Altman Says Laminate Capped OpenAI Sales; Nvidia Supply 2.3x

OpenAI chief executive Sam Altman said the company has at times been unable to sell more AI services because of a shortage of PCB laminate, the base material used to make circuit boards. “We can’t sell more services because the world is out of like PCB laminate or whatever”, Altman said at Semafor’s Silicon Valley & The World conference, in a clip of the interview posted on X. Semafor reported that Altman spoke on Monday, 16 November, in a 30-minute interview with Semafor’s Reed Albergotti.

Our number: on the supplier side, Nvidia more than doubled its supply and capacity commitments in one quarter, from $119 billion to $279 billion as of 26 July, 2.3 times, by our arithmetic from its second-quarter filings. Nvidia said the increase was “primarily related to the procurement of memory”. It reports third-quarter results on Tuesday, 17 November, after the US market close.

Business Pill · The Memory Wall

An evergreen Business Engineer explainer on why memory, not only the chip, limits AI performance. It teaches the general idea only and does not discuss OpenAI or Nvidia.

The key insight: Altman described OpenAI’s sales as set, at different times, by whichever input ran short: memory, rack assembly, data-center capacity or PCB laminate. On the supplier side, Nvidia’s supply and capacity commitments rose 2.3 times in one quarter, to $279 billion, mostly for memory, by our arithmetic from its filings.

What Altman Said

Altman listed the constraints in turn. “The supply chain is very complicated”, he said. “We’ve been constrained by different things at different times”. “Sometimes it’s been memory”. “Sometimes it’s been rack assembly”. “Sometimes it’s been data center capacity”. “Sometimes it’s been some random component”. PCB laminate was his example of such a component. “So it’s just a very complex supply chain now”, he said, “at a huge, huge scale”.

By our count, Altman named four specific constraints in the 30-second clip: memory, rack assembly, data-center capacity and PCB laminate. He did not say when each one bound, how many sales were lost or which customers were affected.

Nvidia’s supply and capacity commitments at the end of its first and second fiscal quarters of 2027, fro
Nvidia’s supply and capacity commitments at the end of its first and second fiscal quarters of 2027, from its second-quarter 10-Q and CFO commentary, which said the increase was primarily related to the procurement of memory. The 2.3x ratio is our arithmetic.

Where It Fits in the Interview

In the same interview, Semafor reported, Altman clarified earlier comments about people keeping RAM on personal devices at home “amid a global memory supply shortage”, saying they were “misinterpreted”. Asked whether there would be enough investor money to go around with Anthropic listing first, he said “That is not a worry”, which we covered in Altman on the Anthropic IPO. “There’s been way too much investor demand and not enough supply”, he said, according to Semafor.

The Supplier Side

Nvidia’s second-quarter 10-Q says the company “significantly increased our supply and capacity commitments from $119 billion last quarter to $279 billion as of July 26, 2026 to meet future demand”. Its CFO commentary for the quarter put the main driver as memory. Memory makers’ margins are running above Nvidia’s own, as we set out in SK Hynix and Micron at 80%. Nvidia reports its third-quarter figures tonight, against a $108 billion revenue guide.

The Business Lever

The lever is capacity, not demand. In Altman’s telling, OpenAI’s sales have at times been set by the scarcest input on a given day, from memory to a circuit-board material. Amazon’s chief executive made a similar point at the same conference, saying the “big chunks of capacity” are “largely reserved already in 27”, which we covered in Jassy on Amazon’s AI business.

The Structural Read

As we read the clip, Altman was describing a rotating constraint rather than a single shortage: four named inputs, each binding at a different time.

As we read Nvidia’s filings, the company’s response to tight inputs has been to commit cash ahead of demand, with memory the main item.

Sam Altman, Semafor Silicon Valley & The World

“So it’s just a very complex supply chain now at a huge, huge scale.”

Three Implications

FOR OPENAI Altman said OpenAI has at times been unable to sell more services because an input ran out; he gave no dates or volumes.

FOR NVIDIA Its commitments rose from $119 billion to $279 billion in one quarter, which Nvidia said was primarily for memory.

FOR SUPPLIERS Altman’s list ran from memory and rack assembly to a circuit-board material, which puts smaller component makers in the same chain as the chip makers.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Nine Layers of AI.

The framework says: “It is a layered industrial stack — and at any given moment, the binding constraint sits at a different layer than it did six months ago.”

As we read it, Altman’s list (memory, then racks, then data-center capacity, then laminate) is that moving constraint described from the buyer’s side. The framework predates the interview and does not discuss OpenAI or PCB laminate.

What Is Not Established

Altman gave no dates, volumes or dollar figures for the shortages he described, and the laminate remark does not appear in Semafor’s written account of the interview; the quotations here are from our transcription of the posted clip, checked against the video’s captions. Nvidia’s commitments are its own disclosure and cover all of its suppliers, not the specific shortages Altman named. We did not contact OpenAI or Nvidia.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

The Nine Layers of AI

A Business Engineer map of the AI industrial stack, from energy and chips to applications, and where the binding constraint sits.

Read the Map of AI →

The Bottom Line

Altman said OpenAI has been constrained at different times by memory, rack assembly, data-center capacity and even PCB laminate. On the supplier side, Nvidia’s supply and capacity commitments rose 2.3 times in one quarter, to $279 billion, mostly for memory, by our arithmetic from its filings.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. Altman’s words are from our transcription of a 30-second clip of his Semafor interview posted on X, checked against the video’s captions; the date, interviewer and his other remarks are from Semafor’s written account. Nvidia’s figures are from its second-quarter 10-Q and CFO commentary. The 2.3x ratio is our arithmetic. We did not contact OpenAI or Nvidia.

Sources: Clip of Sam Altman at Semafor Silicon Valley & The World (posted on X) · Semafor, Altman interview write-up, 16-17 Nov 2026 · NVIDIA Q2 FY2027 Form 10-Q (SEC EDGAR) · NVIDIA CFO Commentary Q2 FY2027, 8-K Ex.99.2 (SEC EDGAR)

Scroll to Top
FourWeekMBA