What happened
IREN Ltd has signed a five-year managed GPU cloud services agreement with Nvidia valued at $3.4 billion, per reporting by CryptoBriefing, AInvest, and Seeking Alpha. Reporting by CryptoBriefing and boerse-global.de describes a separate five-year warrant that gives Nvidia the right to acquire up to 30 million IREN shares at $70 per share, an option that would raise $2.1 billion if exercised in full. CryptoBriefing reports IREN shares rose as much as 27% in after-hours trading following the announcement.
Technical details
Seeking Alpha and boerse-global.de report the agreement covers Nvidia internal AI and research workloads and includes deploying 60 MW of air-cooled Blackwell platform systems at IREN's Childress, Texas data center, with services expected to begin in early 2027. CryptoBriefing and additional market coverage note IREN has publicly stated capacity targets and recent build milestones, and reporting places the company's secured or targeted capacity at roughly 5 gigawatts across North America and Europe (CryptoBriefing; boerse-global.de).
Reported financials and near-term performance
Boerse-global.de (ad-hoc-news) reports IREN posted fiscal third-quarter revenue of $144.8 million, below a cited consensus of $220.2 million, and a quarterly loss of $247.8 million; that coverage also attributes $111.2 million of the quarter's revenue to Bitcoin mining. Reporting by CryptoBriefing highlights that IREN's contracted revenue with Microsoft and Nvidia now exceeds $13 billion, a figure presented against the company's market capitalization in that coverage.
Editorial analysis - technical context
Companies repurposing crypto-mining sites for AI infrastructure commonly leverage existing grid connections, on-site substations, and long-term power contracts to reduce the incremental cost of operating large GPU farms. Industry-pattern observations indicate electricity and grid access are frequent bottlenecks for large-scale GPU deployments, even when chip supply is abundant. For practitioners, this means teams building or procuring large-scale compute should weight site-level power economics and grid interconnection timelines as heavily as rack-level cooling or supplier roadmaps.
Context and significance
Industry reporting frames the Nvidia-IREN arrangement as one of the larger infrastructure deals tying a hyperscaler-class GPU supplier directly to a third-party operator through both services contracts and equity-linked instruments (CryptoBriefing; boerse-global.de). For infrastructure operators and cloud procurement teams, this deal underscores a trend where chip vendors and large AI end users secure fixed-capacity arrangements with specialist hosts to guarantee colocated GPU availability. Industry observers frequently note that such arrangements change how capacity risk and capital commitment are allocated across the supply chain.
What to watch
For practitioners:
- •track IREN's execution on grid milestones and the timing of the Childress Blackwell deployment reported to start servicing Nvidia workloads in early 2027 (Seeking Alpha; boerse-global.de)
- •monitor whether Nvidia exercises the warrant over the five-year term, since full exercise would provide $2.1 billion in capital per CryptoBriefing and boerse-global.de reporting
- •watch IREN's next quarterly results for shifts in revenue mix away from Bitcoin mining toward managed GPU services, given the fiscal Q3 figures reported by boerse-global.de
Final note
Key Points
- 1IREN signed a five-year managed GPU services deal with Nvidia worth $3.4 billion, accelerating its shift from mining to AI infrastructure.
- 2The agreement includes a five-year warrant for 30 million shares at $70, potentially supplying $2.1 billion in capital if exercised.
- 3Industry pattern: grid access and power economics remain the primary constraint on large-scale GPU deployments, not GPU supply alone.
Scoring Rationale
Large, multi-billion-dollar services and equity-linked arrangements between a leading GPU supplier and an infrastructure operator materially affect capacity planning and capital flows for AI compute. The story is notable for practitioners managing large deployments and procurement, but it is not a frontier-model or platform release.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
