Jensen Huang and Trump Move from Friction to Collaboration as Nvidia Trades Manufacturing Commitments for Policy Space
nashnova research
Nvidia CEO Jensen Huang's relationship with President Trump has shifted from the ~$4.5 billion hit triggered by the H20 export ban toward pragmatic alignment, built on a core exchange: expand U.S. manufacturing to earn negotiating leverage on China export licenses — redefining how the AI chip leader survives the U.S.–China tech standoff.
How did the two go from standoff to shared stage?
After Trump returned to the White House in early 2025, Nvidia faced a rapid sequence of blows: H20 export ban, tariff threats, domestic manufacturing demands, and tightened China market restrictions.
The turning point came in April 2025, when Huang announced expanded U.S. manufacturing — meeting political demands with capital expenditure softened the White House stance.
At the recent All-In Summit, Trump called Huang live on stage and endorsed AI infrastructure as a pillar for the next two to three decades. This means → the two have moved from adversarial posturing into mutual endorsement.
How much did the H20 ban actually cost Nvidia?
The H20 chip — a GPU deliberately designed to comply with the previous round of export controls — was banned anyway, leaving Nvidia with ~$4.5 billion in inventory and procurement losses.
Greater China (including Hong Kong) contributed ~$17.1 billion in revenue in Nvidia's fiscal year 2025, making it a core income source the company cannot ignore.
In plain terms = a chip purpose-built to be "weak enough to pass" still got blocked, and the loss landed straight on the balance sheet.
Why does Huang oppose tighter export controls?
He warned publicly that blocking advanced U.S. chips from China would accelerate Beijing's push for domestic alternatives — directly benefiting Huawei Ascend and Cambricon.
This reflects a core contradiction: controls are meant to constrain a rival, but the outcome may be clearing the field for that rival's own suppliers.
This means → Huang's stance is not just corporate self-interest — he is feeding Washington a policy narrative of "controls that backfire."
What did "expanding U.S. manufacturing" actually buy?
Huang announced expanded U.S. production capacity for Blackwell-architecture chips and AI supercomputers, partnering with TSMC, Foxconn, Wistron, Amkor, and ASE.
This means → the manufacturing expansion is not a pure business decision — it is a political leverage play, trading domestic capex for negotiating room on China export licenses.
In plain terms = spend more money, build more factories in the U.S. first — then go back to the White House and ask, "Can I keep selling chips to China?"
Why does Nvidia's supply chain double as a bargaining chip?
Nvidia's production spans multiple regions: the U.S. handles design and software; Taiwan handles advanced process and packaging; South Korea supplies high-bandwidth memory (HBM); Japan provides critical equipment and precursor materials.
This reflects a rare position: Nvidia is one of the few actors that simultaneously connects Silicon Valley, Washington, and Asia's semiconductor industry.
In plain terms = precisely because no one can replace this global network, Nvidia can hold its policy space even under intense regulatory pressure.
Could the AI-safety gap become the next fracture?
In their call, Trump openly dismissed "AI doomsday" talk and pushed infrastructure buildout as a core pillar for the next two to three decades.
Huang endorsed AI's economic upside but kept clear distance on safety — stressing that companies must rely on engineering diagnostics, stress tests, monitoring, and third-party audits, halting development when necessary.
This means → the two agree on "build fast, build big," but fundamentally diverge on the braking mechanism — a gap that could become the next friction point.
市场有风险，内容仅供研究参考，不构成投资建议。