Everyone seems to want exposure to the artificial intelligence (AI) theme, but the real constraint isn't ideas or algorithms -- it's power and compute capacity. While software companies promise AI features, and megacap tech giants hoard graphics processing units (GPUs), a different set of companies are building the physical infrastructure that determines how fast AI actually scales.
Two stocks deliver tangible infrastructure rather than vaporware: a power-rich builder that just landed a $9.7 billion hyperscaler contract, and a pure-play GPU cloud with $30.1 billion in its backlog. The setup rewards investors willing to track energization timelines and megawatt additions instead of waiting for the next chatbot upgrade.
Image source: Getty Images.
The power-rich builder landing hyperscaler validation
Iris Energy (IREN +11.52%) started as a Bitcoin miner but is pivoting hard into AI compute with an unusual advantage -- deep power access. The company has secured 2.75 gigawatts of grid connections and power rights in West Texas, including a new 600-megawatt tie-in. To put that in perspective, 2.75 gigawatts could power roughly 2 million homes, giving Iris rare infrastructure leverage as electricity becomes the defining constraint for data centers.
On Nov. 3, 2025, Iris announced a five-year, $9.7 billion agreement with Microsoft for GPU cloud computing services -- the company's first named hyperscaler customer and validation of its pivot strategy. Microsoft will receive access to Nvidia GB300 GPUs deployed and operated by Iris, with the contract including a 20% prepayment that de-risks execution.
The agreement provides phased deployment of GPU capacity throughout 2026 at Iris's Childress, Texas campus, which has 750 megawatts of total capacity, with four new liquid-cooled data centers supporting 200 megawatts of critical computing load.

NASDAQ: IREN
Key Data Points
Iris also reached an agreement with Dell Technologies covering the purchase of GPUs and additional infrastructure -- servers, cables, software, and licenses -- worth about $5.8 billion. The company intends to use existing cash, Microsoft's prepayments, operating cash flow, and new financing to fund these investments.
Execution risk remains -- campus buildouts and GPU delivery schedules can slip -- and residual crypto exposure creates volatility. However, landing Microsoft as an anchor tenant transforms Iris from a speculative pivot story to validated infrastructure operator, with contracted revenue extending through 2030.
The pure-play AI cloud at scale
CoreWeave (CRWV 5.45%) operates a fast-growing GPU cloud purpose-built for AI training and inference. Since its March 2025 initial public offering (IPO), CoreWeave has posted triple-digit revenue growth and disclosed a $30.1 billion revenue backlog, reflecting multiyear compute commitments from hyperscalers and enterprises. Nvidia holds a roughly 7% strategic stake, signaling confidence in CoreWeave's execution and access to leading-edge silicon.

NASDAQ: CRWV
Key Data Points
Capacity additions, rollouts of Nvidia's newest Blackwell and Grace Blackwell chip architectures, and enterprise AI migration from proof-of-concept to production workloads represent clear catalysts. A fresh $6.3 billion Nvidia agreement backstops unsold capacity through 2032, providing additional revenue visibility.
The cloud model here -- offering specialized GPU infrastructure without legacy enterprise baggage -- captures demand that traditional clouds struggle to serve profitably. Companies need raw computing horsepower for AI work, and CoreWeave delivers it at scale without forcing them into long-term enterprise-software agreements.
The contracted revenue advantage
These two AI stocks won't move in lockstep. Iris Energy carries execution risk on campus buildouts and residual Bitcoin exposure, which creates volatility, while CoreWeave faces capital-intensity pressures and pricing-compression risk as hyperscalers bundle AI services. But the setup rewards focusing on contracted revenue, rather than speculative AI applications.
Iris Energy's Microsoft deal transforms a crypto-miner pivot into a validated infrastructure operator with $9.7 billion in locked-in commitments through 2030. CoreWeave's $30.1 billion backlog demonstrates that purpose-built GPU clouds can capture enterprise spending without competing directly against Amazon, Alphabet, or Microsoft.
Power capacity matters more than algorithms at this stage, and long-term contracts matter more than quarterly revenue beats. While software companies debate which AI features customers want, these two stocks provide direct exposure to the physical buildout that determines how fast AI scales.