There is a straightforward logic to export restrictions on advanced technology.

If you stop a competitor from buying your best tools, they fall behind. They cannot build what they cannot compute. You maintain the lead. Simple.

The US government applied that logic to Nvidia's most advanced chips starting in 2022, and expanded those restrictions multiple times through 2025. The intent was to slow China's development of advanced computing hardware and the models that run on it.

A Sky News investigation published this week, alongside a growing body of independent research, now raises a direct challenge to that assumption. The evidence does not yet say the policy failed completely. But it does say the outcome is considerably more complicated than the original logic suggested.

Don't wait. Sign up for The Hustle to unlock our side hustle database. Unlike generic "start a blog" advice, we've curated 100 actual business ideas with real earning potential, startup costs, and time requirements. Join 1.5M professionals getting smarter about business daily and launch your next money-making venture.

Get the guide

DeepSeek's R1 model, released in January 2025, benchmarked competitively against leading US models at a fraction of the reported training cost.

The significance was not just the performance. It was the method. DeepSeek's team published detailed research explaining how they achieved those results through architectural improvements, more efficient training procedures, and aggressive optimization of compute use rather than brute-force scaling on high-end hardware.

In other words, working without unrestricted access to the best Nvidia chips pushed Chinese researchers to find ways to do more with less. The constraint became an engineering problem, and the engineering problem got solved.

This is not a new dynamic in technology history. When Japanese manufacturers were restricted from certain US markets in the 1980s, they responded with process innovations that outpaced their American counterparts in several categories. Constraints imposed from outside can, under the right conditions, produce focused efficiency gains that a well-resourced competitor with no such pressure would never have been forced to develop.

DeepSeek gets the headline attention. CXMT is the quieter but equally important story.

Memory chips are as critical to computing infrastructure as processors. DRAM, the type of chip CXMT produces, is used in everything from smartphones to servers to the kind of high-density computing clusters that run large-scale models.

China had historically depended heavily on Samsung, SK Hynix and Micron for advanced memory chips. The export restriction framework, and the broader trade tension it sits within, gave Chinese policymakers a concrete reason to accelerate domestic alternatives. State investment in CXMT has been substantial and sustained.

The result, as of 2026, is a Chinese memory chip manufacturer shipping DDR5 at competitive prices and undercutting established players in segments they held comfortably two years ago. The timeline compression from this is not small. It represents years of progress that external pressure appears to have accelerated rather than delayed.

None of this is to say the US strategy was obviously wrong at the time it was made. In 2022, the gap between US and Chinese chip capability was real and wide. Restricting access to the best hardware was a rational attempt to preserve that gap.

The problem is that the strategy appears to have rested on an assumption that did not fully hold: that China's response would be to wait, rather than to build.

China did not wait. It built. The results are imperfect and the gaps are still real in several areas. But the direction of travel is no longer what the original policy anticipated.

This creates an uncomfortable position for US policymakers. The restrictions have cost Nvidia and AMD real revenue billions in deferred sales to Chinese customers who turned to domestic alternatives. Those domestic alternatives are now maturing in ways that will compete globally, not just within China.

For anyone watching the consumer and enterprise tech market, the geopolitical story has a direct commercial translation.

If CXMT's DRAM continues to mature, it introduces genuine price competition into a memory chip market that Samsung, SK Hynix, and Micron have dominated for years. That could eventually lower component costs for device makers and, over time, consumers.

If DeepSeek's efficiency approach proves to be a durable architectural advantage rather than a one-time result, the assumption that compute-scale dominance equals model-quality dominance becomes less reliable. That has implications for every company currently building its strategy around access to large-scale computing infrastructure.

Neither outcome is certain. Both are worth watching more carefully than most mainstream tech coverage currently does.

Policy outcomes in technology are rarely clean. The chip export restrictions were not nothing, and dismissing them entirely would be as careless as declaring them an unqualified success.

But the evidence that has emerged over the past eighteen months a competitive model built on constrained compute, a memory chip manufacturer ahead of schedule suggests that the gap the restrictions were designed to protect is narrowing in ways that were not fully anticipated.

The most useful question now is not whether the policy worked as intended. It is what the US and its allies do next, knowing what they know about how China responded.