Nvidia's Technology Case and Its Funding Case Are Two Different Bets

Jensen Huang used a Bloomberg interview on Friday to push back against the market's gloom over artificial intelligence, arguing the technology will deliver gains across industries and countries. It was a statement of long-term conviction rather than a rebuttal of the specific worry now dogging the sector: how the buildout of AI infrastructure gets paid for.

That gap matters. Broad adoption of AI and the financing of the hardware behind it are separate questions. Huang's argument addresses the first. The selling pressure that has hit chip stocks addresses the second — and it is the second that has been setting the tone.

A Split Tape

Chip shares came under fire Friday as doubts about the AI trade and its funding resurfaced, compounding rate worries already stirred by hotter-than-expected US inflation data. Nvidia closed at €190.52, caught in the same downdraft. Reuters traced Thursday's weakness in AI names to concern that higher interest rates could make AI investment harder to finance, and pointed to TSMC's cautious outlook for 2027 amid macroeconomic uncertainty.

Against that, UBS analyst Timothy Arcuri reaffirmed his buy rating and $300 price target on October 9, reading Taiwanese export data as consistent with healthy underlying AI demand. The two views describe different slices of the business: near-term financing conditions on one side, end-market demand on the other.

What Nvidia Actually Shipped This Week

On the product front, Nvidia said Wednesday that its CUDA-X libraries are being adopted across the semiconductor industry, including a new library called cuPotential. The applications cover AI-assisted chip design and manufacturing. In early access tests, cuPotential accelerated GPU-optimized workflows by two to three times, according to the company — a test result, not a productivity gain achieved at scale.

The automotive side got a lift as well. Nvidia and Mercedes-Benz will jointly develop the next generation of driver-assistance and automated-driving systems, combining Mercedes-Benz MB.OS with Nvidia technology. Volvo has announced plans to use the technology in a next-generation vehicle. Both are development commitments, not booked revenue; the real test is the handoff from engineering promises to production use.

China Is the Counterweight

Not every data point points the same way. Reuters, citing IDC, reported that Nvidia's share of China's AI data-center chip market stood at 34% in the first half of 2026, with shipments down 44% year over year. US export restrictions limited supply, the report said, while domestic rivals led by Huawei gained ground. That sits awkwardly beside UBS's constructive read on global demand — though again, they describe different geographies.

Competition is building elsewhere too. Microsoft unveiled its own AI accelerator, Maia 300, on Monday, with deployment due to begin in 2027 as the primary system for inference across its products and services. That does not translate into a specific loss for Nvidia, but it is a planned alternative that belongs in any long-term demand model.

The Next Data Point

Huang takes the stage at GTC Berlin on October 21, with AI factories, open models, physical AI and accelerated computing on the agenda. No product launches or new contracts have been promised. What investors will be listening for is evidence that this week's announcements are moving toward real deployment — because the case for a broader demand base rests on execution, not on the volume of press releases.