Huang's dire warning on the US-China tech war: what it means for Europe

The numbers behind the warning

Huang's concerns are not purely rhetorical. Current estimates suggest the United States still holds a commanding lead in deployed compute, with roughly 25 times more operational AI infrastructure than China. But the trajectory matters as much as the snapshot. China has integrated artificial intelligence into its national industrial strategy, targeting 90 per cent AI adoption in manufacturing by 2030 through state-directed planning. No equivalent mandate exists in the EU or UK.

"China is going to win the AI race. As I have long said, China is nanoseconds behind America in AI. It is vital that America wins by racing ahead and winning developers worldwide," Huang said, in remarks that circulated widely across the global technology policy community.

For European observers, the phrase "winning developers worldwide" carries particular weight. The EU is home to some of the world's strongest AI research institutions, from ETH Zurich to the Turing Institute, yet European AI companies consistently report losing senior engineering talent to US and, increasingly, Chinese competitors offering larger resource pools and fewer bureaucratic constraints.

Structural advantages China holds, and what Europe should note

Huang's most pointed structural argument concerns energy. China provides substantial subsidies to domestic technology companies for power consumption, a critical advantage given that large-scale AI training and inference are among the most energy-intensive workloads in modern computing. Huang has described this as power being effectively "free" for Chinese firms, allowing them to reinvest operational savings directly into research and development.

European energy costs sit at the opposite extreme. Industrial electricity prices across the EU remain elevated following the 2022 energy crisis, and while they have moderated, they are structurally higher than in the United States or subsidised Chinese markets. This creates a direct cost-competitiveness gap that European AI labs and cloud providers cannot easily close through efficiency alone.

Margrethe Vestager, the former European Commission Executive Vice-President for A Europe Fit for the Digital Age, has previously argued that Europe must compete on quality and trustworthiness rather than scale alone. That is a defensible position, but Huang's analysis suggests it may be insufficient if the scale gap widens further. Meanwhile, Demis Hassabis, CEO of Google DeepMind and arguably the most prominent European-born AI researcher in the world, has repeatedly called for Europe to invest far more aggressively in sovereign AI compute infrastructure, warning that dependency on non-European hardware and cloud services creates long-term strategic vulnerability.

Export controls: a policy backfiring in slow motion

One of the more uncomfortable threads in Huang's argument concerns US export controls on advanced semiconductors. Washington has progressively restricted NVIDIA's ability to sell its most capable chips, including the Blackwell architecture, to Chinese customers. The Trump administration confirmed it will not permit sales of Blackwell to China. The intention is to slow Chinese AI capability development. The evidence suggests it is doing the opposite.

Huang said in May that the chip export crackdown has been a "failure" in important respects, accelerating Chinese firms' drive to develop indigenous AI silicon and software stacks rather than constraining their ambitions. Chinese firms including Huawei, with its Ascend processor line, and a cluster of well-funded domestic chip startups have scaled investment precisely because access to NVIDIA hardware became uncertain.

The implications for Europe are direct. ASML, the Dutch lithography equipment maker and the single most critical node in global advanced chip manufacturing, sits in the middle of this geopolitical contest. The Dutch government, under pressure from Washington, has progressively restricted ASML's export licences for its most advanced extreme ultraviolet machines to China. Peter Wennink, ASML's former CEO, was candid before his retirement that these restrictions would not stop China's chip programme, only slow it, and risked damaging ASML's long-term commercial position without delivering the strategic objective.

The developer dilemma and Europe's talent question

Huang's argument about developer ecosystems has a specifically European dimension that tends to be underplayed. Approximately half the world's AI developers are estimated to be based in China. Policies that fragment the global developer community do not simply hurt US firms; they hurt any country whose AI industry depends on open-source collaboration, shared benchmarks, and internationally mobile talent.

The EU AI Act introduces tiered obligations on AI developers, including conformity assessments, transparency requirements, and restrictions on certain high-risk applications. These are legitimate policy objectives. But Yoshua Bengio, the Turing Award-winning researcher who has been closely involved in advising European AI governance bodies, has argued that regulation must be carefully calibrated to avoid creating compliance burdens that drive cutting-edge development outside European jurisdiction entirely. His concern is not that regulation is wrong but that poorly designed rules could hollow out the very ecosystem Europe is trying to protect.

The stakes extend well beyond market share. As AI becomes foundational to defence, critical infrastructure, pharmaceutical research, and public administration, the jurisdictions that lead in AI development will set global technical standards for a generation. Europe has a genuine interest in that standard-setting process, but influence requires presence in the frontier.

What European policymakers should take from Huang's warning

The honest read of Huang's FT Summit remarks is not that Europe should abandon regulation and race to the bottom on safety standards. It is that the current combination of high energy costs, fragmented capital markets, talent attrition, and compliance uncertainty is compounding into a structural disadvantage that good intentions alone will not reverse.

Several concrete pressure points deserve immediate attention from EU and UK policymakers. Industrial energy pricing for AI compute workloads needs to be addressed, either through direct support mechanisms or accelerated deployment of low-cost renewable capacity. The European Investment Bank's AI investment programmes need to be scaled materially, not incrementally. And the UK, now outside the EU single market, needs a bilateral arrangement with Brussels on AI research collaboration that does not leave both sides weaker by default.

Huang's warning was addressed to an American audience, but the logic travels. The nations that hesitate whilst others commit will find themselves negotiating terms rather than setting them.