Nvidia’s latest Blackwell AI chips hit a snag when a design flaw emerged during production, causing yields to plummet. CEO Jensen Huang publicly admitted the issue was “100 % Nvidia’s fault,” explaining that the flaw required redesigning seven different chip components and ramping them into production simultaneously. The delay, which threatened deliveries to major customers like Meta, Google and Microsoft, was resolved with the help of longtime partner TSMC, allowing the refined Blackwell processors to re‑enter mass production in late October .¹
The setback coincided with a broader challenge: US export controls aimed at curbing China’s access to high‑end AI silicon. Huang criticized the Biden‑era restrictions as “fundamentally flawed,” noting they forced Nvidia to cancel its China‑tailored H20 chip and resulted in a $5.5 billion write‑down. The policy backfired, shrinking Nvidia’s China market share from 95 % to roughly 50 % and spurring rapid growth of domestic Chinese chipmakers such as Huawei .² ³
Despite the turbulence, Huang remains confident that the underlying demand for AI accelerators remains strong. Nvidia’s data‑center revenue surged to $21.9 billion in Q2 2026, and the company is investing heavily in new manufacturing options, including a $5 billion partnership with Intel to develop custom chips. Huang also hinted at potential direct sales of GPUs to OpenAI, signaling a shift toward tighter control of the AI supply chain .⁴
Looking ahead, the industry appears poised for a realignment. While export curbs have accelerated China’s homegrown chip ecosystem, the Trump administration is reportedly considering a rollback of the three‑tier licensing system, which could reopen avenues for Nvidia and other US firms to export AI technology—provided they avoid adversarial markets. The evolving landscape underscores the delicate balance between geopolitical strategy and the relentless drive for AI innovation .
