Is Nvidia's Chip Shortage Real? One Loud Critic Says No

Ed Zitron, who runs a public relations firm and has become one of the loudest critics of AI spending, went on Yahoo Finance's The 8:30 and said that Nvidia's chip shortage is largely manufactured. His line, quoted in the headline of the Yahoo Finance report, was that there is "not scarcity other than that which is created by Jensen Huang", Nvidia's chief executive.

That is an opinion, not a finding. Nvidia was not quoted in reply in the Yahoo report. But the argument touches real numbers, and they are worth understanding.

What is Nvidia actually selling?

Graphics processing units, or GPUs. Originally built for video games, they turned out to be very good at the heavy arithmetic AI needs. Companies like Microsoft, Amazon and Alphabet buy them in huge numbers to fill data centres, the warehouses of computers that run AI tools such as chatbots.

What does Zitron claim?

He makes several separate points, and it helps to keep them apart.

• The shortage is a sales tactic. He says Huang tells buyers to order now or lose supply in a few months, and that this pushes chips into the market faster than real demand would. In finance this is sometimes called channel stuffing: shipping more product to buyers than they need yet. That is his description, not a finding by any regulator or auditor.
• Scarcity makes no sense if there is too much stock. He asked how a product can be scarce when, in his view, the company has too many of them.
• Consumers pay for it. He says the push for GPUs is raising the cost of consumer electronics.
• The payoff is unclear. He asks what all these extra data centres will actually deliver.
• A few AI labs hold it all up. He calls OpenAI and Anthropic "load-bearing" for Microsoft, Alphabet, Amazon, CoreWeave and Nebius, meaning those firms' plans lean heavily on the two labs staying funded. He also argued a government rescue would not stop a market fall if they faltered.

What do Nvidia's own numbers say?

In the quarter ended 26 July 2026, Nvidia reported revenue of $96.2 billion, up 106% from a year earlier. Its data centre business, the part that sells these chips, brought in $89.0 billion, up 117%. It told investors to expect about $108 billion next quarter, give or take 2%, without counting any China data centre sales.

Those figures show demand that is very large. They do not, by themselves, show that the demand is real in the sense of lasting. Customers can order heavily and still end up with more than they use. That is the question the two sides are really arguing about.

Who owes Nvidia the money?

This is the most concrete piece of evidence in the debate. Steve Eisman, the investor made famous by The Big Short, recently pointed to Nvidia's quarterly filing. It shows that five direct customers, who are not named, account for 70% of the money Nvidia is owed for chips it has already shipped. That money is called accounts receivable.

Picture it this way. If customers owe Nvidia ₦100 in total:

• ₦22 is owed by the biggest buyer
• ₦14, ₦13, ₦11 and ₦10 by the next four
• Only ₦30 is spread across everyone else

Two cautions. First, this is receivables, money owed, not revenue, and the two have been confused in some coverage. Second, a supplier of very expensive hardware will naturally have a few giant customers. Concentration is a risk, but it is not proof of anything improper.

Is there other evidence of a slowdown?

The Yahoo report points to Amazon dropping a data centre site in Loudoun County, Virginia. Fortune reports that Amazon withdrew plans for the former George Washington University campus after the community pushed back. That stated reason is local opposition, not weak demand, so it is a thin piece of evidence for Zitron's case.

Separately, Bloomberg reported in September that Microsoft has about 12 gigawatts of data centre capacity, with roughly 2 gigawatts centred on AI chips, and aims for around 38 gigawatts by 2032. A company planning to triple its capacity is not behaving as if it expects a crash. It could also be making the same bet Zitron questions.

The Jargon

• GPU: graphics processing unit, the chip used to train and run AI.
• Channel stuffing: shipping more product to buyers than they need yet, to make sales look stronger.
• Accounts receivable: money owed to a company for goods or services already delivered.
• Hyperscaler: a very large cloud company such as Microsoft, Amazon or Alphabet.
• Gigawatt (GW): a unit of electrical power, used to size data centres.
• Load-bearing: here, a firm other companies' plans depend on.

What to watch

This section is Rainingmoney's reading.

• Whether Nvidia's receivables rise faster than its sales next quarter. That would suggest buyers are taking longer to pay.
• Whether OpenAI and Anthropic keep raising money. Much of the spending chain depends on it.
• Whether any large buyer publicly delays or cuts a chip order, rather than a building project.
• Nvidia's reply. None was in the Yahoo report, and a response with numbers would move this debate.

Nvidia shares closed at $210.96 on 23 October, down 0.59%.

What we do not know

Nobody outside the companies knows how many chips are sitting unused, or how much of today's ordering is real need and how much is fear of missing out. Zitron's view is one side of a live argument, and the sales figures support the other side for now.