IREN delays Microsoft AI capacity at Childress to early 2027, plans $15 billion in spending

What IREN reported

IREN, the former bitcoin miner that now rents out GPU capacity for AI, reported results for the first quarter of its 2027 fiscal year on 5 November 2026. Revenue for the quarter to 30 September rose 66% from the previous quarter to $228 million, slightly above the analysts' average estimate of about $221 million compiled by LSEG, according to Reuters. AI cloud revenue nearly tripled to about $200 million, with an 86% segment gross margin, and now makes up almost all of the company's sales. IREN recorded a net loss of $263 million.

On the earnings call, IREN said its operating annualized run rate had roughly doubled to about $1 billion and that its revenue backlog exceeds $20 billion. Management also said previously contracted Nvidia H100 capacity had been re-contracted at roughly double the earlier price, a sign that demand for older GPUs is holding up even as newer systems ship.

IREN, known as Iris Energy until 2024 and listed on Nasdaq, built its business on cheap renewable power for bitcoin mining before turning the same sites and grid connections into GPU data centers. The quarter shows how far that shift has gone: AI cloud is now the core of the company, and mining is a minor line. That also means IREN's results now move with the build schedules and technical demands of a handful of very large AI customers rather than with the bitcoin price.

The Microsoft delay

The number investors focused on was the date. IREN had aimed to pass more than $4 billion in annualized revenue by December 2026. It now expects to get there in January 2027, possibly February. Co-CEO Daniel Roberts said Microsoft asked for changes to the networking design at IREN's Childress, Texas campus to support newer AI models, and that those changes, together with supplier delivery timing, pushed the schedule back. IREN said the target remains fully contracted and funded and that the delay carries no financial penalties.

Childress is the 750 MW campus behind the five-year AI cloud contract IREN signed with Microsoft on 3 November 2025, worth about 9.7 billion dollars including a 20% prepayment, to host Nvidia GB300 GPUs in liquid-cooled data centers rolled out in phases through 2026. Microsoft accounts for a large share of IREN's backlog, so any slip there moves the whole company's revenue curve.

The link to Microsoft's spending shortfall

IREN shares had already fallen sharply before the report after The Information, citing unnamed sources, linked delays at Childress to Microsoft spending about $5 billion less than planned on capital projects in its September quarter. Microsoft's chief financial officer had said on its own call that some third-party capacity deliveries expected in the quarter were delayed, without naming suppliers. How much of that gap traces to IREN has not been stated by either company.

Microsoft has been one of the largest buyers of capacity from so-called neoclouds, the specialist GPU hosts such as IREN, Nebius and Nscale. Its first-quarter results showed Azure growth of 48% and a cloud backlog of 677 billion dollars, so demand is not the issue; delivery is. For the neoclouds, a customer-requested redesign is a reminder that the buyer, not the host, sets the technical bar as GPU generations change.

The 15 billion dollar plan

IREN said it spent about $5 billion on capital projects in the quarter and plans about $15 billion more over the rest of the fiscal year, which ends on 30 June 2027. The company describes the program as fully funded, a term its release defines as covering remaining capex subject to conditions on committed financing facilities. Independent reviews of the results note that roughly $6 billion of the plan still depends on new financing beyond existing cash, GPU financing and customer prepayments. IREN also said on the call that it had reduced planned spending for the year because capital markets have become more selective.

The plan puts IREN among the most aggressive builders outside the hyperscalers, at a time when lenders are watching AI infrastructure debt closely. Other recent deals, such as TeraWulf's Google-backed leases for Anthropic, show the same model: a former miner, a large AI tenant and a third party carrying credit risk. More company results are in AI business.

What to watch next is whether the January date holds, how much of the remaining financing IREN closes and on what terms, and whether Microsoft reports a catch-up in third-party capacity in its December quarter. Not stated in the results: the exact value of revenue shifted out of the December quarter and whether other customers' sites face similar redesign requests.

Questions people ask

Why did IREN delay its Microsoft capacity?

IREN says Microsoft requested a networking design change at its Childress, Texas campus to support newer AI models, and supplier delivery timing added to the delay. The target moved from December 2026 to January or February 2027.

How much revenue did IREN report?

$228 million for the quarter to 30 September 2026, up 66% on the previous quarter, with about $200 million from AI cloud, according to its results release.

Does IREN face penalties for the delay?

IREN said the delay carries no financial penalties and that the target remains fully contracted and funded.

Sources

1. IREN Reports Q1 FY27 Results (IREN, 5 Nov 2026)
2. AI data center firm IREN delays Microsoft AI deliveries, plans $15 billion spending push (Reuters via Investing.com, 5 Nov 2026)
3. IREN Q1 2027 Earnings Call Transcript (Benzinga, 5 Nov 2026)
4. IREN Secures $9.7bn AI Cloud Contract with Microsoft (IREN, 3 Nov 2025)