IREN (IREN) has delivered the first of four planned AI cloud deployments to Microsoft (MSFT) under their $9.7 billion contract.
The milestone converts a headline-grabbing contract figure into operating infrastructure, a distinction that matters when hyperscaler commitments often precede physical delivery by months.
IREN is targeting 480 megawatts of AI cloud capacity before the end of this year, scaling to 1.2 gigawatts by 2027.
Key Takeaways
IREN completed the first of four planned AI cloud deployments to Microsoft on August 17 The Microsoft contract with IREN is valued at $9.7 billion for AI cloud infrastructure IREN is targeting 480 megawatts of AI cloud capacity before the end of this year Securing grid connection agreements for hundreds of megawatts can take 18 to 36 months in regulated utility markets
IREN AI Cloud Deal Reshapes A Former Miner's Identity
IREN entered the public market as a bitcoin miner, using cheap power and custom cooling to extract cryptocurrency at scale. The business logic was straightforward: locate abundant, low-cost electricity, deploy specialized hardware, and sell the resulting Bitcoin (BTC).
That model worked well when mining margins were wide.
As margins compressed and AI compute demand surged, IREN began redirecting the same core competency, managing large power installations, toward a different customer. The first of four deployments was completed on August 17, demonstrating that the pivot has cleared the hardest threshold- physical delivery at hyperscaler specification.
The $9.7 billion Microsoft contract is the result of that strategic reorientation.
Under the structure, IREN builds and operates AI cloud infrastructure that Microsoft then uses for its own compute workloads. The arrangement is essentially a long-duration power and rack lease dressed in AI vocabulary, IREN supplies the physical plant.
Microsoft supplies the demand.
Delivering the first of four deployments is meaningful precisely because it proves the company can execute at hyperscaler specification, not merely sign term sheets.
What 480 Megawatts Actually Means
A megawatt of AI compute capacity is not a fixed unit. The term describes power draw, not processing throughput, a distinction that gets lost in most coverage of these deals.
A modern AI data center running dense GPU clusters might consume 50 to 100 kilowatts per rack.
At that density, 480 megawatts of capacity represents somewhere between 4,800 and 9,600 fully loaded racks, enough to run tens of thousands of high-end accelerators simultaneously. The cost per unit of inference is ultimately a function of how efficiently that power envelope is filled with productive compute.
That is where operational discipline separates infrastructure owners from infrastructure landlords.
Scaling to 1.2 gigawatts by 2027 would put IREN among the larger dedicated AI infrastructure operators globally. For context, a gigawatt of power is roughly the continuous output of a large nuclear reactor.
The capital intensity of reaching that figure is enormous, which is why the Microsoft contract matters structurally: a committed offtaker at that scale transforms project finance from speculative to bankable, and gives IREN's balance sheet a credibility that no press release about a pivot could manufacture on its own.
Also Read: CZ Moves $965K To Giggle Academy As Binance Founder Shifts Focus To Education
The Miner-to-AI-Cloud Conversion Running Across The Sector
IREN is not the only former miner executing this conversion, but it is the one with the largest single contracted commitment currently in physical execution.
The transition reflects a structural shift in how large AI labs and hyperscalers source compute.
Rather than building every facility themselves, they increasingly sign long-term contracts with operators who can secure land, power interconnects, and cooling infrastructure faster than internal teams. The capacity constraint is real: hyperscalers need compute online faster than greenfield construction timelines allow, and miners with operating power agreements are the shortest path to energized racks.
For IREN specifically, the entire thesis rests on the overlap between its legacy mining infrastructure and AI cloud requirements.
Cheap power contracts, dense liquid cooling, physical security, and on-site electrical infrastructure are fixed costs already sunk into the mining business.
Redirecting that base toward a Microsoft workload generates higher and more predictable revenue per megawatt than Bitcoin ever reliably did. The capital base that looked oversized for mining looks appropriately sized for AI cloud at gigawatt scale.
Bitmine's ongoing accumulation of Ethereum (ETH) as a treasury asset, covered separately this week, represents a different strategic response to the same margin pressure.
IREN has chosen infrastructure revenue. Bitmine has chosen digital asset accumulation.
Both strategies reflect the same underlying reality that pure Bitcoin mining economics no longer justify the capital base most publicly listed miners carry.
What The 2027 Gigawatt Target Requires
Reaching 1.2 gigawatts of AI cloud capacity by the end of next year requires IREN to add roughly 720 megawatts beyond its 2026 target in a single year. That is an aggressive construction and commissioning schedule by any measure, and the constraint is not concrete or steel, it is grid interconnect.
Securing grid connection agreements for hundreds of megawatts often takes 18 to 36 months in regulated utility markets. IREN's ability to hit its 2027 figure depends substantially on whether it has already secured those agreements for the remaining three deployment phases. Power interconnect queues are the binding physical reality behind every gigawatt target in this sector, and no amount of contracted revenue accelerates a utility commission's timeline on its own.
The Microsoft relationship does provide real leverage here.
Hyperscaler demand signals carry weight with utility commissions and grid operators, which can accelerate interconnect approvals for contracted facilities. The delivery strengthens IREN's credibility for exactly that negotiation, a company that has executed phase one of a $9.7 billion contract is a materially different counterparty than one still promising it will.
Read Next: Strategy Sells $334M In MSTR Stock As Bitcoin Holdings Stay Frozen
