NVIDIA will invest up to $2.1bn in IREN as part of a deal that pairs the chipmaker’s reference architecture with the data-centre operator’s 5-gigawatt infrastructure pipeline. The two companies announced the partnership on Thursday.
The structure is unusual. Rather than a straight equity round, IREN has issued Nvidia a five-year warrant for up to 30 million shares at an exercise price of $70. The stock closed regular trading at $56.85, then rose roughly 9% in extended trading after the announcement.
If Nvidia exercises in full, it would put the ticket size at $2.1bn at the strike, with cash outflow timed at the chipmaker’s discretion rather than the deal-close date.
Operationally, the Sweetwater campus in Texas, currently sized at 2GW of planned capacity, comes first. IREN signed a $9.7bn cloud agreement with Microsoft last year for capacity at the same site; the Nvidia tie-up extends the same playbook with the chipmaker as both partner and shareholder.
What Nvidia is buying
The deal slots IREN into a category that has become a familiar Nvidia portfolio shape over the past eighteen months. The chipmaker has taken positions in CoreWeave, Nebius and several other so-called “neoclouds”, the firms that buy Nvidia GPUs at scale and rent them back to hyperscalers and frontier-model builders.
Nebius buying Eigen AI to maximise tokens per GPU was one of the most-watched plays in the same arc. The thesis is straightforward: AI demand outruns hyperscaler capacity faster than the hyperscalers can build, and Nvidia benefits twice if the gap is filled by neoclouds running its silicon.
The numbers behind that thesis have grown larger each quarter. The four largest US tech companies have collectively guided to more than $700bn in 2026 capex, much of it earmarked for AI infrastructure.
Even at that pace, model providers and emerging enterprise users keep landing on waitlists for capacity. IREN at 5GW is a meaningful slice of the answer; the larger Texas grid build-out reframes that as a multi-year commitment.
IREN started life as Iris Energy, a Bitcoin miner with grid-tied infrastructure spread across Texas and British Columbia. It pivoted to AI compute in 2024, dropping the Iris Energy name in favour of the IREN ticker and the cloud-centric brand.
The Microsoft deal in 2025 confirmed the pivot at scale; this Nvidia announcement fixes the supply side.
The pivot from crypto to AI cloud is increasingly common. Hut 8, Applied Digital and Core Scientific have all redirected at least part of their capacity.
What separates IREN at this stage is its portfolio of grid interconnects in regions with cheap, abundant power, the constraint that has become more binding than capital for new AI builds.
Sweetwater is the demonstration site. Texas regulators have been faster to approve large interconnects than other states, and the campus sits adjacent to renewable generation that IREN can lock in on long-term contracts.
A 2GW campus is comparable in scale to the largest hyperscaler builds in the same region.
The neocloud structure
Calling IREN a neocloud is shorthand for a particular business model: build dedicated capacity, sign long-term contracts with one or two anchor customers (here Microsoft), and finance the GPU acquisition through a mixture of vendor financing, equipment leasing, and equity.
Meta’s $21bn add-on with CoreWeave is the largest example of an anchor-tenant deal of this shape; CoreWeave’s multi-year Anthropic agreement is another. The pattern is now well enough established that public investors price these companies on contracted capacity rather than on traditional cloud revenue multiples.
NVIDIA’s role in those structures has become more direct over time. Rather than just selling GPUs, the company increasingly takes warrants or equity in the buyers, both to share in the operating uplift and to ensure that capacity dedicated to its silicon stays priced and serviced as Nvidia would prefer.
The IREN deal extends that posture into a public-company partnership rather than private financing.
The arithmetic that has to work
5GW is a striking number, but the binding constraints are timing rather than capital. Substation interconnects, transformer lead times, GPU shipment schedules, and renewable PPAs all have multi-year cycles that don’t compress.
The Sweetwater campus, being closest to operational readiness, is where most of the early Nvidia-IREN co-investment will land.
Beyond that, the partnership functions as a framework rather than a build plan; future deployments are pitched as conditional on demand and on the regulatory and supply environment.
The pattern echoes Nscale’s Portugal expansion alongside Microsoft, where the build-out timeline trails the headline capacity number by years.
The risk for IREN is execution: turning grid rights and announcements into operational gigawatts at the price points that anchor customers will pay. The risk for Nvidia is concentration.
The chipmaker is now equity-exposed to multiple customers in the same arc; if AI demand normalises or shifts toward in-house silicon at hyperscalers, those positions revalue.
For now, the demand signal is in the other direction, and Nvidia is taking the bet.
Get the TNW newsletter
Get the most important tech news in your inbox each week.