AI companies can spend billions on GPUs, but those chips aren’t much use without enough electricity to keep them running.
Anthropic has reportedly reached a $3.4 billion agreement to secure 1.7 gigawatts of gas-fired generator capacity through equipment orders previously placed by OnePWR Solutions, according to the Financial Times. The arrangement reflects how access to power equipment is becoming a major constraint on AI data center expansion.
For systems integrators, the deal highlights a growing challenge: AI hardware can be ready for installation long before the power infrastructure needed to run it.
Earlier Caterpillar orders provide access to power equipment
Under the agreement reached in late September, Anthropic secured commitments for 1.7 gigawatts of heavy-duty generators from OnePWR Solutions, a San Antonio, Texas-based supplier. The transaction is the largest known single purchase of power-generation equipment by an AI laboratory.
OnePWR had ordered the Caterpillar generators approximately two years earlier but needed financing to complete the purchase. The AI developer initially planned to become a power customer before deciding to take over those equipment orders instead.
Engineering and construction capabilities needed to install the generators were included in the arrangement. Anthropic did not acquire OnePWR itself, and no employees transferred to the company.
Earlier orders give the AI lab access to engines expected to become available in the near term and over the next two years, following a period of rising equipment prices. Neither party commented on the transaction when contacted.
AI companies expand direct control over energy infrastructure
Manufacturing lead times for gas turbines and generators now extend years into the future as AI data center electricity demand continues to rise. Major technology companies are increasingly securing power equipment themselves after traditionally depending on utilities and specialist energy providers.
Other AI and technology companies have pursued substantial energy investments:
- xAI: The SpaceX-owned AI company reportedly purchased $2 billion in mobile gas turbines earlier in 2026 to support its data center operations.
- Alphabet: Google’s parent company announced a deal to acquire energy developer Intersect in late 2025, expanding its involvement in power infrastructure development.
Anthropic’s purchase comes amid an extensive infrastructure expansion. The company has reported nearly $900 billion in computing commitments through 2045, having relied heavily on partners to obtain capacity for its growing operations.
Commissioning schedules determine when AI hardware can go live
Data center systems integrators should distinguish equipment procurement from operational power capacity when planning customer AI infrastructure projects. Securing generators does not guarantee that a facility can support high-density GPU deployments, even when installation capabilities are included in the procurement agreement.
Before confirming hardware delivery dates, integrators should obtain the facility operator’s commissioning schedule and verify the electrical capacity allocated to the project. Designs for rack-level power distribution must align with those limits, and liquid-cooling systems need sufficient capacity for the selected server configurations.
Project contracts should establish responsibility for electrical acceptance testing, infrastructure handover, and delays that could affect installation schedules. Without those safeguards, expensive AI hardware could arrive before the facility is ready to operate it.
Anthropic’s reported agreement illustrates a broader challenge for the AI infrastructure market: securing power equipment and delivering usable computing capacity are two different milestones. For channel partners, coordinating those timelines is becoming just as important as sourcing the hardware itself.
More AI news: Investors are reportedly reconsidering Anthropic’s valuation, with some favoring a range of $1.1 trillion to $1.3 trillion ahead of its planned IPO.


