Epoch AI now estimates that OpenAI’s Stargate site in Abilene, Texas, holds 1,018,696 H100-equivalents of compute, up from 764,022 in its previous entry, according to its Frontier Data Centers data, which we pulled on 1 November. By our arithmetic, that is a 33% increase, and it moves Abilene from fourth to second among the sites Epoch tracks.
The new entry, dated 1 November, reads “Buildings 7-8 are estimated to be operational”. Epoch adds that “Oracle gave no date for the last two buildings on its September 10 call”, so the step is Epoch’s estimate from satellite imagery, not a confirmed handover.
Business Pill · Lead Time Explained
A short explainer on lead times: why big infrastructure takes years from plan to operation. It teaches the general idea only and says nothing about this site or Epoch’s estimates.
The key insight: Epoch AI now estimates OpenAI’s Abilene site at 1,018,696 H100-equivalents, second only to Colossus 2 at 1,111,673 in its data; the step from 764,022 rests on Epoch marking buildings 7 and 8 as estimated to be operational, a date Oracle has not given.
Epoch’s timeline for the site lists six operational buildings, 618 megawatts of IT power, 764,022 H100-equivalents and a total capital cost of about $23.4 billion at its 10 September entry. The 1 November entry lists eight buildings, 824 megawatts of IT power, 1,018,696 H100-equivalents and about $31.2 billion. Epoch’s hub lists Oracle as owner, OpenAI as user, and B200 and B300 as the chips.
The September entry rests on Oracle’s earnings call. Epoch quotes Oracle’s Clay Magouyrk: “6 of the 8 campus buildings, representing 618 megawatts and 75% of total capacity have now been delivered to the customer”.
For the last two buildings, Epoch reasons from the build pace. It writes that on 24 March “Building 8 roughly looked like where Building 3 was in September”, and that Building 3, by Epoch’s estimate, became operational seven months later, which “naively suggests Building 8 will be operational by early November”.
The five largest sites by H100-equivalents in Epoch AI’s Frontier Data Centers data. OpenAI’s Stargate site in Abilene rose from 764,022 on our 31 October pull to 1,018,696 on 1 November, after Epoch marked buildings 7 and 8 as estimated to be operational; the other four did not change.
Where Abilene Now Ranks
In the same Epoch data, xAI’s Colossus 2, listed under SpaceXAI, still ranks first at 1,111,673 H100-equivalents. By our arithmetic, Abilene’s new figure is about 92% of Colossus 2’s, a gap of 92,977. Musk has said Colossus 2 will hold about 1.2 million Blackwells by year-end, as we reported in our Colossus 2 piece; Epoch’s figure is its own estimate, in H100-equivalents.
Behind the top two, Epoch lists the Anthropic-Amazon site in New Carlisle, Indiana, at 818,428 H100-equivalents and Microsoft’s Fairwater Atlanta at 768,769. On our 31 October pull of the same data, Abilene ranked behind both.
The Business Lever: Capacity in Two-Building Steps
As we read Epoch’s timeline, Abilene has added capacity in steps of two buildings, from two buildings at the end of 2025 to an estimated eight now. Our power piece worked through the electricity a site of this scaleneeds, and our US compute-share piece and Malaysia piece used the same Epoch data.
Our earlier coverage of a report on off-balance-sheet financing described the Abilene chain: it says OpenAI leases the facility and its chips from Oracle, Oracle leases it from Crusoe Energy, and Crusoe co-develops the site with Blue Owl Capital and Primary Digital Infrastructure.
The Structural Read
As we read Epoch’s timeline, every step in Abilene’s estimate since the end of 2025 has come in pairs of buildings, with H100-equivalents and IT power each rising by a third in the latest step.
By our arithmetic, Abilene’s new figure is about 92% of Colossus 2’s, and the 92,977 gap is smaller than any single step Epoch has recorded at Abilene since September 2025.
As we read the notes, Epoch dates its September entry to Oracle’s call and its November entry to satellite-based pacing, so the two steps rest on different kinds of evidence.
Clay Magouyrk, Oracle, 10 September 2026, as quoted by Epoch AI
“6 of the 8 campus buildings, representing 618 megawatts and 75% of total capacity have now been delivered to the customer”
Three Implications
FOR READERS OF COMPUTE RANKINGS Epoch labels the last two Abilene buildings as estimated, so the No. 2 rank rests on its pacing method rather than an Oracle statement.
FOR OPENAI’S CAPACITY Epoch’s hub lists OpenAI as the site’s user and B200 and B300 as its chips.
FOR THE LEADERS’ GAP In Epoch’s data, the gap between Colossus 2 and Abilene is 92,977 H100-equivalents, by our arithmetic.
The framework says: “The $1T capex cycle is being underwritten by a $1T concentrated demand commitment from two counterparties whose combined free cash flow is deeply negative.”
As we read it, Abilene is one of the sites that capex cycle pays for; the framework does not discuss Epoch’s estimates or this site.
Epoch marks the last two buildings as estimated, and Oracle has not given a date for them. H100-equivalents are Epoch’s way of putting different chips on one scale; they are not a count of chips. The ranking covers only the sites Epoch tracks. The 33% rise, the 92% ratio and the 92,977 gap are our arithmetic from Epoch’s figures.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Epoch AI now estimates OpenAI’s Abilene site at 1,018,696 H100-equivalents and 824 megawatts of IT power, up 33% by our arithmetic from its September entry, after marking buildings 7 and 8 as estimated to be operational. That puts Abilene second in Epoch’s data, behind Colossus 2 at 1,111,673.
A note on sourcing. Site figures and construction notes come from Epoch AI’s Frontier Data Centers data (CC-BY), which we pull daily; the earlier figures are from our 31 October pull and Epoch’s own timeline. The Oracle quote is as Epoch quotes it. Comparisons labelled as our arithmetic are ours.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
