Nvidia Just Bought The US Government For $4.95 Million

0:00 America sold China the most powerful AI chips on Earth… for a cut of the profits. That’s not a conspiracy. It’s official policy. After years of promising to cripple China’s AI industry Washington reversed course and allowed Nvidia’s chips for direct sale to Beijing. No espionage or cyberattacks. Just a lobbyist and ink drying on a contract. With a flick of a pen, the most strategically important technology of the 21st century landed back in the hands of America’s biggest rival. Because the AI Cold War was never really about stopping China. It was about who gets paid. Chapter 1: The Beijing Reversal The press release that announced the H200 deal in December 2025 reads like ordinary trade paperwork. At first glance.

0:45 Read it again, and something starts to feel different. Buried inside what is basically a one-page memo is one of the larger policy U-turns in American technology history. The kind of pivot that would normally roll out over months. Instead, it landed almost out of nowhere. The U.S. government would now take a cut of every Nvidia H200 chip shipped across the Pacific. A quarter on every dollar. The H200 isn’t something you’d find in a gaming PC. This is different. It’s the kind of hardware built for one job.. Training advanced AI systems at scale. There’s so much raw data passing through it that it behaves less like a chip and more like a small factory for intelligence. Stack a few thousand together and you get an industrial compute cluster that can train elite AI models in weeks rather than years. Under U.S. export controls, that kind of system

1:33 was never meant to end up in places like China. At least, that was the rule. But before we go any further, imagine this. You’re online every single day. You check your email, open a few apps, look things up, stream videos, maybe even do all of that while traveling. And to you, it feels totally normal. But behind the scenes, your internet provider, advertisers, network admins, and sometimes even governments can build a surprisingly detailed picture of what you’re doing online. Now, to be clear, no VPN can protect you from everything. You still have to be smart online. Don’t click suspicious links, don’t hand over personal information to sketchy emails, and definitely don’t trust the so-called Prince of Nigeria. But a good VPN is an important layer of privacy, because privacy shouldn’t be something you only think about after something goes wrong. It should be the default. And that’s exactly what Proton VPN is built for.

2:20 Proton VPN helps keep your browsing private wherever you are, whether you’re at home, traveling, or just trying to stop your online activity from being tracked. Their no-logs policy has been verified by independent auditors, and Proton is backed by a foundation dedicated to privacy, transparency, and user rights. Proton VPN is also fully open source, meaning its code is publicly available for inspection. So instead of just asking you to trust them, Proton gives people a way to verify what they’re doing. And privacy doesn’t have to mean slow. Proton VPN offers high-speed connections with VPN Accelerator, plus NetShield, which blocks malicious ads and trackers before they ruin your browsing experience. It also gives you more freedom online. You can connect through servers around the world, access content more securely while traveling, and use it for streaming, gaming, and P2P file sharing, all while keeping your online activity more private.

3:08 So whether you’re trying to stop advertisers from following you, keep your browsing private from your ISP, or just take back control of your online life, Proton VPN makes privacy simple. Click the link in the description or go to protonvpn.com/theinfographicsshow to get up to 70% discount when you sign up to Proton VPN 2-year plan. Back in 2023, the Bureau of Industry and Security drew a hard line. They are the ones who decide what technology can cross which borders. The concern was simple. China’s AI labs were making inroads in the AI arms race, and Washington was desperately trying to slow So officials started limiting how much computing power could legally be exported. The problem? The best chips were already way beyond those limits. So companies like Nvidia did what companies do… they adapted. They built a downgraded version

3:56 designed to slip just under the rules, keeping the market open without crossing the line. But even that workaround didn’t hold for long. Instead of a clean ban, Washington shifted to something more flexible, a one-year waiver system with individual approvals. In practice, the government itself would decide - case by case - who in China was allowed to buy advanced chips. The press called it the “China Chip Review.” The list of who qualified is the real story. Because the approved buyers weren’t obscure startups or low-risk firms. They included Alibaba, ByteDance, and Tencent. There were also research institutions with links to the People's Liberation Army (PLA), and state-backed cloud providers tied into China’s defense infrastructure. The same organizations American policy had spent two administrations trying to starve of compute were suddenly eligible buyers. Once the hardware leaves, the rules and

