other. That's the reason why we're [32:21] doing it. [32:21] >> We need people to invent a future that's [32:23] optimistic. But we have to do it [32:25] responsibly. Can you create a future [32:27] that people want to be part of if you're [32:29] not optimistic? [32:30] There is no question that this [32:32] technology has great potential to do [32:34] good, but we have to advance it safely [32:37] and responsibly. It is the [32:39] responsibility of the leaders that are [32:41] creating creating it to do it safely, [32:43] you know. So, it's good that we warn [32:45] people. [32:45] >> What I would advise all children is [32:47] this. Engage AI. [32:50] >> Use it. Use it to empower yourself. Use [32:53] it to lift yourself. So he's going for [32:55] the get them hooked while they're yan [32:57] approach which I think you get one [33:00] generation deep into that where it's [33:02] just integrated with youth culture and [33:05] learning and whatever and suddenly [33:08] everyone else becomes the old man [33:10] yelling at the data center shaped cloud. [33:12] Why are you so weird about using your ID [33:14] to log into websites grandpa? It's [33:17] normal. Everybody does that. It's not a [33:20] problem. That's the direction I think [33:21] we're going with it. Jensen's public [33:23] image contrasts with what we know of and [33:25] hear about Nvidia's internal practices [33:27] as well. He fronts this very friendly [33:31] nice guy. But the reality is over the [33:34] years Jensen's former staff have [33:36] repeatedly described him to us in [33:38] conversations as cutthroat and as [33:40] brutal, including in all hands meetings [33:42] where he once allegedly asked employees, [33:45] "Who did this to me about a failed [33:47] launch?" until he could, as I was told [33:49] at least, single people out. That was as [33:52] the story was told to us. Nvidia is the [33:54] same company that it has been for years [33:57] now. We're just seeing it more in some [34:00] of the political spaces. Nvidia has long [34:02] been accused of taking punitive actions [34:04] toward partners that tried to use AMD [34:06] GPUs, like with the story of its GeForce [34:09] partner program in 2018. Nvidia bends [34:11] its messaging, we think, to avoid tough [34:13] questions about chip smuggling. It's [34:16] trying to make a sale to anybody, [34:17] including military contractors. Uh and [34:20] throughout all of this, Jensen will [34:21] fight tooth and nail to avoid [34:24] regulation. [34:25] >> Uh I think the the balance of of precise [34:29] regulation uh don't regulate [34:31] hypothetical theoretical harm, regulate [34:35] actual and pragmatic harm. Right now the [34:38] technology is in its formative stages [34:40] and and the um the advice uh that I [34:43] would have is is um uh to regulate [34:47] practical and actual harm and not uh uh [34:53] regulate theoretical and hypothetical [34:55] harm. And the reason for that is of [34:57] course the technology is still [34:58] advancing. We want to make sure that the [35:00] technology advances in a safe way and [35:02] and the the more advanced it is, the [35:04] safer and more practical it could be, [35:06] more useful could be. [35:07] >> However, Nvidia's ambitions may be [35:09] catching up with them. On October 28th, [35:12] Reuters reported that the Department of [35:13] Justice is poking around about potential [35:15] issues with Nvidia's purchase of AI [35:18] development host Hugging Face. According [35:20] to Reuters unnamed sources, the DOJ's [35:23] questions cover quote, "Nvidia's [35:24] acquisitions, investments, sales [35:26] practices, and whether the chip designer [35:28] has exerted pressure on its customers." [35:30] End quote. Just a just a chip in there, [35:32] DOJ, I can give I can I can help with [35:36] that one. Let me know. Anyway, the [35:39] officials ask, quote, whether the deal [35:41] would let Nvidia capture a fundamental [35:43] distribution point for AI, end quote. [35:45] Nvidia declined to comment to Reuters. [35:48] Senator Warren has also petitioned for [35:49] Jensen to speak to Congress and [35:51] committees multiple times. Most recently [35:53] asking Jensen to speak in front of [35:55] Congress. This follows another [35:57] investigation into Nvidia by the Justice [35:58] Department. According to a New York [36:00] Times investigation, the government is [36:02] looking into whether Nvidia tried to get [36:04] around antitrust rules relating to its [36:06] Grock with a Q agreement. We'll see [36:08] where these inquiries go, if anywhere, [36:11] but in the very least, some regulators [36:12] are paying attention to Nvidia. In its [36:14] last earnings, Nvidia said that it [36:16] expects to grow its revenue by 70% in [36:18] the next fiscal year. They of course [36:19] also talked about the rentals and [36:21] software rentals which may be a part of [36:23] future revenue growth. Further, Nvidia [36:25] forecasts revenue of $ 108 billion plus [36:27] or minus a few billion dollars of [36:29] rounding error at 2% in the next [36:31] quarter. Nvidia reports its next results [36:33] on November 17th. If Nvidia meets those [36:35] figures, it'll have another [36:37] record-breaking quarter. [36:38] >> But as as you know, Nvidia is also the [36:40] infrastructure for the planet. [36:41] infrastructure [36:42] for the planet. That's what Jensen seems [36:46] to want to provide. And if Jensen has [36:48] his way, no one will stop him. That's it [36:51] for this one for your update on the [36:53] Nvidia everything that they're doing. [36:56] It's unbelievable that this is where [36:57] we've gotten from 2016 where you look at [37:00] the [37:02] I was just happy to be able to find MSRP [37:04] $700 1080 Ti with shitty heat sinks on [37:07] them and rip them off and turn them into [37:09] hybrid mods with like a closed loop [37:11] liquid cooler strapped to the GPU and [37:13] then some shitty fan pointed at the VRM. [37:15] That's what I was happy to do with [37:17] Nvidia hardware 10 years ago. Now you [37:21] might be happy if you could even get [37:22] Nvidia hardware. Although depending on [37:24] your feelings about all this, maybe you [37:27] wouldn't be. But thanks for watching. Go [37:30] to store.gamersac.net [37:31] to support our deep dive reports like [37:33] this. These take a lot of focused effort [37:35] to put together and a lot of really sort [37:38] of specific uh research and [37:40] documentation. And so as always, we [37:42] could use your support to continue doing [37:43] these to document what these companies [37:45] are doing and do our investigations [37:47] which we have some other big ones coming [37:48] up like the data centers one uh and more [37:51] on smart TVs and spying. So check back [37:54] for all that. Go to store.camesac.net to [37:56] help us out directly like by grabbing [37:57] one of these documentaries on Blu-ray. [37:59] There's USB options on there if you just [38:01] want to plug a USB drive in. It is of [38:03] course available for free on the [38:05] channel. We've hired someone to help [38:06] with converting the whole documentary [38:08] into a booklet insert which is only [38:11] available obviously by buying it because [38:13] it's a physical booklet. Uh so we're [38:15] excited about those things and uh yeah [38:17] you can get it on the store. They ship [38:18] in December, start shipping in December. [38:20] It probably take us through January. But [38:22] uh that's it for now. Subscribe for more [38:24] and you can go to patreon.com/gamersex [38:26] if you want to throw a few bucks our way [38:27] that way. And we'll see you all next