NVIDIA to Build AI Supercomputers in the US for the First Time

NVIDIA will produce AI supercomputers entirely in the United States for the first time. The company announced the move on April 14, 2025, framing it as a four-year buildout sized at up to $500 billion in AI infrastructure production. The plants will cover more than a million square feet of manufacturing space across Texas and Arizona, with Foxconn in Houston, Wistron in the Fort Worth area, and TSMC in Phoenix.

Until now, NVIDIA designed its chips in the US and outsourced production to contract manufacturers, largely in Taiwan. The new plants pull the full supercomputer assembly onto US soil for the first time, while the company has also tapped Amkor and SPIL for chip packaging and testing in Arizona.

NVIDIA’s First US-Built AI Supercomputers

NVIDIA’s April 14, 2025 announcement of US supercomputer manufacturing described the plan in the company’s own words. The release said the company has commissioned more than a million square feet of manufacturing space to build and test NVIDIA Blackwell chips in Arizona and AI supercomputers in Texas. Five manufacturing partners are part of the build, including TSMC for chip wafers, Foxconn and Wistron for supercomputer assembly, and Amkor and SPIL for chip packaging and testing. The plan pulls final assembly onto US soil for the first time in NVIDIA’s history.

The engines of the world’s AI infrastructure are being built in the United States for the first time.

Jensen Huang, founder and CEO of NVIDIA, said it in the company’s April 14, 2025 announcement. The same release said NVIDIA will use its Omniverse platform to build digital twins of the new factories and its Isaac GR00T model to develop the manufacturing robots. Mass production at the two Texas plants is expected to ramp up within 12 to 15 months of the announcement.

Blackwell wafer production is already underway at TSMC’s chip plants in Phoenix, per the announcement. Amkor and SPIL will handle chip packaging and testing at separate Arizona sites. Until 2025, NVIDIA designed its GPUs in the US and outsourced manufacturing to contract partners, most of them in Taiwan. The new plants move the final assembly step into Texas for the first time, but the Taiwan link remains.

A $500 Billion, Four-Year Buildout

NVIDIA has framed the US AI infrastructure buildout at up to $500 billion in production value over four years, through partnerships with TSMC, Foxconn, Wistron, Amkor, and SPIL. The figure is the production value of the AI infrastructure NVIDIA expects to flow through its US partners, not a lump-sum check the company has committed to write. The same announcement frames the buildout as supply chain hardening, with the Texas and Arizona plants sitting closer to US AI customers.

Wistron has confirmed $761 million for its two Fort Worth plants and more than 800 jobs at the Alliance development. Both Fort Worth sites are expected to be operational by early 2026, with local officials approving tax abatements to land the deal. The Fort Worth site joins the Houston Foxconn plant and the Phoenix and statewide Arizona sites in the four-year build. Manufacturing partners are deepening their own US footprints as part of the arrangement, growing their businesses while expanding their global reach.

What Forced the Map to Change

On April 2, 2025, the Trump administration announced reciprocal tariff rates that included a 32% tariff on products from Taiwan, where NVIDIA largely manufactures its GPUs. China was hit with 145% tariffs the same week. The Taiwan rate was among the highest announced that day.

On the Friday before NVIDIA’s Monday announcement, Trump exempted chips, smartphones, computers, and other tech devices and components from the new tariffs. On the Sunday before the announcement, Trump said he would announce tariffs on imported semiconductors within the week. The exemption was a reprieve, not a permanent shield, since semiconductor-specific tariffs were promised days later. The sequence put NVIDIA’s manufacturing footprint back in the crosshairs of US trade policy right as the company prepared its largest US expansion.

When asked whether the US manufacturing announcement was tied to the tariff fight, NVIDIA declined to comment. The silence leaves the link between the two events unconfirmed by either side. NVIDIA’s blog post does not mention tariffs in any form.

The White House put its own framing on the record the same day, calling the move the Trump Effect in action. Investors and analysts read the shift as supply-chain rerouting under tariff pressure. The map changed because the cost calculus of building in Taiwan had moved. The US plants put NVIDIA’s manufacturing footprint closer to its biggest customers and its biggest political ally. The shift did not sever the company’s ties to Taiwan, but it did redirect the next four years of capex into US soil.

