Ed Zitron · Sep 22, 2026
Ed Zitron argues GPU sales are outrunning usable AI data center capacity
Markdown · JSON · Original source
In a newsletter published September 22, 2026, Ed Zitron argues that sales are being mistaken for evidence of working capacity. He says Microsoft’s broad claims about data center capacity obscure how much is specifically for AI and how much is installed and usable. He estimates that Microsoft has $50 billion to $100 billion in GPUs waiting to be installed, and that hundreds of billions of dollars in AI hardware across the industry may be sitting in warehouses or unpowered data centers. His case draws on company statements, financial disclosures, analyst estimates and sources familiar with infrastructure; he acknowledges that some short-lived assets may be CPUs rather than GPUs and that his estimates are imperfect. Zitron argues that this gap matters because investors may read chip purchases and capacity announcements as evidence of revenue-generating infrastructure, while projects may take years to build, connect to power and serve customers. He warns that purchases made well ahead of installation, combined with demand concentrated among a few AI labs, could leave companies with costly equipment and debt if demand does not materialize. He calls for clearer disclosure of operational capacity and GPUs in storage. The scale of installed, revenue-generating AI capacity remains unclear, and Zitron’s estimates depend on incomplete company disclosures and assumptions.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
Ed Zitron · Article
Source details
Original source date: Sep 22, 2026 (2026-09-22T14:35:46.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-10-31T21:27:03.294Z. First observed: 2026-10-23T19:47:43.443Z. Retrieved: 2026-10-23T19:49:39.628Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 23, 2026; originally published September 22, 2026. Zitron says his industry-wide figures are estimates and rely on assumptions about asset types, costs and the share of attributable to AI infrastructure. Company disclosures do not consistently distinguish AI-specific capacity from overall capacity, or clarify whether reported capacity is operational, connected to power or merely secured. The text includes claims based on unnamed sources and allegations of deception; the retrieved evidence does not independently verify those allegations.
Open question
How many sold to hyperscalers are installed, powered and generating revenue, rather than waiting for data centers or power?
Source-stated
Ed Zitron argues that Microsoft’s capacity announcements do not make clear how much capacity is specifically for AI or operational.
Microsoft claims, as reported, that it has 12GW of capacity — 3GW of which came online in the last three quarters! — but only 2GW of that is AI data center capacity
Source-stated
Zitron estimates that Microsoft has between $50 billion and $100 billion in GPUs that have not been installed.
In other words, Microsoft is warehousing anywhere from $50 billion to $100 billion in GPUs
Source-stated
Zitron estimates that the companies in his analysis have more than $374 billion in construction in progress, excluding several companies and projects.
Across hyperscalers including Google, Meta, Oracle, Amazon, SpaceX, and Tesla, neoclouds like CoreWeave and IREN, and colocation companies like Core Scientific and Applied Digital, there is over $374 billion in construction in progress
Source-stated
Zitron argues that chip sales and purchases made ahead of installation are not reliable evidence of present demand for operational AI compute.
NVIDIA has created a remarkable illusion perpetuated by the media — that GPU sales are a direct measurement of the actual demand for AI compute, rather than a measurement of how a few companies are willing to invest in an idea two years in advance
Source-stated
Zitron argues that demand concentrated among a few AI labs can make demand for compute appear broader than it is.
If “demand is outstripping supply” because of millions of customers begging for AI compute, that’s very different to “demand outstripping supply” because three customers are taking up most or all of the capacity.
Source and contributor context
Ed Zitron · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Graphics processing unit
A chip that performs many calculations at once; the main hardware used to train and run AI models.
AI compute
The computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Capital expenditure
Money a company spends on long-lasting assets such as buildings, chips and power equipment.
Editorial provenance
Version 1. Publication ID: 879c6879-4631-ba67-eada-e3a7e5c088eb.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-10-31T21:27:02.217Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_screen.
Observation: 59cf257e2df9593205e6e2ac1922d6e1199f7f5715e42c4965b8bb2c9f399adf. Source text hash: e4a73110d40278b0d817b34f1d521cb29fe4337c29cd5b938f43352f315a59f7.
Revision and source-check history
Version 1 · 2026-10-31T21:27:03.294Z
Ed Zitron argues GPU sales are outrunning usable AI data center capacity
In a newsletter published September 22, 2026, Ed Zitron argues that sales are being mistaken for evidence of working capacity. He says Microsoft’s broad claims about data center capacity obscure how much is specifically for AI and how much is installed and usable. He estimates that Microsoft has $50 billion to $100 billion in GPUs waiting to be installed, and that hundreds of billions of dollars in AI hardware across the industry may be sitting in warehouses or unpowered data centers. His case draws on company statements, financial disclosures, analyst estimates and sources familiar with infrastructure; he acknowledges that some short-lived assets may be CPUs rather than GPUs and that his estimates are imperfect. Zitron argues that this gap matters because investors may read chip purchases and capacity announcements as evidence of revenue-generating infrastructure, while projects may take years to build, connect to power and serve customers. He warns that purchases made well ahead of installation, combined with demand concentrated among a few AI labs, could leave companies with costly equipment and debt if demand does not materialize. He calls for clearer disclosure of operational capacity and GPUs in storage. The scale of installed, revenue-generating AI capacity remains unclear, and Zitron’s estimates depend on incomplete company disclosures and assumptions.
