In a lively keynote dripping with patriotic fervour, Huang reeled off a list of new partnerships with Uber, Palantir, Amazon and Microsoft while pledging his company’s loyalty to American manufacturing.
He spent most of his time boasting about the company’s latest Blackwell graphics processing units, which he called the most advanced artificial-intelligence chips ever made. Huang highlighted the Grace Blackwell NVL72, a “thinking machine” built mostly in US factories. It crams 72 GPUs into a single rack, weighs 1.3 tonnes and costs several million dollars.
The Nvidia boss said six million Blackwell chips had already shipped since their release last year, with orders for another 14 million over the next five quarters, totalling half a trillion dollars in sales.
Huang said: “We are manufacturing in America again. It is incredible. The first thing that President Trump asked me for is, bring American manufacturing back…. Nine months later, we are now manufacturing, in full production, Blackwell in Arizona.”
That claim stretches the truth a little. While the first silicon wafers for Blackwell GPUs are rolling off lines at Taiwan Semiconductor Manufacturing Company’s (TSMC) plant in Phoenix, much of the advanced packaging is still done in Taiwan. Huang told the audience that TSMC was working to bring its chip-on-wafer-on-substrate process to the US soon.
“Everything, from the beginning, the idea, silicon to the idea of intelligence will be manufactured here,” he promised.
Huang’s speech marked his latest attempt to charm Washington at a time when export policies could make or break Nvidia’s Chinese sales. The CEO will head to South Korea this week for the Asia-Pacific Economic Cooperation meetings, where Trump and China’s Xi Jinping are due to meet. Beijing’s access to Nvidia chips has been a hot bargaining point for both sides.
The CEO admitted that the US risked surrendering its AI edge if too much of its technology stack stayed dependent on foreign suppliers.
“I really hope President Trump will help us find a solution. Right now we’re in an awkward place.”
Nvidia’s growing presence in Washington reflects Huang’s shift from geek icon to political operator. The company’s near-total grip on the AI chip market has made him a celebrity and a target for critics who accuse him of undermining US national security.
In July, Huang called Trump America’s “biggest advantage” in the AI race and reportedly struck a deal giving the government 15 per cent of Nvidia’s China chip sales in exchange for export licences.
Earlier this year he popped into Mar-a-Lago to plead for favourable export rulings and to brag about Nvidia’s hundreds of billions in US investments.
During his speech, Huang rattled off a list of projects meant to make Nvidia sound indispensable to the national interest. Palantir will use Nvidia gear to create digital twins of corporate supply chains. Uber and Lucid Motors are teaming up to push autonomous driving. Drugmaker Eli Lilly is building a pharmaceutical supercomputer on Nvidia hardware.
In telecoms, the company has tied up with Nokia to weave AI into 6G networks, sinking $1 billion (about €930 million) into the Finnish outfit to boost its AI data-centre ambitions.
Energy efficiency got a mention. Huang praised US energy secretary Chris Wright for boosting power capacity, noting that without it “we could have been in a bad situation” given the energy demands of data centres.
Electricity shortages remain a headache for the sector, but Huang brushed that off with his trademark optimism.
“We’re going to do it hopefully every year,” he said of the new Washington summit. “Thank you all for your service and helping make America great again.”