Neocloud Brief · neocloud stocks · what is a neocloud · neocloud ETF

IREN — from Bitcoin miner to Microsoft’s neocloud

A miner that bet the company on converting its cheap-power sites into AI cloud capacity — anchored by Microsoft, with more than $4B of ARR already contracted and a $25–30B capex plan to fund.

Ticker IREN · Nasdaq Price $45.99 Mkt cap ~$16.4B 52-wk $28.93–76.87 Snapshot Sep 18, 2026

01The 30-second read

Fast read

IREN’s pivot is working faster than the financials suggest: AI cloud revenue hit $70.5M in the June quarter (+110% sequential), run-rate crossed $1B after Microsoft accepted Horizon 1, and more than $4B of ARR is already contracted for 2026 capacity. The catch is scale and funding: FY27 capex of $25–30B against ~$14B of funding identified and a residual gap of ~$8B.

FQ4 booked a ~$684M loss — mostly noncash mining impairments. Every bull case runs through one question: can IREN convert contracted ARR into operational ARR before the funding treadmill tightens?

$70.5MFQ4 AI cloud revenue · +110% QoQ
>$4Bcontracted ARR, 2026 capacity
$25–30BFY27 capex plan
~$8Bresidual funding gap

02What IREN actually is

IREN (formerly Iris Energy) built massive, low-cost power sites for Bitcoin mining, then realized those same assets — land, substations, cooling, cheap energy — are exactly what AI compute needs. Today it converts that infrastructure into GPU cloud capacity, anchored by a multi-year relationship with Microsoft (Horizon 1 accepted; Horizons 2–4 in progress), a new multi-year contract with a leading frontier AI lab, and a ~$700M-ARR Nvidia cloud contract ramping in 2027.

The transition shows in the numbers: from 3 MW of self-built AI cloud two years ago to 480 MW and ~150,000 GPUs targeted by year-end 2026, and 1,210 MW by end-2027. Mining is expected to wind down by decade-end — the power and land were always the moat; mining was the bridge.

03The bull case

1 · The ARR ladder is already built

  • ~$500M at end of FQ4 → $1B operating after MS accepted Horizon 1 → >$4B contracted → >$4B ARR targeted by the December quarter
  • Each step is a signed contract, not a hope

2 · Microsoft is a real anchor

  • A $3.65B investment-grade GPU financing package funds ~96% of the $5.81B GPU capex program tied to the Microsoft contract
  • Investment-grade financing with this counterparty is a structural vote of confidence

3 · Funding options keep multiplying

  • ~$6.5B of GPU financing secured in the last three months (incl. $2.8B for non-IG deployments at 9%)
  • Data-center portfolio kept unencumbered — a large asset base available for the ~$8B gap

4 · The Street sees the ramp

  • Cited targets: JPM $65 (OW), H.C. Wainwright $90 (Buy), B. Riley $83, Compass Point $105
  • Roughly +41–128% upside from $45.99, on the December-timeline Horizons landing

04The bear case

1 · The financials are still ugly

  • FQ4 revenue $137.2M, down 5.2% sequentially as mining wound down
  • FQ4 net loss ~$684M (mostly noncash impairments); FY26 loss ~$703M

2 · Capex dwarfs revenue

  • FY27 capex $25–30B against a ~$1B revenue run-rate
  • ~$14B funded; ~$8B must still be raised — many quarters of flawless capital-markets access

3 · Execution is the whole game

  • Horizons 2–4 landing at Microsoft on the December timeline is the difference between “>$4B ARR by December” and a credibility break
  • The Q4 report carried execution concerns; any slip hits cash flow and financing terms

4 · The mining exit is a one-way door

  • Mining revenue and margin are deliberately retired; if AI cloud ramps slower than contracted, there’s no cushion to fall back on

05The numbers

MetricValueNote
FQ4 (Jun) revenue$137.2M−5.2% QoQ; mining wind-down
FQ4 AI cloud revenue$70.5M+110% sequential
FQ4 net loss~$684MMostly noncash impairments
FY26 full-year loss~$703MEPS deeply negative
Operating ARR today~$1BAfter MS accepted Horizon 1
Contracted ARR (2026)>$4BTarget >$4B by Dec quarter
FY27 capex plan$25–30BMS delivery + 2027–28 buildout
Funding identified~$14BCash, GPU financing, prepayments
Residual gap~$8BTo be raised (stated)
GPU financing (3 mo.)~$6.5BIncl. $3.65B IG for MS deal
Capacity targets480 MW2026; ~150K GPUs; 1,210 MW by 2027
Cited targets$65–105JPM / Wainwright / B. Riley / CP
Implied upside~+41–128%vs. $45.99

06What could change the story

Bull triggers

  • Horizons 2–4 accepted at Microsoft on the December-quarter timeline — the single biggest catalyst
  • Contracted ARR flips to operational ARR early 2027, as guided
  • The ~$700M-ARR Nvidia cloud contract ramps on schedule in 2027
  • The ~$8B gap closes at reasonable rates via the unencumbered portfolio

Bear triggers

  • Microsoft delivery slips past the December timeline — credibility and funding both break
  • The $8B gap requires punitive-rate financing or heavy dilution
  • AI cloud ramps slower than contracted ARR while mining revenue is already gone
  • A big customer (frontier lab) delays or renegotiates

07What’s next

NOV
2026

Q1 FY2027 earnings (verify date). December-quarter ARR target and Horizon delivery updates are the headline.

DEC
2026

Microsoft Horizons 2–4. Acceptance on the December-quarter timeline converts the ARR story into cash flow.

2027

Capacity + NVIDIA ramp. 1,210 MW targeted; the ~$700M Nvidia ARR contract ramps; mining winds down.

08FAQ

Is IREN profitable?

No. FQ4 2026 net loss ~$684M (mostly noncash mining impairments); FY26 ~$703M. AI cloud scales fast, but the whole company is still loss-making with heavy capex.

What does IREN actually do?

Converts former Bitcoin-mining power sites into AI cloud capacity — land, substations, cooling, cheap power — anchored by Microsoft, with mining wound down by decade-end.

Why is the stock down if the pivot is working?

June-quarter revenue miss and a large loss, plus fear about the ~$8B funding gap against $25–30B of FY27 capex. The market prices execution and financing risk against a very visible ARR ramp.

Is IREN a buy?

Cited bank targets run $65–105 (+41–128% upside), but this is a high-risk execution story, not a consensus name. This page is research, not a recommendation.

Sources & method