Skip to content

Invest1 publisher3 min readPublished

Nvidia guided 70% growth and can supply 70% of demand. The second number is the thesis.

The gap between Nvidia's forecast and the 44% sell-side consensus is an argument about other companies' factories, and the Korean stocks that jump when Jensen Huang books dinner are pricing it.

The Investor · Invest desk

What happened

  • Huang said supply currently covers 70% of demand from existing customers, and that demand runs much higher.
  • Asked to rank his tightest constraint, Huang declined and said the entire supply chain is challenged with everyone running flat out.
  • The stock rose more than 4% in after-hours trading on the preliminary forecast.

Compiled by The InvestorSomething wrong?How this is made

Why it matters

  • constraint A capacity plan keyed to Nvidia's guide is keyed to Nvidia's supplier negotiations, and the buyer has no way to hedge that assumption.
  • decision Buyers must choose between committing earlier on Nvidia's terms or planning around a delivery date they cannot independently verify.
  • exposure Any supplier named as the binding constraint gets re-rated on the naming, so the incentive runs against telling planners which input is short.
  • precedent The next quarter is now a test of whether supply arrived rather than whether customers did, and a miss against this guide reads as a supply-chain failure.

The fill rate is the number worth doing arithmetic on. If the $96.2 billion Nvidia booked in the quarter [1] is roughly what it managed to ship, and that shipment covered 70% of what existing customers asked for [6], then those accounts wanted something like $137 billion of product and went home about $41 billion short in a single quarter [13]. The calculation is crude, since recognised revenue and shipped supply are not the same thing. No version of it produces a market where the buyer sets terms.

Keep that 70% away from the other one. Next fiscal year's growth guide is also 70%, against a 44% consensus [3][4], and the two measure different quantities: one is the share of demand Nvidia can serve, the other is how much bigger the shipments get. Read together, the guide is a supply forecast. Consensus at 44% was a more cautious guess about what Nvidia could extract from power, shells, memory and wafers, and it was 26 points light, roughly 1.6 times off in relative terms [14]. Fortune put the gap at more than $100 billion of revenue above where analysts were modelling [5]. Nvidia's read of its own supply chain moved further than anyone's read of its customers.

Which is why the withheld answer matters. A Goldman Sachs analyst asked for the constraints ranked: data centre power, shell availability, memory pricing, wafer foundry access [7]. Huang said there was something funny he could say, then did not say it, offering instead that the entire supply chain is challenged and everyone is running flat out [8]. CFO Colette Kress also pointed to heavy constraints, per Fortune's account of the call [9].

The reason for the non-answer sits a few paragraphs further into the same article. LG rose 30% on reports that its chairman would meet Huang, with listed affiliates up between 17% and 29% [12]. Huang's own joke on the call was that his dinner companions' stock prices double the next day [11]. A ranked bottleneck names a supplier and re-rates it before Nvidia has finished negotiating with that supplier. So the disclosure a capacity planner most needs, which input gates delivery, is the one Nvidia has the clearest commercial reason to keep general.

In the meantime, allocation does the work price would otherwise do. Larry Ellison's account of a 2024 dinner at Nobu Palo Alto with Elon Musk was that the two of them begged for GPUs [16]. That is what procurement looks like when 30% of the order book cannot be served. The after-hours move of more than 4% [10] priced the guide. Nothing prices the 30% that is not for sale.

What to watch

  • Whether Nvidia repeats the fill-rate disclosure next quarter or quietly drops it, which is the only public read on unserved demand.
  • Capacity or pricing announcements from memory and foundry suppliers, which would identify the constraint Huang declined to rank.
  • Whether large buyers begin disclosing slipped data-centre dates or deferred deliveries in their own filings.
Loading claim ledger
Loading source directory links
Loading share composer
Loading topic controls
Loading related stories