A post on X by @scaling01 on 31 October, with Epoch AI’s ranking attached, said two of the ten largest AI data centers are in Malaysia, both operated by DayOne, and named ByteDance as DayOne’s largest customer. DayOne’s own filing does not name that customer.
Malaysia’s four AI data centers in Epoch AI’s Frontier Data Centers dataset hold 1,439,059 H100 equivalents of compute, 4.5 times the 320,440 of the dataset’s Chinese sites, by our arithmetic. Two of them, DayOne Nusajaya at 583,837 and DayOne Kempas at 480,639, rank sixth and eighth in the world.
Both sit in Johor. A post on X by @scaling01 pointed out the two DayOne sites on Epoch’s ranking and named ByteDance as DayOne’s largest customer; DayOne’s own IPO filing does not name the customer.
Short explainer · Malaysia Has 4.5x China’s Tracked AI Compute
A vertical news Short summarising this story. Figures come from Epoch AI and DayOne’s SEC filing; our arithmetic is labelled.
The key insight: As we read Epoch’s dataset, Malaysia’s four tracked AI data centers hold 4.5 times the compute of its Chinese sites, and all four are in Johor.
What the Epoch Data Shows
Epoch AI’s Frontier Data Centers dataset lists 96 data centers with an estimate of their current compute in H100 equivalents, a common unit for AI chips. In the copy we downloaded on 31 October, Colossus 2 leads at 1,111,673, followed by Anthropic-Amazon New Carlisle at 818,428 and Microsoft Fairwater Atlanta at 768,769.
DayOne Nusajaya is sixth, at 583,837 H100e and 473 MW, and DayOne Kempas is eighth, at 480,639 H100e and about 422 MW. Every other site in the top 10 is in the United States. Epoch leaves the owner and user fields blank for both DayOne rows.
Together the two DayOne sites hold 1,064,476 H100e, 16.4% of the top 10’s compute and 96% of Colossus 2’s, by our arithmetic. Epoch puts their combined capital cost at about $33.9 billion.
The 10 largest AI data centers by current compute in H100 equivalents, from Epoch AI’s Frontier Data Centers dataset, downloaded 31 October 2026. Shares are our arithmetic.
How Fast Johor Filled Up
Epoch’s timeline for DayOne Kempas puts it at 91,408 H100e on 22 January 2026 and 480,639 on 1 October, after its notes say buildings 7 and 8 were “assumed operational”. That is 5.3 times in about eight months, by our arithmetic.
Across all 96 sites, Epoch’s figures put 86.1% of tracked compute in the United States, 9.3% in Malaysia and 2.1% in China, by our arithmetic. Malaysia’s other two tracked sites, Bridge Data Centres MY06 and YTL Johor, are also in Johor.
What DayOne’s Filing Says
DayOne’s F-1 registration statement, filed with the SEC on 5 October, says its largest customer is “a global technology company with a leading short-form video platform serving a worldwide audience”. That customer was 69.4% of revenue in 2025 and 69.2% in the first half of 2026.
The filing says the anchor customer of its Johor Nusajaya campus grew its deployments “from an initial anchor commitment of 64MW to 838MW as of September 20, 2026”, 13.1 times, by our arithmetic. It also says Malaysia was 87.0% of revenue in the first half of 2026.
Chinese demand for AI computing rose 417% in the first quarter while supply rose 128%, per CAICT figures reported by Caixin, as we reported. Epoch’s data now shows more tracked AI compute in Johor than at its Chinese sites.
Epoch data also showed the United States holding 72% of the world’s AI chip compute, as we reported on 30 October. At the top of the ranking, Elon Musk has said Colossus 2 will hold about 1.2 million Blackwells by year-end, as we reported on 31 October.
The Structural Read
As we read the ranking, DayOne’s two Johor sites together hold 96% of Colossus 2’s compute, by our arithmetic.
As we read Epoch’s timeline, DayOne Kempas went from 91,408 to 480,639 H100 equivalents between 22 January and 1 October, 5.3 times.
As we read the filing, DayOne’s Johor anchor customer grew from a 64MW commitment to 838MW, 13.1 times by our arithmetic.
DayOne Data Centers, Form F-1, 5 October 2026
“a global technology company with a leading short-form video platform serving a worldwide audience”
Three Implications
FOR DAYONE Its filing says one customer provided 69.2% of revenue in the first half of 2026, and Malaysia 87.0%.
FOR THE COMPUTE MAP Epoch’s figures put 9.3% of tracked compute in Malaysia and 2.1% in China, by our arithmetic.
FOR THE TOP OF THE RANKING Colossus 2 leads Epoch’s list at 1,111,673 H100 equivalents, ahead of Anthropic-Amazon New Carlisle at 818,428.
The framework says the infrastructure constraint is “creating geopolitical leverage that operates independently of who has the best models.”
As we read Epoch’s data, two of the world’s ten largest AI data centers sit in Malaysia. The framework is the owner’s argument about AI infrastructure and does not discuss DayOne.
Epoch’s compute figures are its own estimates, and its dataset covers only the sites it has identified, so compute elsewhere, including in China, may be undercounted. Epoch does not name the users of the DayOne sites. DayOne’s filing does not name its largest customer, and neither source says who runs which chips in Johor.
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.
Malaysia’s tracked AI compute in Epoch’s data, 1,439,059 H100e, is 4.5 times China’s 320,440, by our arithmetic, and DayOne’s two Johor sites rank sixth and eighth in the world. DayOne’s filing says one short-video customer provided 69.2% of its revenue.
A note on sourcing. Compute figures come from Epoch AI’s Frontier Data Centers dataset, downloaded on 31 October 2026; customer and capacity figures come from DayOne’s Form F-1 filed with the SEC. Shares and ratios are our arithmetic. We did not contact DayOne or ByteDance.
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.
