Jensen Huang says Nvidia's AI chips are now being manufactured in Arizona (www.cnbc.com)

Nvidia announced at GTC that its Blackwell AI GPUs — previously manufactured solely in Taiwan — are now in full production and assembly in Arizona, with the first Blackwell wafers produced in a Phoenix facility by TSMC. Jensen Huang framed the move as driven by national-security and job concerns after a presidential request, and emphasized supply‑chain resilience as U.S. export controls have already dented revenue (Nvidia said 6 million Blackwell GPUs shipped in the past four quarters and projects roughly $500 billion in GPU sales across Blackwell and next‑gen Rubin). The onshoring helps Nvidia make a policy case that U.S. access to advanced AI silicon is critical, even as China sales remain constrained by licensing rules (the H20 chip required U.S. export licenses and Nvidia estimated roughly $10.5B in lost H20 sales over two quarters). Beyond manufacturing, Nvidia unveiled strategic partnerships and products that extend its role in the broader telecom, quantum and supercomputing stack. It will take a $1B stake in Nokia and supply an ARC platform (Grace CPU + Blackwell GPU + networking) for 5G/6G base stations, and launched NVQLink to connect quantum devices to GPUs for error correction and orchestration. Nvidia also announced collaboration with the DOE on seven supercomputers. Together these moves reinforce Nvidia’s bid to cement U.S. AI infrastructure, reduce dependence on foreign hardware, and shape policy around export controls and technology leadership.