The United States is trying to tax its way through the AI arms race, and the numbers do not add up. By letting Nvidia resume advanced chip exports to China while skimming a 25 percent cut of those sales, President Donald Trump has embraced a policy that pleases almost no one in the security community and leaves economists scratching their heads. The move risks strengthening China’s AI capabilities, undercutting Washington’s own export-control strategy, and confusing allies who had treated U.S. chip rules as a clear line rather than a negotiable toll.
Instead of choosing between access and denial, the White House has tried to split the difference, turning Nvidia’s China business into a revenue stream for the U.S. Treasury. Experts argue that this hybrid approach is strategically incoherent: it sacrifices leverage over Beijing’s AI development in exchange for a slice of corporate profit, while doing little to protect the technological edge Washington says it wants to preserve.
Trump’s 25 percent toll on Nvidia’s China sales
At the center of the controversy is President Donald Trump’s decision to allow Nvidia to sell advanced AI chips into the Chinese market as long as Washington takes a 25 percent cut of the revenue. The arrangement effectively turns a national security question into a licensing deal, with the U.S. government acting as a silent partner in Nvidia’s China business. According to reporting on the policy, experts have already warned that a U.S. decision to take a 25 percent share of Nvidia chip sales “makes no sense,” because it treats the AI race as a tax opportunity rather than a strategic contest that hinges on who controls the most capable hardware.
Critics point out that the plan blurs the line between regulation and participation. Instead of using export controls to limit China’s access to cutting edge accelerators, the administration is monetizing that access, even as analysts caution that Nvidia’s most powerful products are exactly the tools Chinese firms need to close the gap with U.S. rivals. One detailed account of the proposal notes that Donald Trump’s decision to permit Nvidia exports while collecting a quarter of the proceeds has been framed as a way to offset security risks, yet specialists quoted in that same analysis argue that no amount of tax can compensate for the strategic value of the compute capacity Washington is handing over to Beijing, especially when the chips in question are central to training frontier models.
Why security experts say the math does not work
From a national security perspective, the 25 percent cut looks less like a clever compromise and more like a mispriced trade. The United States has spent years arguing that advanced accelerators are a chokepoint technology that can slow China’s progress in military and dual use AI systems. By instead allowing Nvidia sales to proceed in exchange for a share of the revenue, the administration is effectively admitting that it is willing to sell that chokepoint for cash. Analysts quoted in one in depth critique of the policy describe this logic as fundamentally flawed, warning that the marginal tax revenue cannot offset the long term cost of helping Chinese firms acquire the very hardware Washington previously tried to keep out of their hands.
The security community’s concern is not abstract. The same reporting notes that by allowing Nvidia to ship high end chips into China, the United States is enabling more powerful training runs for large language models, computer vision systems, and other applications that could be adapted for surveillance, cyber operations, or autonomous weapons. In that context, the 25 percent levy looks like a discount, not a deterrent. One analysis of the decision argues that the policy undercuts the logic of earlier export rules that treated access to Nvidia’s top tier accelerators as a strategic asset, and that the new approach risks sending a message that Washington’s red lines can be negotiated down to a royalty.
“Only helps China”: the economic critique
Economists and industry analysts have been just as blunt as security experts in their assessment of the 25 percent cut. One prominent critic described President Donald Trump’s plan to take a quarter of Nvidia’s AI chip sales to China as “nuts,” arguing that it “only helps China” by normalizing the flow of advanced hardware into its data centers while giving Beijing a clear price tag for U.S. political risk. That assessment reflects a broader view that the policy effectively subsidizes China’s AI ambitions by guaranteeing Nvidia a path back into a lucrative market, even as Washington claims to be worried about the strategic consequences of that very access.
The economic critique also focuses on incentives. By turning the U.S. government into a beneficiary of Nvidia’s China revenue, the policy risks creating a constituency in Washington for continued exports, regardless of how the security environment evolves. Analysts warn that once the Treasury is collecting a steady stream of cash from Nvidia’s China business, it will be harder to tighten controls later without facing internal resistance. The same commentary that labeled the plan “nuts” notes that Chinese buyers are likely to treat the 25 percent surcharge as a cost of doing business, especially if they believe that access to Nvidia’s most advanced accelerators will help them leapfrog domestic competitors and secure a larger share of the global AI market.
