How Money Works on the US AI lobby gravy train, while China is overtaking the US in practical AI

**TL;DR:** The US AI industry has become a massive lobbying force, spending over $100 million in a single year to push the narrative that China must not surpass the US in AI, while simultaneously lobbying to sell China the very chips needed to do so, revealing a stark hypocrisy where the real goal is financial gain rather than genuine national security. ## The Rise of the AI Lobby America’s leading AI companies have quietly become one of the nation’s largest lobbying forces. In the past year alone, they spent over $100 million influencing policy. OpenAI’s lobbying expenditure surged dramatically year-over-year, likely marking its first major investment intended to generate financial returns. Nvidia, a company that had long stayed silent in Washington, multiplied its lobbying spending fivefold to ensure its voice was heard. These well-funded voices all push a single, overwhelming message: China must not be allowed to surpass the US in AI. The argument is framed as an existential threat: either we dominate, or China dominates. And if China beats the US on the ultimate goal of AGI (Artificial General Intelligence), the consequences would be catastrophic. So the lobby goes: overregulate AI, and China wins; under-regulate, and China wins; stop funding AI development, and China wins; sell cutting-edge chips to China, they win; don’t sell chips, they also win. This China-centric argument has made AI lobbying one of the most effective in Washington. And as absurd as it sounds, they might actually be right about the threat—but for all the wrong reasons. ## The Chip Contradiction Despite endless talk about Nvidia being indispensable in the current AI ecosystem, the company has admitted its market share in China has effectively dropped to zero. Chinese labs are already developing domestic alternatives, matching or even surpassing US models using supposedly inferior hardware and fewer resources. If it weren’t true, why would Nvidia admit it? Meanwhile, it’s becoming increasingly clear that current AI models may perform poorly in many advertised applications but are frighteningly effective for mass surveillance, warfare, cyber warfare, and propaganda. If this truly is an existential threat where we cannot afford China to gain an edge, we must at least understand the hypocrisy in our response. Herein lies the first massive irony: the same companies warning about the Chinese threat are the ones lobbying hardest to sell China the most advanced technology. For nearly two years, the headline message from Nvidia, OpenAI, and most major AI labs has been: “Letting China take the lead is a civilization risk we cannot allow.” Then, late last year, when the AI industry seemed to face a market correction, that message became more flexible. In December, the government announced that Nvidia would again be allowed to sell its H200 chips to China, with the rule taking effect at the Commerce Department in January. Nvidia itself describes the H200 as one of the most advanced mass-produced AI chips on the planet. Theoretically, shipments to China are capped at 50% of US sales and require inspection for transshipment. But in practice, just the H200 cap could allow up to 1 million chips into China, plus possibly another 1 million H100s. According to the Council on Foreign Relations, that’s enough to build the world’s largest AI data center and roughly multiply Chinese AI computing power by 2.5 times compared to using only domestic chips. So in the same year we’re told that China overtaking the US in AI is a generational threat, we also decided that as long as US companies get paid, selling the actual chips needed to achieve that is perfectly acceptable. And Nvidia didn’t stop there. They are now lobbying to sell even more advanced chips, arguing that the H200 isn’t enough to win the Chinese market. This is impressive and confusing—especially when you remember that Nvidia spent two years convincing investors that its CUDA software made its hardware nearly irreplaceable throughout the entire AI stack. One of these claims must be false because they directly contradict each other. Either Nvidia chips are unique and irreplaceable—in which case Chinese labs would pay a premium regardless of downgrades—or they’re not, meaning the valuations of all downstream AI companies are built on much weaker foundations than the market prices. This contradiction isn’t limited to chips. The same company can tell a Senate committee on Tuesday that its products are essentially harmless and need no liability framework, then tell a podcast host on Wednesday that this is the most powerful creation in human history, capable of actually extinguishing species. When asked, “Do you have a bunker?” the reply is, “I have an underground reinforced concrete basement, but nothing to… wait, I can’t let that one slide.” The harmless version is for regulators, to avoid liability frameworks. The extinction version is for investor roadshows and government grant committees—higher stakes mean bigger checks. Both can’t be true, but depending on which room the lobbyist is in, both are extremely useful. ## The Phantom Data Centers Last week, actor Kevin O’Leary—who occasionally plays a businessman on TV—announced he would build the world’s largest data center in