4:46 regulations don’t leave with it. Compute doesn’t stay supervised. A data center doesn’t announce what it’s training. The United States was now collecting revenue from selling hardware it had publicly called a national security threat just years before. What’s most telling is how the people who built the original blockade found out it had changed. According to later reporting, many of them found out the same way the public did, through news agencies. Beijing didn’t sit still either. Chinese regulators told state-linked firms to favor Huawei chips over American ones. But Alibaba, ByteDance, and Tencent moved fast on the H200 anyway, with reported orders running into the hundreds of thousands of units. So how did this move from unthinkable to operational without a single congressional vote? Just follow the money trail.

5:32 Chapter 2: The Tithe In August 2025, an agreement was already taking shape between Nvidia, AMD, and the executive branch. In exchange for export licenses, the chipmakers agreed to send 15% of their Chinese chip revenue back to the U.S. government. This wasn't a tax or a tariff. It was something genuinely new in American trade history. A voluntary revenue share, negotiated directly between private companies and the executive branch, all tied to approval for exports. Access, in exchange for a cut. For Nvidia, the numbers are staggering. China revenue from the H20 alone is measured in tens of billions. The 15% share, plus other commitments, comes in at $12 billion. That’s in the same range as the entire annual economies of places like Madagascar or the Bahamas. And none of it went through the usual political machinery.

6:22 What this really created was a precedent. Export controls stopped behaving like hard rules and started behaving like pricing terms. Something you could adjust, negotiate, trade against revenue. The December deal pushed it even further. With the more powerful chips, the surcharge reportedly climbed to 25%. The pricing model became obvious. Pay more, ship more sensitive hardware. The fine print added a 50% cap. Shipments to China can’t exceed half of what’s sold in the U.S. And the whole arrangement runs on a one-year clock, expiring in December 2026. After that, it gets renegotiated or rolled forward. But by then, the precedent will already be set. A government that funds itself partly from a corporation's exports has a financial interest in those exports continuing. And once that happens,

7:07 the incentive structure shifts. Blockades stop acting like a deterrence. They start behaving like checkpoints that collect revenue. What was sold to the public as smart policy looks, on closer look, more like a protection racket. To pull this off, Nvidia needed more than slides and a corporate affairs team. It needed people who could rewrite the rules from inside the building. And in 2025, it found them. Chapter 3: The Gatekeepers Nvidia, for most of its corporate life, was almost an afterthought in Washington. In 2024, its federal lobbying filings totaled roughly $640,000. By the standards of a corporation with a market value north of $3 trillion, that figure was nothing. Pharmaceutical firms one-tenth its size routinely spent 10 times that. In 2025, something changed.

7:53 Nvidia's lobbying spend climbed to $4.95 million, a jump of more than 600% in a single year. The main goal wasn’t tax policy or antitrust. It was the Bureau of Industry and Security and the export rules that decided which Chinese addresses could receive American silicon. A key hire landed at Brownstein Hyatt Farber Schreck, the Denver-based lobbying firm that has steadily embedded itself in Washington’s corridors. They created a dedicated team for the Nvidia account led by Ed Royce. Royce isn’t a peripheral figure. He spent years as chair of the House Foreign Affairs Committee, one of the key places where U.S. policy on technology and rival states is shaped. Which means he didn’t just understand export controls on paper. He understood how they actually move. Who writes the language before

8:40 it becomes policy. Which offices quietly steer decisions. And where the real pressure points sit. It’s the kind of experience that turns regulation into something you can work with, not just something you read. But companies hire former officials all the time, lobbying numbers go up every year, so why is this any different? The answer comes down to timing. The Brownstein hire, the lobbying surge, and the policy shifts didn’t unfold across a multi-year arc. It was a matter of months. Export categories that had been closed in May 2025 became unclear by August and were basically open by December. There’s a version of this story where it's a coincidence. But it doesn’t really read that way anymore. Because the record shows something too targeted to ignore. Specific pressure on specific rules. The exact levers that end up moving.