Original source
Ed Zitron · Article
Source details
Original source date: Sep 22, 2026 (2026-09-22T14:35:46.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-10-31T21:27:03.294Z. First observed: 2026-10-23T19:47:43.443Z. Retrieved: 2026-10-23T19:49:39.628Z.
Limits
Text as retrieved on October 23, 2026; originally published September 22, 2026. Zitron says his industry-wide figures are estimates and rely on assumptions about asset types, costs and the share of attributable to AI infrastructure. Company disclosures do not consistently distinguish AI-specific capacity from overall capacity, or clarify whether reported capacity is operational, connected to power or merely secured. The text includes claims based on unnamed sources and allegations of deception; the retrieved evidence does not independently verify those allegations.
Open question
How many sold to hyperscalers are installed, powered and generating revenue, rather than waiting for data centers or power?
Source-stated
Ed Zitron argues that Microsoft’s capacity announcements do not make clear how much capacity is specifically for AI or operational.
Microsoft claims, as reported, that it has 12GW of capacity — 3GW of which came online in the last three quarters! — but only 2GW of that is AI data center capacity
Source-stated
Zitron estimates that Microsoft has between $50 billion and $100 billion in GPUs that have not been installed.
In other words, Microsoft is warehousing anywhere from $50 billion to $100 billion in GPUs
Source-stated
Zitron estimates that the companies in his analysis have more than $374 billion in construction in progress, excluding several companies and projects.
Across hyperscalers including Google, Meta, Oracle, Amazon, SpaceX, and Tesla, neoclouds like CoreWeave and IREN, and colocation companies like Core Scientific and Applied Digital, there is over $374 billion in construction in progress
Source-stated
Zitron argues that chip sales and purchases made ahead of installation are not reliable evidence of present demand for operational AI compute.
NVIDIA has created a remarkable illusion perpetuated by the media — that GPU sales are a direct measurement of the actual demand for AI compute, rather than a measurement of how a few companies are willing to invest in an idea two years in advance
Source-stated
Zitron argues that demand concentrated among a few AI labs can make demand for compute appear broader than it is.
If “demand is outstripping supply” because of millions of customers begging for AI compute, that’s very different to “demand outstripping supply” because three customers are taking up most or all of the capacity.
Source and contributor context
Ed Zitron · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Graphics processing unit
A chip that performs many calculations at once; the main hardware used to train and run AI models.
AI compute
The computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Capital expenditure
Money a company spends on long-lasting assets such as buildings, chips and power equipment.
Editorial provenance
Version 1. Publication ID: 879c6879-4631-ba67-eada-e3a7e5c088eb.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-10-31T21:27:02.217Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_screen.
Observation: 59cf257e2df9593205e6e2ac1922d6e1199f7f5715e42c4965b8bb2c9f399adf. Source text hash: e4a73110d40278b0d817b34f1d521cb29fe4337c29cd5b938f43352f315a59f7.
passed · 2026-10-31T21:27:02.217Z · anthropic/claude-opus-5.5
Publication: 879c6879-4631-ba67-eada-e3a7e5c088eb
attribution — pass: The piece is a single newsletter essay by Ed Zitron on his own site. The reading attributes the arguments, estimates and allegations to him rather than presenting them as established fact.
Today’s newsletter is about a cruel truth: that NVIDIA’s revenue growth is almost entirely the result of speculative purchases
summary — pass: The reading leads with Zitron's thesis, his Microsoft and industry-wide estimates, why he thinks the gap matters, and his call for disclosure. It keeps his caveats and his force. GPU, AI compute and capex are defined beside the reading. 'Hyperscalers' in the question is unexplained, but the summary itself stays readable without it.
In other words, I think everybody is completely wrong about data center capacity, and it’s going to be a nightmare to untangle once they work it out.
date — pass: The stated date of September 22, 2026 matches the UTC day of originalPublished. modifiedAt equals the publication time. The evidence is a publisher page, so no archive disclosure is needed. The reading frames no forecast as an outcome and uses no hindsight.
with Bloomberg reporting on September 11, 2026 that it had “around” 2GW
support — pass: Each claim matches the source: - Microsoft's $50–100B warehoused-GPU estimate. - Over $374B in construction in progress, with exclusions. - The concession about CPUs. - The 'imperfect' estimates. - Concentrated demand among a few AI labs. - Debt and write-down risk. - The disclosure questions. No implied benefits or costs go beyond the source.
The following are imperfect yet, I believe, well-founded estimates.
media — pass: No media candidates were supplied and none was selected, so no media identity or relevance issue arises.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
A chip that performs many calculations at once; the main hardware used to train and run AI models.
Definition referenceThe computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Definition referenceMoney a company spends on long-lasting assets such as buildings, chips and power equipment.
Definition referenceA chip that performs many calculations at once; the main hardware used to train and run AI models.
Definition referenceA chip that performs many calculations at once; the main hardware used to train and run AI models.
Definition referenceThe computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Definition referenceMoney a company spends on long-lasting assets such as buildings, chips and power equipment.
Definition referenceA chip that performs many calculations at once; the main hardware used to train and run AI models.
Definition reference