Utah. The project is expected to consume more electricity than the rest of the state combined and cost tens of billions of dollars. Despite wearing two expensive watches at once, Kevin O’Leary does not have tens of billions of dollars. So most of his work focuses on securing favorable government approvals, fast-tracked regulations, and state-level tax breaks, vaguely hoping these conditions will attract the consortium of investors who will eventually fund the project. This announcement came about 16 months after O’Leary declared he would build the world’s largest data center in Alberta, Canada, with similar talk of reduced red tape and access to nearly unlimited cheap natural gas. Unsurprisingly, neither project has made any real progress. We’ve already done a video about the massive cancellation of data center construction sites over the past year. Among all the data centers that will never be built, these are the least likely. But there is one possibility—I can’t believe I’m saying this—that O’Leary isn’t as foolish as he looks. These proposals are worth noting only because he uses a narrative that makes multiple governments roll out the regulatory red carpet for a project they know will never see the light of day. “This is different,” he says. “We’re building the pipe and power from scratch. Pipeline brings natural gas, we burn it clean in turbines, and we feed power back into the grid. It’s good for the community. For the country, we need to compete with China.” That narrative clearly works. And to be fair, we probably don’t want China to take a leading position in this technology. But the way we are addressing this threat has three giant ironies. ## The Dark Money Network The test of a threat narrative is whether those pushing it actually act as if the threat were real. They don’t. The same companies warning that a Chinese AI takeover is a generational crisis are lobbying to sell China the most powerful chips, telling regulators their products are too harmless to require liability frameworks, and actively opposing any regulation that would slow their expansion. These tools exist in a “tech bro super-contradiction state”: claiming to be the greatest existential threat to American dominance since the Cold War, while also claiming to be too harmless for domestic regulation. It’s a stupid message, but the industry is pouring huge money to amplify it. Beyond the nine-figure direct political lobbying and acquisitions of major podcasts to promote the arguments, some companies have been exposed running what’s called “dark money” operations that perfectly evade formal disclosure. For example, OpenAI founders and early investors have funneled millions into an organization called “Building America’s AI,” which then passes funds to paid influencers to push almost the same narrative as Mr. Wonderful, but with less transparency. The message: (A) AI must stay in America; (B) we can’t let China win. The ultimate goal is blatantly to pave the way for a government bailout of OpenAI when needed. Overall, they understand AI is politically unpopular. Government bailouts are also unpopular, but they hope China’s AI is even more unpopular. ## What the Lobby Actually Buys The second irony is even larger. Even if you accept the AI arms race as a real and urgent threat, almost all the deregulation pushed under the “beat China” banner is not designed to actually win the race. Most recent deregulation efforts focus on two areas: environmental protection and energy prioritization, both aimed at accelerating private data center construction. The vast majority of computing power from these data centers will not be used for advancing defense applications, securing cyber infrastructure, or any monitoring and combat systems critical in a real strategic competition with another superpower. They will be used for targeted advertising, AI customer service chatbots, image generators, the tenth iteration of a chat product, and season five of *Fruit Love Island*. If this were truly a national security priority, the simple version of policy would be: narrow the scope of strategic AI projects, greenlight those projects, and treat the rest of the industry as the consumer product business it is. Instead, policy treats the entire industry as if it were the Manhattan Project. Not the best analogy, but consider steel: it clearly has civilian and military uses, so maintaining a stable supply of steel is a matter of national security. Yet we do not—The point is that the US response is designed not to win a race, but to keep money flowing to a handful of companies and financial institutions at the public’s expense. The third irony, not fully spelled out in the transcript but implied, is that the loudest voices warning we are losing this arms race may be the very ones ensuring we lose it. Because the “win” they are actually buying looks nothing like national security. It looks like permanent, government-backed monopolies for the biggest players, with regulatory burdens so high that new startups cannot comply, cannot afford the cost, and cannot enter the market. The result is a permanent monopoly for big companies, enabled by the government, all under the pretext of beating China. --- **Source:** [How Money Works on the US AI lobby gravy train, while China is overtaking the US in practical AI](https://www.youtube.com/watch?v=UgDEyQ1h-EA)