Nvidia CEO sets sights on making ‘several hundred billion’ dollars worth of electronics in the USA over the next four years, increasing the chance of your next GPU being made in America
In the wake of tariffs set by the Trump administration and new factories in the US from world-leading chip foundry TSMC, Nvidia’s CEO says it aims to produce “several hundred billion” dollars worth of electronics in the US in the next four years. That means there’s a growing chance your next GPU will be made in America.
According to a Financial Times interview with Nvidia CEO Jen-Hsun Huang, Nvidia will “procure, over the course of the next four years, probably half a trillion dollars worth of electronics in total”. Of that total sum, Huang says “We can easily see ourselves manufacturing several hundred billion of it here in the US.”
It was revealed this week that Nvidia is now making chips in the USA. We don’t yet have specifics on the type or quantity of chips but the announcement is a further declaration of Nvidia’s shift from Taiwan-based manufacturing to closer to home (in theory, as the chips still have a long way to go around the world until they’re in a functioning graphics card).
In a move that might help Nvidia get on the Trump administration’s good side, Nvidia is also seemingly pretty happy with the administration’s recent moves to deregulate AI. Just this week, the official White House website shared a quote from Huang saying “Having the support of an administration who cares about the success of this industry and not allowing energy to be an obstacle is a phenomenal result for AI in the U.S.”
President Trump’s tariffs are an ongoing proponent of his plan to push for more manufacturing in the US. Effectively, a 25% tariff on Chinese goods will make them more expensive for the consumer to buy, and less worthwhile to import. The logic is that this will incentivize in-country manufacturing.
This has pushed many companies to adapt, in order to continue supplying to America. Some graphics card manufacturers have moved their production of actual cards out of China as a result.
Huang tells Financial Times “At this point, we know that we can manufacture in the US, we have a sufficiently diversified supply chain.”
Nvidia is not the only major computing manufacturer to show increased investment in US production after threats of tariffs. Just earlier this month, TSMC and Trump announced a $100 billion investment into the US, with the intent to make three new factories. It is not yet clear if this is different from the $65 billion TSMC has already announced, which it had earmarked for building fabs in the US, some of which came from President Biden’s CHIPS Act.
TSMC is the world’s biggest semiconductor manufacturer and designer and doesn’t just work with Nvidia. AMD and Intel, among many others, utilise it, with the former even said to be making some Ryzen chips in the US. Based out of Taiwan, its Taiwanese fabs are currently developing the most advanced node processes for chips. Despite continued investment in America, a spokesperson for Taiwan’s president’s office clarified, “The most advanced processes will remain in Taiwan.”
Nvidia’s Blackwell architecture, which can be found in RTX 50-series cards, uses a custom 4 nm node process, which is a spin on a slightly older process from TSMC. The same is true of Nvidia’s current AI chips. Whereas the likes of the Apple M4 chip use a newer 3 nm process. Effectively, the smaller the process, the higher the transistor density, and the more efficient a chip can be.
Nvidia has rarely been on the cutting edge of the latest process, however, which means TSMC keeping the most advanced processes in Taiwan may not substantially affect its plans if it were to focus on more US production. Still, if it had wanted 3 nm chips, it wouldn’t have been able to get those in the US. At least not from TSMC. Perhaps it could tap Intel for its cutting-edge 18A node once that’s up and running.
TSMC fabs in the US are due to update to the 3 nm process in 2027, which could be soon enough for the RTX 60-series, though it will not be on the cutting edge by that time.
Source link
