Jensen Huang denied reports of delays to Nvidia's Vera Rubin AI systems, declaring the chips already in production with "giant amounts of production incoming," a pledge that matters because the company has hung a $1 trillion revenue target on Rubin and its predecessor Blackwell for 2026 and 2027.
KeyBanc analyst John Vinh and research firm SemiAnalysis had flagged thermal issues, high-bandwidth memory qualification snags, and networking manufacturing troubles as potential obstacles. Huang's rebuttal, relayed by Bloomberg, was blunt: the delay talk is "not true."
The revenue math is striking. Nvidia now sees $1 trillion from the two architectures across 2026 and 2027, double the $500 billion it projected for the overlapping 2025-2026 period. Vera Rubin is supposed to drive that step-up by slashing AI inference costs, a market where Nvidia still dominates despite rising competition.
Wall Street is largely convinced. Sixty-two of 66 analysts rate the stock a buy, with a median 12-month price target of $300 implying a 45% climb. Yet the same models show earnings-per-share growth accelerating to 88% in fiscal 2027 from 60% last year, and a scenario where a Nasdaq-100 multiple of 25.5 times applied to $16.06 of EPS in fiscal 2029 yields a $409 share price, roughly double where the stock trades today.
That $409 figure requires a straight line from today's "giant amounts" to 2029's earnings, assuming the multiple holds and the inference cost curve bends exactly as planned. It is a lot of faith to place on a chip that analysts were worrying about overheating last week.
The near-term test is whether Huang's "giant volumes" materialize on schedule and whether the inference cost advantage actually widens the moat. If the thermal and HBM headaches were real, the production ramp will show it; if they weren't, the delay narrative was just noise. Either way, the trillion-dollar forecast is now on the clock.