DayOne Data Centers Limited on Oct. 5 filed a preliminary Form F-1 registration statement for a U.S. IPO, offering public investors one of the clearest operating snapshots yet of the AI data-center buildout outside the United States and China. The Singapore-headquartered company says it expects to list American depositary shares on Nasdaq under the symbol DODC, though the filing leaves the number of shares, price range and deal size blank.
What matters immediately is not the unfinished offering but the operating evidence inside it. DayOne reported $512.024 million in revenue for the six months ended June 30, up from $151.500 million a year earlier, alongside a net loss that widened to $77.209 million from $12.573 million. As of Sept. 20, it listed 2,281 megawatts of customer bookings, 953 MW of billings, 1,328 MW of backlog, 962 MW of capacity in service and 1,328 MW under construction. For investors, cloud buyers, utilities and equipment suppliers, the prospectus turns the AI-infrastructure boom into a more answerable question: is this becoming a durable operating business, or is much of the value still trapped in a capital-intensive pipeline?
The delivery ladder behind the gigawatts
DayOne’s filing is useful because it separates terms that are often blurred together in the AI-infrastructure rush. The company says it has secured 4.6 gigawatts of Resources across ten markets and about 2.3 gigawatts of Bookings, primarily from seven global hyperscale and technology customers. But Resources, as DayOne defines them, are not live compute. They include capacity already in service, capacity under construction and secured powered land with binding land and power arrangements plus the necessary approvals and permits.
That makes DayOne’s own operating ladder more important than its headline gigawatt figure. Bookings are legally binding customer commitments. Billings are the portion already producing income. Backlog is contracted capacity that is not yet income-generating. Capacity in service is ready for customers; capacity under construction is not. Put differently, 2,281 MW of demand has moved further toward monetization than 4.6 GW of Resources, but even that booked volume is not the same thing as energized, billable capacity.
The progress is still substantial. Billings rose from 213 MW in June 2025 to 666 MW in June 2026 and then to 953 MW by Sept. 20, according to the filing. Utilization stood at 99.1%, and pre-commitment at 99.3%, both signs that DayOne is not building into empty rooms. Substantially all revenue comes from long-term data-center service contracts for space, power and cooling, with customer power consumption typically passed through. That supports the argument that hyperscalers are outsourcing a hard piece of AI expansion to specialist operators. It also means the key test is execution: getting land, power, permits, interconnection, labor, cooling and equipment aligned so that a signed contract actually turns into billings and, eventually, cash.
Demand is real; concentration is, too
The company can fairly point to fast adoption. DayOne was incorporated in 2022, delivered its first campus in Johor, Malaysia, in 2023, and has since expanded into Batam, Singapore, Greater Bangkok, Tokyo, Greater Helsinki, Kuala Lumpur, Zaragoza and Kyushu. As of Sept. 20, it said it had bookings from 15 customers, with three hyperscale customers each deploying around or above 200 MW across multiple markets. For cloud buyers hunting AI capacity outside the most constrained U.S. hubs, that suggests DayOne is solving a real procurement problem.
But the filing also shows how narrow the business still is. Malaysia supplied 87.0% of first-half 2026 revenue, and the top two customers accounted for 84.3%. Those figures do not negate the growth story; they define its current fragility. A platform concentrated in the SIJORI corridor and nearby markets may be able to secure land and power faster than operators in more saturated locations, but it is also more exposed to local grid policy, permitting, weather, currency and network conditions. And when a small number of hyperscalers represent most of the revenue, they tend to hold significant leverage over schedules, pricing and future expansion plans.
That concentration matters for suppliers too. Utilities, transformer makers, cooling vendors, contractors and fiber providers can read DayOne’s numbers as evidence of real build demand, especially with 1,328 MW under construction. They can also see that the pace of procurement depends heavily on a few customers continuing to take down capacity on time.
The IPO test is financing as much as demand
If the operational question is delivery, the market question is balance-sheet stamina. DayOne says it raised $1.9 billion through Series A and B financings in 2024, $1.3 billion in 2025 and $3.2 billion in 2026 in connection with its Series C financing. As of June 30, the filing showed about $4.35 billion of long-term borrowings and $430.4 million of finance leases, alongside a shareholders’ deficit. First-half revenue grew 238% year over year, but the net loss widened sharply, reminding investors that data-center development can scale revenue well before it proves equity returns after depreciation, interest and construction costs.
That is why the filing’s non-GAAP profitability picture needs context. DayOne reported a first-half net loss margin of 15.1% and an adjusted EBITDA margin of 40.2%. The adjusted figure may help show the earning power of operating sites before major non-cash and financing charges, but it does not erase the cash demands of turning backlog into live capacity. In this sector, a contract signed ahead of energization can be economically valuable and still leave the operator carrying months or years of delivery risk.
The missing terms of the IPO underline how early the public-market test still is. The preliminary prospectus does not disclose the number of ADSs, price range, expected proceeds, valuation, use of proceeds or final governance terms. It also stops short of giving outsiders the project-by-project schedule they would need to map every megawatt from secured land to service date. Nor does it identify the seven main hyperscale customers or spell out termination rights, renewal economics or the margin structure on contracted power. The company also says its market estimates rely in part on a commissioned Structure Research report and warns investors not to give those estimates undue weight.
So what does DayOne’s prospectus actually settle? It makes the AI data-center boom more legible as an operating business: there is real revenue, real utilization, real contracted demand and a real multi-country buildout. But it also shows that the sector’s headline gigawatt story still rests on a narrower set of facts than the marketing often implies: a few customers, a few key markets, a lot of capital and a long chain of things that must go right before booked capacity becomes dependable, billable infrastructure. The next filings will matter less for the slogan-size pipeline number than for a simpler measure of proof: how fast backlog turns into billings without the debt and concentration swallowing the economics.
By
By
By
By