Texas and Arizona Take the Work

Foxconn is leading the AI supercomputer assembly plant in Houston under the partnership, per NVIDIA’s announcement. The Houston site is one of two Texas facilities in the buildout. Mass production at the Houston plant is expected to ramp up within 12 to 15 months of the April 2025 announcement, the same window as the second Texas site.

Wistron is taking the lead in North Texas, where local officials confirmed two AI supercomputer manufacturing facilities at the Alliance development in Fort Worth. Per local reporting, Wistron is investing $761 million in the two plants and creating more than 800 jobs in the area, with both sites expected to be operational by early 2026. Governor Greg Abbott welcomed the Texas investment in a statement on the Texas investment.

The Arizona leg is split across partners, with TSMC’s Phoenix plants producing Blackwell wafers and Amkor and SPIL handling chip packaging and testing at separate Arizona sites. The packaging step, in particular, is the bottleneck that has constrained AI chip supply for two years.

• 1,000,000+ square feet of manufacturing space commissioned
• $761 million Wistron investment in Fort Worth
• 800+ Wistron jobs announced in North Texas
• 12 to 15 months to mass production at both Texas plants
• 5 NVIDIA partners with US production roles: TSMC, Foxconn, Wistron, Amkor, SPIL

Even Arizona-Made Chips Still Head Home

The US plants do not finish the picture. Industry trade outlets reported in October 2025 that the first Blackwell wafer produced at TSMC’s Fab 21 in Phoenix still has to travel back to Taiwan for advanced packaging. Wafer fabrication and final assembly are now in Arizona and Texas, but the most advanced packaging step, the chip-on-wafer-on-substrate (CoWoS) technique, remains in Taiwan. Until that step moves onshore, US-built Blackwell chips will still cross the Pacific twice before they reach a data center rack.

Adding American manufacturing helps us better meet the incredible and growing demand for AI chips and supercomputers, strengthens our supply chain and boosts our resiliency.

Jensen Huang said it in NVIDIA’s April 14, 2025 announcement, framing the move around supply chain hardening. The announcement does not address the packaging wrinkle. The shift is real, and so is the remaining dependence on Taiwan for the most advanced step.

The White House Claimed the Win

The same day as NVIDIA’s announcement, the White House put out a statement on the announcement that named the move the Trump Effect in action. The release credited President Trump’s priority on US-based chip manufacturing. The administration has owned the link between the tariff package and the NVIDIA buildout from its side.

The market has rewarded the strategy. On October 29, 2025, NVIDIA became the first company in history to close above a $5 trillion market valuation, three months after becoming the first to clear $4 trillion. The Trump Effect framing, the AI boom, and the US buildout announcement were cited in coverage at the time. As of June 11, 2026, NVIDIA’s investor relations page listed the share price at $204.87.

NVIDIA has not confirmed a link between the announcement and the tariff fight. The White House has named the link from its side. The market reward for the strategy is on the record, and the company’s own framing of the decision is not.

Frequently Asked Questions

When did NVIDIA announce it would build AI supercomputers in the US?

April 14, 2025. The same business day, the Trump administration had imposed a 32% reciprocal tariff on products from Taiwan, and on the Friday before the announcement the White House exempted chips from that tariff package.

Where will NVIDIA build the supercomputers?

Texas and Arizona. Foxconn is leading a Houston plant, Wistron is leading two North Texas plants in the Fort Worth area, TSMC is making Blackwell wafers in Phoenix, and Amkor and SPIL are handling packaging and testing in Arizona.

How much is NVIDIA investing in US manufacturing?

NVIDIA has put a four-year, up to $500 billion price tag on US AI infrastructure production through five partners: TSMC, Foxconn, Wistron, Amkor, and SPIL. The figure is the production value of AI infrastructure flowing through those US plants, not a single cash deposit. Wistron alone has confirmed $761 million for its two Fort Worth plants and more than 800 jobs.

When will mass production start at the US supercomputer plants?

Both Texas plants are scheduled to reach mass production roughly 12 to 15 months after the April 2025 announcement, with the Wistron Fort Worth sites targeted to be operational by early 2026.

Is the move a response to Trump’s tariffs?

NVIDIA declined to comment when asked. The White House has called the move the Trump Effect in action. Coverage in 2025 reported the link from the administration’s side, and NVIDIA has not confirmed it.