NVIDIA earlierAnnounceIn a highly symbolic milestone, CEO Jensen Huang personally unveiled the first Blackwell wafer manufactured "in the U.S." at a celebratory event. This wafer, the core foundation of NVIDIA's next-generation AI chip, was produced at TSMC's advanced semiconductor manufacturing facility (Fab 21) in Phoenix, Arizona.
This move not only marks the readiness of NVIDIA's latest Blackwell architecture to enter volume production in the United States, but also represents a crucial first step in strengthening domestic chip manufacturing capabilities and ensuring the resilience of critical supply chains.
"The most important chip in modern American history" is produced locally
At the event, Jensen Huang emphasized the historic significance of this collaboration, stating, "This is the first time in modern American history that this 'single most important chip' is being manufactured in the United States by the most advanced foundry (TSMC)."
These remarks also highlight the absolute importance of the Blackwell platform to the current AI revolution. Since its release in 2024, NVIDIA has emphasized its promise to revolutionize the AI industry. Nearly every tech giant, including Amazon (AWS), Google, Meta, Microsoft, and OpenAI, has pledged to adopt this next-generation architecture for their next-generation large-scale language model training and inference infrastructure.
According to data previously released by NVIDIA, the Blackwell platform can provide up to 25 times lower costs and energy savings for specific AI workloads compared to the previous generation Hopper architecture (such as the H100). Ensuring the supply of this "strategic" chip is naturally a top priority for NVIDIA.
TSMC's Arizona plant plays a key role
TSMC's Arizona facility, responsible for producing Blackwell wafers, is a key component of TSMC's global manufacturing risk mitigation efforts in response to the US CHIPS Act. Although the facility faced delays and cost challenges during construction, its successful production of NVIDIA's next-generation wafers provides significant support for TSMC's operational prospects and mass production capabilities.
TSMC's Arizona wafer fab (Fab 21) is currently primarily utilizing 4nm (N4/N4P) process technology. NVIDIA's Blackwell GPUs, such as the B100 and B200, utilize TSMC's customized 4NP process, which aligns with the fab's technology blueprint. The premiere of wafers produced there by a key customer like NVIDIA further demonstrates the close partnership between the two companies in advanced processes.
Core goal: Reduce geopolitical impact
NVIDIA also stated that the key strategic goal of this collaboration is to strengthen its "chip manufacturing capabilities in the United States." NVIDIA stated in a statement that with the start of Blackwell wafer production at TSMC's Arizona plant, NVIDIA will be able to "more effectively insulate itself from evolving tariff situations and geopolitical tensions."
Over the past few years, the tech war and trade frictions between the US and China have increasingly exposed the risks of a highly concentrated semiconductor supply chain in Asia (particularly Taiwan). For NVIDIA, its high-end AI chips are not just commercial products; they are also considered critical strategic materials by the US government, with strict export restrictions placed on certain countries.
Therefore, establishing a production line in the United States capable of producing cutting-edge AI chips provides NVIDIA with a critical buffer and redundancy solution, both in meeting the US government's "local manufacturing" requirements and in responding to potential future supply chain emergencies.
Continued expansion of U.S. AI infrastructure investment
With the Blackwell architecture officially entering mass production, NVIDIA's current goal is not only to meet the market's explosive demand for AI chips, but also to continue to expand its overall manufacturing footprint in the United States.
Earlier this year, NVIDIA announced plans to invest up to $5000 billion to build a more comprehensive AI infrastructure in the U.S. This ambitious plan will not be accomplished solely by NVIDIA, but will be achieved through close collaboration with TSMC, Foxconn, and other supply chain partners.
From upstream advanced wafer manufacturing (TSMC) to mid- and back-end chip-on-a-sheet (CoWoS) and server assembly (Foxconn), NVIDIA is clearly attempting to gradually replicate or back up key links in the AI industry chain within the United States, creating a more resilient and autonomous ecosystem. The Blackwell wafer launch in Arizona is a crucial step in this grand strategy.