Nvidia’s H200 and the AI arms race
The stakes are higher because the policy is not about legacy hardware. President Donald Trump has given Nvidia and other U.S. chipmakers a green light to sell their H200 AI accelerators into China, a class of chips that sits at the heart of the current wave of large scale model training. According to one detailed analysis of the decision, the authorization for Nvidia and its peers to resume exports of these advanced parts has significant implications for U.S. China technology competition, because it restores Chinese access to the same kind of high bandwidth memory and compute density that leading American labs rely on for their own frontier systems.
Another report on the move underscores how central Nvidia’s H200 has become to the global AI race. It notes that the Trump Administration has cleared Nvidia’s H200 chips for sale to China, reshaping the AI arms race by reopening a channel that earlier export rules had tried to close. That same account points out that the decision affects not only Chinese hyperscalers but also companies like Huawei, which can use imported accelerators to complement domestic designs and work around some of the constraints imposed by U.S. semiconductor export controls. In that context, the 25 percent cut looks less like a marginal adjustment and more like a structural shift in how Washington manages the flow of its most sensitive computing technology.
China’s AI ecosystem is no longer standing still
Supporters of the 25 percent policy sometimes argue that if the United States does not sell advanced chips to China, others will, or that Chinese firms will simply build their own. That argument misses how far China’s AI ecosystem has already come. As one close observer of the sector notes, the industry has already hit its stride, with large scale training and the migration of large models onto domestic platforms well underway. The same analysis explains that multiple Nvidia products have now been cleared for sale, and that for Nvidia, returning to the Chinese market sets the stage for a new round of competition in which U.S. and Chinese players race to capture the most demanding AI workloads.
In other words, Washington is not deciding whether China will have AI, but what kind of AI hardware stack it will build on. By allowing Nvidia back into the market, the United States is giving Chinese firms the option to lean on world class accelerators instead of being forced to optimize around domestic chips that may lag in performance or efficiency. That choice matters because it shapes how quickly Chinese labs can iterate on large models, how much capital they must devote to catching up on hardware, and how much leverage Beijing feels from U.S. export controls. The reporting on China’s reaction to the H200 decision makes clear that local players see Nvidia’s return as both an opportunity and a challenge, since it raises the bar for domestic designs even as it gives them access to the best foreign parts.
Beijing’s self sufficiency push and the risk of backfire
Even as Washington reopens the door for Nvidia, Beijing is not abandoning its drive for chip self sufficiency. One detailed newsletter on the subject highlights that self sufficiency push, noting that investors have shown great enthusiasm for upcoming listings by Chinese semiconductor and AI firms that promise to reduce reliance on foreign suppliers for the coming few years. That enthusiasm reflects a broader policy agenda in Beijing that treats advanced chips as a strategic vulnerability and is channeling capital into domestic fabs, design houses, and cloud providers that can absorb and deploy homegrown accelerators at scale.
The risk for Washington is that the 25 percent policy ends up accelerating this transition rather than slowing it. By signaling that access to Nvidia’s best products can be turned on or off by political fiat, the Trump administration is giving Chinese leaders another reason to double down on indigenous alternatives, even as it temporarily restores access to U.S. hardware. The same reporting that describes China’s reasons to say “no thanks” to Trump’s AI chip offer notes that Beijing’s planners are already thinking in multi year horizons, where short term access to foreign accelerators is less important than building a resilient domestic ecosystem that can withstand future sanctions or export bans.
How the policy reshapes the global AI race
The Trump administration has framed its Nvidia decision as a way to manage risk while keeping U.S. firms competitive, but the broader impact is to reshape the global AI race in unpredictable ways. One in depth business analysis explains that The Trump (President Donald Trump) administration is allowing Nvidia to sell its advanced computer chips in China, a move that could help Beijing narrow the gap or even win in the AI race if it uses that hardware to scale up training and deployment faster than U.S. rivals. That same reporting notes that China’s access to Nvidia’s accelerators will influence not only commercial applications but also the balance of power in areas like autonomous systems and cyber defense.
For U.S. allies and partners, the policy sends a mixed signal. On one hand, Washington is telling Europe and Asia that advanced chips are sensitive enough to justify strict export controls and joint screening of outbound investment. On the other, it is carving out a special arrangement for China that turns those same chips into a taxable commodity. The result is a more fragmented landscape in which some countries face hard limits on their access to Nvidia’s best products while China can buy them at a premium. That asymmetry could complicate efforts to build a coordinated front on AI governance and security, especially if other governments start to question whether U.S. rules are driven by strategy or by the prospect of revenue.