9:26 What they delivered was something more valuable than meetings. A foothold inside the rooms where licensing exceptions were decided. Once Nvidia's policy team had that foothold, the next part of the machine did most of the work on its own. Chapter 4: The Corporate Loyalty Scorecard According to Axios, a rating system exists within the West Wing, one that tracks corporate America. It covers 553 companies and trade associations, scored on how hard each one worked to support the administration's signature bill, the One Big Beautiful Bill. Companies are rated as strong, moderate, or low supporters. The factors include social media posts, press releases, ads, video testimonials, and attendance at White House events. Senior officials told reporters the document is evolving to cover other priorities. Those who score well are said to receive what staff describe as fast-track treatment.

10:15 From everything visible on the outside, Nvidia's score is exceptional. Their separate $15 billion domestic commitment landed on the right side of every metric the system appears to track. The company announced datacenters in politically meaningful states, gave executives plenty of face time at administration events, and tied their massive supply partnerships to the Stargate buildout. The connecting thread runs through the chief of staff’s office. Susie Wiles, who managed the 2024 campaign before taking the role, came to the White House after a long career in political consulting and lobbying work That background has shaped how some inside the system describe the current setup: a West Wing where senior staff are closely familiar with how corporate and policy interests move through federal decision-making. For Nvidia, the advantage doesn’t come from a

11:01 single approval or decision. It builds in the background, through how applications are categorized before they’re even reviewed. Over time, that classification determines how smoothly requests move through the system. A competitor who shows up to the licensing window with a stronger technical case but a weaker scorecard rating does not lose because their argument is wrong. They lose because the scorecard has pre-sorted the queue. People familiar with the process describe it in simple terms: by the time you’re debating the decision, the decision has often already been framed. Chapter 5: The Green Channel Nvidia doesn’t just lead the AI chip market. It dominates it. In the commercial sector used for training and running advanced AI systems, Nvidia controls roughly 94% of the market. AMD holds 5%, and Intel

11:47 takes most of the rest. So when export approvals become case-by-case decisions, the system isn’t really choosing between equal options. And that matters. Because modern AI isn’t just about hardware, it’s about what that hardware runs on. Customer demand is overwhelmingly for Nvidia silicon. That is what the Chinese labs have built their software stacks around. Switching it out isn’t like swapping brands. It’s closer to rebuilding your entire infrastructure while the system is still running. So even though, on paper, multiple suppliers exist, in practice the demand funnels toward one. Nvidia did not need a clear policy granting it preferred treatment. Its market dominance, combined with how deeply its software is embedded in the industry, meant that any export pathway for advanced chips would almost automatically route through it. So when a revenue-share system was introduced,

12:36 it didn’t need to explicitly pick winners. It just followed the path the market had already carved. On paper, the arrangement looked neutral, something any supplier could, in theory, participate in. In practice, there was only one meaningful option. And everyone else was left outside watching the terms get set. The knock-on effect makes the structure feed itself. Each chip shipped pushes Nvidia’s products further inside Chinese datacenters and generates revenue. A percentage flows to the U.S. Treasury, which strengthens the political case for continued shipments. But the rulebook still officially rests on technical benchmarks written into law. S o how, exactly, do you legally ship a chip that, by the explicit limits set in 2023, should be banned? Chapter 6: The Loophole Factory

13:22 This is where the system gets clever, and where the technical details start to matter. Inside the Bureau of Industry and Security, rules don’t only change through formal policy shifts. They also evolve through technical questions and guidance notes. These are documents written to explain how existing rules apply in practice. No hearings. No votes. Just clarifications that shape how the rules work. It’s where policy gets adjusted in real time… without ever looking like it was changed at all. And over time, those adjustments tend to move in one direction. As senior meetings between Nvidia and U.S. officials take place, new guidance keeps refining what counts as acceptable under export limits. Each update, on its own, looks minor. But together, they don’t just interpret the rules. They move the boundary, just enough for

