Nvidia wants to build and sell up to half a trillion US dollars of American-made AI supercomputer equipment over the next four years, with the help of Taiwan Semiconductor Manufacturing Co, aka TSMC, and its partners.
The pledge means Nvidia will begin production of its latest generation of Blackwell accelerators and systems in the United States within the next 12-15 months, or so it hopes. It also assumes there will be sufficient buyers, domestic or otherwise, of this half-a-trillion-bucks of made-in-America gear over the coming years.
The corporation’s promise comes as American companies wrestle with the Trump regime’s on-again, off-again tariffs on foreign imports.
“The engines of the world’s AI infrastructure are being built in the United States for the first time,” CEO Jensen Huang said in a canned statement Monday. “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.”
To support this goal, Nvidia has commissioned more than a million square feet of manufacturing space to assemble chips in Arizona and build servers in Houston and Dallas, Texas. According to the GPU giant, these sites will eventually support the deployment of tens of gigawatt-scale datacenters – or “AI factories” as Huang prefers to call them – in the States.
Production of Nvidia’s Blackwell chips – which relies on TSMC’s chip-on-wafer-on-substrate (CoWoS) packaging technology – is said to have begun earlier this year at TSMC’s Phoenix, Arizona, chip fabs.
Nvidia has previously used TSMC’s CoWoS tech to incorporate high bandwidth memory modules in its highest-end accelerators, though only outside of the US. Late last year, TSMC announced plans to bring both its CoWoS and integrated fan-out packaging tech to America as part of a collaboration with Arizona-based chip packaging and test provider Amkor, which will work with Taiwan’s Siliconware Precision Industries Co. (SPIL) to assemble and test its US-made Blackwell accelerators.
To produce its increasingly power hungry servers and rack systems in America, Nvidia is working with contract manufacturing outfits Foxconn and Wistron to stand up manufacturing facilities in Houston and Dallas, respectively. Wistron appears to have a budget of $50 million to purchase property or land in the US.
These plants are expected to use Nvidia’s Omniverse digital twin and Isaac GR00T robotics platforms to automate manufacturing.
Performing final assembly in the US benefits Nvidia and its American customers, because AI systems are sometimes unique and a fault or misconfiguration on an imported system would be hard to address. Final checks conducted in the US could save a lot of back-and-forth.
“We actually used to have standards for chassis. We had power standards. We had keep-out zones. [Nvidia’s server design spec] HGX was one of those standards, but those didn’t really comply with wattage or even rack height,” Moor Insights and Strategy Chief Analyst Patrick Moorhead told El Reg. “Now, pretty much all of these are bespoke.”
To Moorhead’s point, Nvidia’s system design has become considerably more complex and the tolerances tighter over the past few years as the GPU giant has transitioned from semi-standard air-cooled systems to rack-scale machines containing more than a hundred liquid-cooled CPUs and GPUs connected by miles of copper cabling.
$500 billion of what, exactly?
One wonders if Nvidia is really promising to build and sell $500 billion of just its own chips and DGX-branded servers in the States, or if that figure will include sales of entirely-US-made Nvidia-powered machines from the likes of Dell, HPE, Supermicro, and Lenovo, too. It’s not terribly clear from the varying language in the GPU giant’s announcement.
Moorhead argued if the $500 billion is just homegrown DGX server sales, that’s quite ambitious, and if it includes HPE et al, it’s intentionally conservative to avoid being interpreted as long-term guidance.
“Five-hundred billion dollars over four years of Nvidia servers, that would be very, very aggressive,” he said, adding that this would also signal a business model change because today the ecosystem is offered a mix of Nvidia-branded servers, equipment for hyperscalers, and traditional boxes from the likes of HPE and Dell.
Nvidia declined to comment.
The GPU titan’s commitment to US manufacturing comes a week after it was reported the Trump regime planned to pause controls that prevented the export of Nv’s H20 GPUs to China, following an appearance by Huang at a $1-million-a-head dinner at the US President’s Mar-a-Lago resort home.
Nvidia is not alone in moving to manufacture its products in the USA as the White House pushes corporations to “reshore” their factories. Last month, TSMC revealed plans to spend upwards of $100 billion to construct in the States three new fabrication plants, two advanced manufacturing facilities, and an R&D center.
And in February, Apple announced it plans to spend $500 billion in the US including a commitment to purchase a large quantity of chips from TSMC’s Arizona fab and build a server-making plant in Houston – the same city where Foxconn is setting up facilities with Nvidia. That sounds suspiciously like one factory built by Foxconn could end up serving both tech giants. ®
Bootnote
Apple and Nvidia aren’t the only ones eyeing TSMC’s US capacity. AMD late Monday revealed it had successfully validated its 5th-gen Epyc datacenter chips, codenamed Turin, for production at TSMC’s Fab 21 in Arizona. AMD also said its 6th-gen Venice Epycs would be built using TSMC’s 2nm process tech when they arrive in 2026.
- SEO Powered Content & PR Distribution. Get Amplified Today.
- PlatoData.Network Vertical Generative Ai. Empower Yourself. Access Here.
- PlatoAiStream. Web3 Intelligence. Knowledge Amplified. Access Here.
- PlatoESG. Carbon, CleanTech, Energy, Environment, Solar, Waste Management. Access Here.
- PlatoHealth. Biotech and Clinical Trials Intelligence. Access Here.
- Source: https://go.theregister.com/feed/www.theregister.com/2025/04/15/nvidia_made_in_america/
