When Nvidia CEO Jensen Huang makes a big claim about where artificial intelligence (AI) is headed, he rarely walks it back, especially when the numbers keep lining up in his favor.
On September 10, Jensen Huang stood on stage at one of Wall Street’s biggest technology gatherings and repeated a forecast he first made exactly one year earlier.
The timing came just days after Nvidia posted another blowout quarter, and as fresh questions swirl around a close partner, CoreWeave, and the physical limits on how fast the AI buildout can grow.
Jensen Huang sticks with his $4 Trillion AI market call for 2030
Speaking at the Goldman Sachs Communacopia and Technology Conference on September 10, Huang reiterated his expectation that the global AI infrastructure spending could reach between $3 trillion and $4 trillion by 2030, according to Yahoo Finance.
It was exactly one year earlier, at the same conference, that he first floated the number.
“The semiconductor industry is going to just keep getting larger and larger,” Huang said, pointing to a new layer of computing and the end of Moore’s Law as the two forces feeding the buildout.
He added that demand for smarter AI models, particularly the ones that produce fewer wrong answers, should compound into a much bigger industry over time.
More Nvidia:
- Nvidia just made a move Wall Street wasn’t ready for
- Nvidia just locked down deal that changes AI race
- Nvidia stock is doing something it hasn’t done in years
Goldman Sachs analysts at the same event described visibility across the semiconductor supply chain as historically unprecedented. Equipment makers and foundries reported sharply higher order backlogs, while the bank’s analysts flagged downstream power and data center capacity as potentially becoming the bigger bottleneck going forward.
Huang also argued that Nvidia has evolved beyond a chip supplier into something closer to a financial platform and infrastructure around AI.
He pointed to system pricing that has climbed sharply across generations, with Hopper-based systems running about $18,000 per GPU, Blackwell near $25,000, and the upcoming Vera Rubin platform closer to $40,000,as reported by Benzinga.
The earnings numbers behind Huang’s confidence
Huang’s strong defense did not come out of nowhere. Nvidia’s latest quarter, reported in late August, delivered adjusted earnings of $2.22 per share on revenue of $96.2 billion. Beating Wall Street’s estimates of $2.09 per share and $92.3 billion.
The company’s Data Center segment, which covers hyperscalers, AI clouds and enterprise customers, brought in $89 billion versus a projected $85.8 billion, as reported by Yahoo Finance. The smaller unit covering gaming and physical AI generated $7.2 billion against expectations of $6.6 billion. Overall revenue grew 106 percent year over year and 18 percent from the prior quarter.
Nvidia also told investors it expects 70% revenue growth for fiscal year 2028, well above the 45 percent growth Wall Street had penciled in. Huang said the increase would have been even larger, potentially topping 100%, if not for supply constraints, including an ongoing shortage of memory chips that has squeezed the entire industry.
That memory crunch is already showing up in the numbers. Nvidia expects gross margin to slip from 75% last quarter to roughly 74% in the current one. Guidance for the October quarter sits near $108 billion, a figure that assumes zero data center chip sales to China.
Cheng Xin / Getty Images
CoreWeave shows just how tight supply really is
Few companies illustrate the demand side of Huang’s argument better than CoreWeave.
“We are struggling to meet demand every day,” CoreWeave CEO Michael Intrator said at the same Goldman Sachs conference. “Every GPU we have could be sold to multiple different clients. It is a unique moment, and it continues to be.”
Nvidia is not a passive bystander in that story. The company supplies effectively all of the advanced GPUs running CoreWeave’s specialized AI data centers and holds an 11.5 percent equity stake in the company, making it CoreWeave’s biggest supplier and one of its major strategic investors.
That closeness feeds a broader debate over how AI infrastructure spending gets valued. On CoreWeave’s second quarter call, Intrator said a batch of older H100 GPUs coming off an expired contract was immediately rebooked at 95% of its original price, TheStreet reported. A data point he has used to counter claims that Nvidia’s chips depreciate faster than balance sheets suggest.
Short seller Jim Chanos challenged that argument after Huang shared GPU rental data on social media in early September, questioning why companies renting Nvidia hardware do not simply raise prices instead.
Michael Burry, who disclosed a short on Nvidia earlier this year, has raised similar depreciation concerns, and CoreWeave’s coming earnings calls should add fresh data to that fight.
What Nvidia investors should watch next
The memory shortage sits at the top of the list. Nvidia has described “extreme pricing conditions” for high-bandwidth memory that have already exceeded its earlier expectations. And supply across the three major HBM suppliers remains tight. Any further tightening could pressure the margins beyond current guidance.
China remains the other overhang. Export restrictions have already dropped Nvidia’s China data center revenue to zero, and its own guidance assumes that stays the case.
Chinese rivals Huawei and Cambricon have raised prices on competing chips by as much as 50% because of their own memory access problems. A sign Nvidia’s absence has not simply handed the market to cheaper alternatives.
The Burry and Chanos critiques around depreciation deserve attention too, even though Nvidia has not directly addressed them beyond Huang’s social media posts.
With Wall Street still overwhelmingly bullish on the stock, investors would be smart to keep watching CoreWeave’s rental pricing and Nvidia’s own margins for early signs of whether that skepticism has merit.
Related: Jensen Huang just answered Michael Burry’s Nvidia bear case
