AI critic Ed Zitron has accused Nvidia chief executive Jensen Huang of manufacturing the GPU scarcity his company sells against, arguing on Yahoo Finance’s The 8:30 that there is no shortage except the one Huang himself has created. Zitron, of public relations firm EZPR and a well-known critic of the AI spending boom, said Huang pressures buyers to purchase graphics processors now or risk going without them in a few months, a practice he described as channel stuffing and called morally reprehensible. The article carrying the interview, published on October 24, contains no direct response or rebuttal from Nvidia, so the claims stand here as Zitron’s criticism, not as established fact. It lands as scrutiny of AI infrastructure spending intensifies.
Zitron’s Scarcity Accusation
Zitron’s central argument is that Nvidia is creating the scarcity it then uses as a selling point. In his characterisation, Huang tells buyers to purchase now or risk unavailability in a few months, forcing GPUs down the channel. He argued that this raises prices for consumer electronics and benefits Huang and Nvidia shareholders at other people’s expense.
When the hosts put it to him that Huang’s pitch might simply be a demand signal, given semiconductor lead times that stretch over years, Zitron rejected the idea. There is no scarcity, he argued in substance, except that created by Huang, asking how there could be a shortage of a product of which Nvidia has too many. He questioned what additional data centres actually deliver, pointing to a gap between the capacity being announced and the capacity actually running.
The Concentration Figures Behind The Claim
Zitron backed his case with figures on how concentrated Nvidia’s customer base has become. Five companies accounted for about 70% of Nvidia’s accounts receivable in the first half of the year, a concentration risk he noted that ‘Big Short’ investor Steve Eisman has also recently flagged. If a handful of buyers dominate what Nvidia is owed, Zitron’s argument runs, the company’s fortunes rest on very few relationships.
Citing a Bloomberg report, he said Microsoft has about 2 gigawatts of AI compute online out of 12 gigawatts reported, and noted that Amazon recently cancelled a data centre in Loudoun County, Virginia. He further argued that OpenAI and Anthropic have become ‘load-bearing’ for Microsoft, Alphabet, Amazon, CoreWeave and Nebius, and said that if those two AI labs falter, they should be allowed to fail, in his words: ‘Let them die.’
Hosts Push Back On Crash Warning
The hosts pushed back on Zitron’s bleakest conclusion. They noted that a record share of Americans own stocks, and that Bloomberg’s Eric Balchunas has argued this could pressure the government to intervene in a crash. Zitron answered that a bailout would not hold the market up anyway, because GPU sales are what matter and are keeping the market alive. Unless buyers keep purchasing more GPUs at ever-higher prices, he argued, the market will crash either way.
The article closes with a disclosure that internal AI tools helped analyse interview transcripts, research and draft the post, and that humans reviewed, edited and fact-checked it before publication.
Why It Matters
Zitron’s attack matters because it goes after the premise underneath the AI build-out rather than just its price tag. If GPU demand is partly manufactured by sales pressure rather than genuine shortage, then the spending plans of hyperscalers, neoclouds and AI labs rest on shakier ground than their announcements suggest. His concentration figures point the same way: a market where five buyers account for most of what Nvidia is owed is vulnerable to any one of them slowing down. Investors have already begun pricing some of that doubt in, with AI infrastructure names sliding even as their suppliers insist demand is real. The debate now is whether the sceptics are early or simply wrong, and Nvidia’s next results will be read as evidence either way.
Background
Concern about AI spending has grown alongside the borrowing funding it, with AI-related debt more than quadrupling to $466 billion on Oracle and CoreWeave borrowing alone, while Nvidia’s own hardware, including the DGX Spark desktop system, keeps pushing deeper into workplaces and data centres.
Source: Yahoo Finance
0 Comments