Alphabet Commits $1.8 Billion to 564MW Wyoming Power Facility for AI Data Centers
The AI compute build-out continues to focus on securing substantial power capacity and enhancing networking infrastructure to meet escalating demands.
The story
As the demand for AI compute capacity outpaces traditional grid capabilities, Alphabet has moved to directly finance a significant power infrastructure project. The company announced a definitive agreement to invest $1.8 billion in a natural gas and microgrid expansion with Black Hills Corp in Cheyenne, Wyoming.
This initiative aims to establish a 564-megawatt power facility by approximately 2029, alongside managing a 2.1-gigawatt microgrid composed of third-party contracted resources. This direct investment underscores a growing trend among hyperscalers to bypass slower traditional grid procurement processes, ensuring dedicated and scalable energy for their AI data centers. The move highlights how power availability has become a primary constraint for AI expansion, pushing technology firms to integrate energy solutions directly into their infrastructure strategies.
Silicon
| Chip | Maker | What | For Whom |
|---|---|---|---|
| Tomahawk 6, Tomahawk Ultra, Jericho 4 (Ethernet switches) | Broadcom | New generation Ethernet switches for scaling AI infrastructure. | AI labs, hyperscalers, and cloud providers for next-generation AI clusters. |
| Thor Ultra 800G, Thor 2 400G (AI Ethernet NICs) | Broadcom | High-speed AI Ethernet Network Interface Cards. | AI labs, hyperscalers, and cloud providers for AI cluster networking. |
| TH6-Davisson Co-Packaged Optics (CPO) | Broadcom | Third-generation integrated optics to meet exponential bandwidth demands. | AI labs, hyperscalers, and cloud providers for frontier AI models. |
| Token Fabric (networking chips) | Upscale AI | Hardware and software platform to connect AI processors from different suppliers across a data center. | Customers building data centers with multi-vendor AI chips, including neoclouds and large cloud providers. |
The build-out
| Project | Who | Scale | Where |
|---|---|---|---|
| US Data Center Portfolio Expansion | Gorilla Technology Group | Initial 6 MW operating facility with pathway to 36 MW, targeting over 100 MW total | United States (first acquisition planned) |
| Data Center Land Purchases | Hyperscalers, neocloud providers, energy-first developers | $6 billion in H1 2026, up 79% year-over-year | United States |
Supply & policy signals
TSMC's Q3 2026 sales up 51% year-over-year, Samsung's Q3 operating profit up 8.8 times year-over-year, with some signs of easing demand.
Implication: While AI hardware demand remains very strong, the pace of growth might be moderating slightly from previous peaks, though overall revenue is still substantially higher than last year.
Rising rack-level power requirements pushing industry towards 800VDC power architecture for AI data centers.
Implication: Power delivery is becoming a system-level design issue, creating new opportunities and challenges for power semiconductor technologies like GaN and SiC across the AI infrastructure supply chain.
Memory lead times up to 30-40 weeks, and network switch availability constrained.
Implication: AI infrastructure bottlenecks extend beyond compute scarcity to include critical components like memory and networking equipment, impacting the ability to connect distributed AI workloads.
What we'll be watching
- Broadcom's presentations at the Open Compute Project (OCP) Global Summit from October 12–15 in San Jose, detailing new AI networking innovations.
- Nvidia RTX Spark pre-orders opening on October 16, and the subsequent shipping of compact desktops in November.
- Gorilla Technology Group's targeted completion of its first US data center acquisition by the end of November 2026.
Reporting + analyst voices: grounded via Google Search at publish time.