Huawei unveils LogicFolding chips to challenge Nvidia in China

Huawei announced plans to roll out new smartphone chips this fall using an internal engineering method called LogicFolding, a development that arrives as American export controls continue to shape competition in advanced semiconductors. The Chinese company, long a target of U.S. restrictions, presented the approach at an industry event in Shanghai and projected that the technology could reach performance levels comparable to 1.4-nanometer processes by 2031.

The timing matters because Nvidia has already signaled it is losing ground in China. Its chief executive recently described the company as having conceded the market for high-end chips there, a direct result of licensing rules that limit sales of its most sophisticated products. Analysts tracking the sector note that this outcome is accelerating Beijing’s push for domestic alternatives, with Huawei’s earlier Mate 60 device already helping it regain smartphone share from Apple inside China.

These moves sit inside a broader pattern of concentrated gains. Over the past two years, nearly all of the S&P 500’s advance has come from a handful of companies tied to AI infrastructure, including Nvidia, Microsoft, and Broadcom. Removing those names from the index would have cut the benchmark’s return roughly in half. The same concentration shows up in valuations: several of the leading chip-related stocks trade at multiples well above historical norms even as overall consumer sentiment remains weak and Treasury yields sit near 5 percent. Portfolio managers who treat broad index funds as diversified exposure are effectively carrying heavy bets on continued AI-driven demand.

That demand is itself evolving. Bank of America analysts recently raised their Nvidia price target to $350, citing the shift from chat-based tools to “agentic” applications that require far more compute per task. The firm points to $119 billion in supply commitments already on Nvidia’s books and expects sustained revenue and margin growth. At the same time, smaller but still sizable players such as AMD and Broadcom have reported triple-digit increases in AI-related revenue, underscoring how the entire semiconductor supply chain is being reshaped.

Security questions add another layer. Anthropic has developed an internal model, Mythos, capable of identifying software vulnerabilities at scale. The company has so far limited access to vetted partners precisely because the same capability could be misused by malicious actors. Early users report that the tool surfaces large numbers of issues, sometimes overwhelming existing patching processes. Governments in Asia have already launched reviews of their own systems in response. The episode illustrates how rapid capability gains in one domain quickly create new policy and operational challenges in another.

Taken together, the developments point to a feedback loop in which export controls intended to slow one nation’s progress instead speed domestic substitution, while commercial success in AI hardware concentrates both market returns and systemic risk inside a narrow set of firms. Industrial policy on both sides of the Pacific is now being written in real time through licensing decisions, investment flows, and corporate road maps rather than through comprehensive legislation. How those choices balance security concerns, economic concentration, and long-term technological capacity will determine whether the current cycle produces durable advantages or simply raises the stakes for future disruptions.