14:07 the chips to keep moving through. None of this is technically illegal. Most of it is the system working as intended. The U.S. Department of Commerce has the authority to interpret its own export rules. That interpretation happens constantly through technical updates and guidance. The problem is what the combined effect of dozens of small clarifications looks like once you step back. Stacked together, they start to change what the rules actually are in practice. A fixed limit turns into a negotiated threshold. Technical specs that were meant to be objective limits start reflecting input from the very industry they govern. When a company can influence how the rules governing its exports are interpreted, it’s no longer just being regulated in a simple sense. It’s helping shape the rulebook it operates inside. The memory in Nvidia’s H200 chip is more than just a memory spec. In frontier model training,

14:57 it’s the difference between a system that can hold a top-tier language model in active context and one that can’t. It’s also the line between a chatbot and a cyberwarfare engine. By engineering chips right up to the very edge of the rules, Nvidia is now writing the working version of American export policy in real time. Chapter 7: The Shadow State Department Jensen Huang's 2025 calendar reads like a study in parallel diplomacy. The Nvidia CEO held high-profile international meetings in Taiwan, Japan, the United Arab Emirates, Saudi Arabia, and the United Kingdom. At each stop, he announced compute commitments and buildout deals at a scale that would historically have been worked out by trade representatives or cabinet secretaries. In several cases, the agreements the CEO announced set the agenda

15:43 for the nation-to-nation conversations that followed, rather than the other way around. The Gulf compute deals are perhaps the clearest example. Saudi Arabia's announcement of a sovereign AI initiative, anchored by Nvidia hardware, set the terms of the American foreign policy meetings that followed with the Kingdom. Nvidia built the policy and the State Department adapted to it. The same pattern played out in Abu Dhabi. The G42 partnerships and tied chip commitments came before, not after, the formal U.S. rulebook for AI cooperation with the UAE. It’s not how foreign policy is supposed to work. The Department of State exists to coordinate how the United States approaches strategic technology and its transfer abroad. But now, a company can announce commercial commitments first, and the government is left reacting afterward. They have to

16:29 adjust policy to match what’s already in motion. What emerges is an unusual kind of sovereignty. Nvidia is not a state in the traditional sense. It has no army and no treasury. But it does sit at the point where AI capability is actually determined through the chips that define what systems can and cannot do. And it has enough political access to ensure its commercial choices don’t just operate inside policy, they actively shape it. That’s a different relationship than regulation alone. It starts to look less like a company operating under rules, and more like a company operating alongside the formation of those rules. Not replacing the state. But, in key areas, setting the terms the state then responds to. Chapter 8: The Reckoning The first H200s under the new waiver are

17:15 starting to land inside Chinese datacenters, with reported orders already running into hundreds of thousands of units. The training runs will follow. The models that come out will end up applied to use cases spanning the full range, from commercial chatbots to military targeting systems. The compute, once delivered, becomes whatever the operator chooses to do with it. What makes this hard to unwind is that it’s no longer just policy, it’s income. Once the revenue starts flowing into federal planning, it stops being a clean on/off switch. Federal budget projections include it. Domestic buildout programs partly funded by it have constituencies that will defend it. The bureaucratic machine of the U.S. government has, in effect, taken a financial stake in continued exports of the very hardware it once labeled a

18:00 controlled good. Reversing the deal would require not just a political decision but the removal of revenue already locked into agency planning. When Nvidia raises its revenue projections, the government's projected share rises with them. It only strengthens the political case for expanding the waiver. The privatization of foreign policy stops being a metaphor and starts being a balance sheet relationship. Once a company has shown that exemptions to national security policy can be bought at a fixed price, the precedent becomes permanent. What’s emerging here looks less like a one-off deal and more like a template for turning regulatory pressure into a negotiated cut of revenue. And other industries are already watching closely. Pharma, satellites, advanced biotech… sectors where the same product can be both commercial and strategic. For decades, the rule was simple:

18:48 some technologies stay out of open markets. That rule hasn’t been rewritten. It’s just… stopped being enforced in the same way. Nvidia is just one player in the AI race. And no one wants to be the one that slows down. But an arms race only works if it’s sustainable. Right now, the spending is starting to look less like growth and more like pressure building in the system. And when it breaks, everyone feels it. Find out what happens next in “$115 Billion Burn Rate. The AI Bubble Just POPPED.” Or watch this video instead.