What this means for Nvidia and U.S. industry
For Nvidia, the 25 percent cut is both a constraint and a lifeline. The company has spent the past two years navigating a shifting maze of export rules that have repeatedly forced it to redesign products for the Chinese market. The new arrangement gives Nvidia clearer, if more expensive, access to Chinese buyers who still see its accelerators as the gold standard for training and inference. One policy focused analysis notes that President Donald Trump’s decision to allow Nvidia and other U.S. chipmakers to sell their H200 AI accelerators into China has significant implications for U.S. China technology competition, but it also stabilizes a key revenue stream for Nvidia at a time when demand from hyperscalers in the United States and Europe is starting to normalize.
Other U.S. firms are watching closely. Cloud providers, systems integrators, and AI startups that depend on Nvidia’s roadmap are trying to gauge whether the 25 percent policy will lead to more predictable export rules or simply add another layer of political risk. Some executives worry that if Washington is willing to tax Nvidia’s China sales today, it might impose similar levies on other forms of cross border AI collaboration tomorrow, from model licensing to cloud services. The broader tech sector is also grappling with a more mundane but still important question: how to price products and services in a world where key components can be subject to sudden, politically driven surcharges, much like the way consumer electronics makers must factor in tariffs when listing a new product in a major online marketplace.
China’s likely response and the limits of leverage
Beijing’s response to the 25 percent policy is still taking shape, but early signals suggest that Chinese firms will be selective in how they engage. One detailed account of China’s reaction to the H200 decision notes that But the industry has already adapted to earlier export controls by investing heavily in domestic platforms, and that Chinese companies now face a choice between doubling down on those efforts or re embracing Nvidia’s accelerators as their primary compute backbone. The same analysis points out that for Nvidia, returning to the Chinese market sets up a new round of competition in which local chipmakers will be judged directly against the U.S. leader, rather than being shielded by sanctions.
At the policy level, Beijing is likely to treat the 25 percent surcharge as both a warning and an opportunity. On one hand, it underscores the risk of relying on foreign technology that can be taxed or restricted at Washington’s discretion. On the other, it gives Chinese firms a window in which they can access world class hardware while continuing to build out domestic alternatives. A separate report that examines why China has reasons to say “no thanks” to Trump’s AI chip offer emphasizes that highlighting the self sufficiency push has become a political priority, and that investors are betting on Chinese firms that promise to reduce reliance on foreign chips for the coming few years. In that environment, the United States may find that its leverage is more limited than it assumes, because Chinese leaders are already planning for a future in which Nvidia is a nice to have, not a necessity.
Why experts see a strategic own goal
Put together, the security, economic, and industrial critiques converge on a simple conclusion: the 25 percent cut trades long term strategic advantage for short term revenue and corporate relief. Analysts who have examined the policy in detail argue that it undermines the logic of earlier export controls, which treated access to Nvidia’s top tier accelerators as a lever Washington could pull to shape China’s AI trajectory. By turning that lever into a toll booth, the Trump administration risks normalizing the very flows of technology it once tried to restrict, while sending a confusing signal to allies and adversaries alike about what the United States is actually trying to achieve.
The irony is that the policy arrives just as Washington’s broader export control strategy was starting to bite. Chinese firms had begun to re architect their AI stacks around domestic chips, and Beijing was pouring capital into self sufficiency precisely because it feared being cut off from Nvidia’s best products. By reopening the door in exchange for a 25 percent fee, the United States may end up giving China a last infusion of world class hardware that helps it bridge the gap until local alternatives mature. As one sharply worded critique put it, the plan is “nuts” because it only helps China, and the more closely I look at the reporting, the harder it is to disagree.
The road ahead for AI export policy
Where the policy goes from here will depend on how quickly its consequences become visible. If Chinese firms use their renewed access to Nvidia’s H200 and related accelerators to roll out more capable models and applications, pressure will grow in Washington to revisit the decision, regardless of how much revenue the 25 percent cut is generating. One detailed technology policy report has already warned that by instead allowing Nvidia sales to proceed under a revenue sharing scheme, the United States is accepting a higher level of strategic risk than earlier export rules contemplated, and that the logic behind the shift remains unconvincing to many of the experts who helped design those rules in the first place.
For now, the policy stands as a case study in the difficulty of balancing economic interests with security concerns in the AI era. The Trump Administration has chosen a path that tries to have it both ways, keeping Nvidia competitive in China while asserting that a 25 percent tax can offset the dangers of exporting cutting edge compute. The reporting from across the security, business, and China watching communities suggests that this bet is unlikely to pay off. Instead, it may leave the United States with the worst of both worlds: a more capable Chinese AI ecosystem and a less coherent export control regime, all in exchange for a slice of revenue that looks smaller and smaller the further out one projects the strategic costs.
More from MorningOverview