Nvidia CEO Defends AI Buildout After $1 Trillion Tech Selloff
TREE NEWS reports: Global technology stocks suffered a broad-based rout this week, with the world’s twenty largest tech companies collectively erasing more than $1 trillion in market value. At the center of the storm was mounting skepticism over the return on investment in AI infrastructure — specifically, who will ultimately pay for the trillions of dollars being poured into data centers and compute capacity.
Nvidia CEO Jensen Huang moved swiftly to counter the narrative. In a Bloomberg interview from Stockholm, he publicly backed CoreWeave, stating the company “does not need Nvidia to provide backstop support” and that its “cost of capital is actually falling.” The remarks triggered an intraday rebound in new cloud computing stocks, with CoreWeave closing up 5.94% against the broader downtrend. The Nasdaq subsequently recovered all of its losses.
The CoreWeave Stress Point
CoreWeave became the most closely watched pressure point in this week’s selloff. Five-year credit default swap spreads insuring against CoreWeave debt default spiked as much as 74.5 basis points intraday, hitting an all-time intraday high of 1,194.5 basis points — implying a five-year default probability exceeding 60%.
The anxiety stems from CoreWeave’s debt-fueled expansion model. Nvidia is both its equipment supplier and its second-largest shareholder with roughly a 10% stake. This circular financing structure — where a supplier finances a buyer who then purchases the supplier’s products — has raised widespread questions about the financial sustainability of AI infrastructure expansion.
Huang’s comments directly countered credit market pricing. “CoreWeave is getting stronger every day,” he told Bloomberg. “I’m very optimistic. Their cost of capital is actually coming down.” He also explicitly stated that Nvidia does not need to provide any form of backstop for CoreWeave.
Deconstructing the Misreading
In a nearly 140-minute interview on the BG2 podcast released just before the U.S. market open, Huang addressed the market’s core anxiety: if OpenAI and Anthropic revenue growth slows, who pays for trillions in compute buildout?
Huang called this framing “asking the wrong question.” He broke down Nvidia’s compute demand into three roughly equal segments:
- About one-third: Basic computing entirely unrelated to AI labs — molecular dynamics, physics simulation, quantum chemistry, classical machine learning (such as Visa and Mastercard fraud detection), and recommendation systems. “Even with no AI at all, Nvidia isn’t zero,” he said. “This business grows 30% to 40% annually.”
- About one-quarter to one-third: Frontier lab compute demand. But Huang emphasized the market typically sees only two players, while he sees five or six — Meta, Grok, Gemini, Anthropic, and others emerging. “If you tell me that represents less than a third, about a quarter of our business, that’s fine.”
- About one-third: Thousands of startups renting compute — which he likened to “the next-generation AWS.” About 80% of these companies use open-source models, building specialized intelligence in vertical domains like physical AI, biological AI, and materials science.
“Betting trillions of dollars of buildout on two labs’ revenue is a big mistake,” Huang said. “I’m happy to work with them and support them. They will become among the most important tech companies ever and achieve brilliant success. But their revenue doesn’t need to approach $1 trillion for this to work.”
Why Demand Accelerates
Huang articulated why compute demand won’t decelerate linearly but will continue to accelerate. The core mechanism involves two compounding factors: first, inference (thinking) is orders of magnitude more expensive than recitation-style generation — AI must explore multiple options, plan, reflect, prune, and retain “strategic options,” consuming substantial compute. Second, the more capable the model, the more scenarios it gets “hired” for.
“It uses more compute to do the work, but it does it well, so we hire it to do more work,” he said. “These two things stack, and demand accelerates, not decelerates.”
He identified last year’s key unlock as the “harness” — working memory management, knowledge graph construction (similar to offline compression during sleep), and tool use — which transformed large language models from chatbots into agents capable of enterprise work.
Power Scarcity and Nvidia’s Share
Looking toward 2027, Huang gave a clear assessment: demand will be “no less, only more,” but land, permits, power, and construction labor will be the primary obstacles. Next year will add 43 gigawatts of AI compute — exceeding today’s global installed base — followed by another 40 gigawatts the year after.
Notably, power scarcity actually strengthens Nvidia’s market position in Huang’s framework. His argument: Nvidia’s architecture delivers the highest tokens-per-watt globally, so as power tightens, customers gravitate toward platforms with the highest output per unit of energy.
He disclosed that Nvidia’s market value per gigawatt of compute has risen from roughly $18 billion to over $20 billion, now exceeding $40 billion, and still climbing. Nvidia accounts for about 40% of global cloud service provider capital expenditure and about 80% on the non-CSP side.
Long-Term Scale
Huang provided a framework for AI infrastructure’s long-term scale. Before AI, global computing infrastructure construction ran at roughly $500 billion annually. Combined with the end of Moore’s Law and AI-driven demand, he estimates annual buildout will reach $3-4 trillion by decade’s end, closer to $4 trillion, with about $1.5 trillion being pure compute.
“If someone tells me annual infrastructure construction and refresh reaches $10 trillion, I wouldn’t be surprised at all,” he said.
On financial strategy, Huang revealed Nvidia currently uses over 50% of free cash flow for share buybacks, reaching about 60% last quarter. He also highlighted two new growth vectors: enterprise software licensing (annual fees based on GPU count) and revenue-sharing arrangements with some cloud service providers — both independent of hardware sales.
Key Takeaways for Investors
- Credit markets vs. equity narrative: The CoreWeave CDS spike signaled genuine default concern, but Huang’s intervention shows Nvidia’s willingness to defend its ecosystem. Watch whether CDS spreads normalize in coming sessions.
- Diversification of AI demand: If Huang’s three-segment breakdown is accurate, AI infrastructure demand is far less concentrated than bears assume — reducing single-point-of-failure risk from OpenAI or Anthropic.
- Power as the real constraint: The bottleneck for 2026-2027 is physical, not financial. Companies with secured power and land positions may outperform.
- Nvidia’s pricing power: Rising market value per gigawatt and share gains in non-CSP capex suggest Nvidia’s dominance may strengthen even as total buildout faces constraints.
- Watch the buyback: With over 50% of free cash flow going to buybacks, Nvidia is signaling confidence in its own valuation — a meaningful signal during tech selloffs.
