Jensen Huang Was Handed the Phrase Investors Are Trained to Distrust
Note: This image contains AI-generated elements.
Jensen Huang was handed the phrase investors are trained to distrust.
Asked by Axios co-founder Mike Allen in July whether the AI chip boom was due for a bust, Huang said no, not for a while. Allen gave him a way out: "So this time is different?" Huang turned the phrase around instead.
"This time is different because this is not demand-driven. This is industrially driven, meaning the fundamental technology of computers is changing."
That phrase has preceded many of the market dislocations of the past century.
Then the quarter landed. NVIDIA reported fiscal second-quarter revenue of $96.2 billion, up 106% year-over-year, with data center revenue of $89 billion, up 117%. Third-quarter guidance came in at $108 billion, and that outlook assumes no China data center compute revenue at all.
CFO Colette Kress gave the spending numbers behind the claim. Capital spending by the top five customers is tracking toward roughly $800 billion in 2026 and $1.3 trillion in 2027. Cloud-provider backlogs already under contract run to about $2 trillion.
Kress's 2026 figure sits about $200 billion above consensus.
Committed spending on that scale is the strongest version of Huang's argument, and it has one gap. Every one of those figures measures promised spending. None measures what the spending earns.
At Eagle Talon we weigh a leader's stated thesis against three measures. The capital allocation record behind it, the incentive design that rewards it, and how precisely that leader talks about returns. On the first two, Huang has put roughly $279 billion of Nvidia's own supply commitments behind the same view, mostly memory for Vera Rubin. On the third, he's specific about why the demand exists and he hasn't named a timeline on which the spending pays back.
When a cloud provider finally publishes a return figure on the compute it already bought, does Nvidia's $279 billion look early or exposed?
