🧭 Decision Radar
Relevance for Algeria
Low
▾
Infrastructure Ready?
No
▾
Skills Available?
No
▾
Action Timeline
24+ months
▾
Algérie Télécom, ARPT, Ministry of Post and Telecommunications, Algeria Venture (digital infrastructure investment)
Decision Type
Educational
▾
Quick Take: Algerian digital infrastructure planners should note the Iren-Microsoft model — an independent operator delivering phased, liquid-cooled AI capacity faster than a hyperscaler could build it alone — as an illustration of how AI compute capacity deals are increasingly structured globally, even though Algeria is years away from having the power and cooling infrastructure to compete for this kind of contract.
From Bitcoin Mining Infrastructure to Hyperscaler AI Capacity
Iren’s Childress, Texas campus delivered its first operational milestone under the company’s landmark agreement with Microsoft: Horizon 1, a 50MW phase of dedicated AI cloud capacity, has been completed and formally accepted by Microsoft. The handover marks the first concrete delivery under the five-year, $9.7 billion cloud services agreement the two companies announced in November 2025 — a deal that saw Iren, a company with roots in Bitcoin mining infrastructure, pivot decisively toward becoming a dedicated AI cloud capacity provider for one of the largest hyperscalers in the world.
Horizon 1 is the first of four 50MW (IT load) direct-to-chip liquid-cooled buildings scheduled for delivery during 2026, which together will create a combined 200MW liquid-cooled AI infrastructure footprint dedicated entirely to the Microsoft contract. The facilities use direct-to-chip liquid cooling to support high-density NVIDIA GB300 NVL72 systems, a deployment that earned Iren NVIDIA Exemplar Cloud status — the dense, power-hungry hardware configurations that have become standard for frontier AI training and inference workloads in 2026.
Why Independent Operators Are Winning Hyperscaler Contracts
The Iren-Microsoft arrangement is one example of a broader pattern reshaping the data center industry in 2026: hyperscalers like Microsoft are increasingly contracting directly with independent data center operators to secure AI compute capacity faster than they could build and commission it themselves. Rather than waiting years for a fully self-developed campus, Microsoft is effectively renting speed — Iren’s existing land, power interconnection, and construction pipeline let capacity come online in a matter of months per phase rather than the multi-year timelines typical of greenfield hyperscaler-built facilities.
For operators like Iren, the shift represents a fundamental change in revenue model: a company that built its infrastructure and expertise around power-hungry, always-on compute for cryptocurrency mining has repositioned that same core competency — cheap, reliable, high-density power delivery — toward AI workloads, where demand and contract values are currently far higher.
1. Expect more crypto-to-AI infrastructure pivots as capacity crunches continue
Iren’s transition from Bitcoin mining infrastructure to hyperscaler AI capacity provider is a template other crypto-infrastructure operators with existing power interconnection and land are likely to follow, given the scale of unmet AI compute demand.
2. Track phased delivery as the new normal for hyperscaler capacity deals
The 50MW-at-a-time delivery structure — four phases building to 200MW over 2026 — reflects how hyperscalers are now structuring capacity contracts to get usable compute online incrementally rather than waiting for a single large facility to be fully built.
3. Watch liquid cooling and GB300-class hardware become the baseline, not the exception
The use of direct-to-chip liquid cooling to support NVIDIA GB300 NVL72 systems at Childress reflects that dense, liquid-cooled infrastructure is now the default requirement for any operator seeking hyperscaler AI contracts, not a differentiator reserved for cutting-edge facilities.
Advertisement
What Comes Next for the Childress Campus
With three more 50MW phases scheduled through the remainder of 2026, the Childress campus is on a trajectory to reach its full 200MW commitment to Microsoft within the year, assuming construction and power interconnection stay on schedule — part of a broader company target of 480MW of gross AI Cloud capacity by the end of 2026, scaling toward 1.2GW in 2027. The deal’s structure — a fixed five-year, $9.7 billion commitment tied to phased physical delivery — gives both sides clear milestones to track, and the successful Horizon 1 handover suggests the broader timeline remains realistic. For the wider data center industry, Iren’s experience is likely to encourage other independent operators to pursue similar direct hyperscaler partnerships rather than competing purely on wholesale colocation pricing.
Frequently Asked Questions
What did Iren deliver to Microsoft, and when?
Iren delivered “Horizon 1,” the first 50MW phase of dedicated AI cloud capacity at its Childress, Texas campus, which Microsoft has formally accepted as the first operational milestone under their five-year, $9.7 billion cloud services agreement announced in November 2025.
How much total AI capacity will Iren deliver to Microsoft?
Horizon 1 is the first of four planned 50MW IT-load deployments scheduled for delivery during 2026, building toward a combined 200MW liquid-cooled AI infrastructure footprint dedicated to the Microsoft contract.
What hardware and cooling technology does the Childress campus use?
The facilities use direct-to-chip liquid cooling to support high-density NVIDIA GB300 NVL72 systems, the dense hardware configuration now standard for frontier AI training and inference workloads.














