TL;DR
30-second read- Anthropic purchased Caterpillar gas turbines capable of producing 1.7 gigawatts for $3.4 billion.
- Long waits for grid connections and turbine backlogs are driving AI labs to build on-site power plants.
- Watch how exposure to volatile fuel prices affects Anthropic's cost structure leading into its IPO.
- My take: “From my perspective, heavy equipment manufacturers like Caterpillar are the undeniable winners in this dynamic.” Read my full take
What happened
Anthropic has agreed to spend $3.4 billion on natural gas turbines to supply its growing computational needs. The company is taking over a two-year-old order for Caterpillar generators initially placed by equipment supplier OnePWR. The hardware can generate up to 1.7 gigawatts of electricity, representing the largest known purchase of power gear by a standalone artificial intelligence lab.
The purchase bypasses Anthropic's traditional operating model of renting third-party computing capacity. In an S-1 filing from May, the startup revealed it pays xAI $1.25 billion each month through May 2029 for computing power. OnePWR was unable to arrange the necessary financing to finalize its original turbine deal, prompting Anthropic to step in and buy the equipment directly from Caterpillar.
Tight energy markets pushed Anthropic toward direct hardware ownership before its planned public debut. Waiting times to connect newly built facilities to regional electric grids can extend into the next decade. Similarly, delivery backlogs for commercial natural gas turbines now reach into the early 2030s, leaving firms few options to secure timely electricity.
Why it matters
Anthropic is following tech giants like Google, Amazon, Microsoft, and Meta, which are turning to on-site natural gas installations instead of relying solely on public utilities. The rapid expansion of artificial intelligence clusters has revitalized domestic natural gas demand. By 2035, American data centers are projected to burn more natural gas than Germany and Japan use together.
Investing directly in power generation shields AI developers from infrastructure bottlenecks, but it introduces major commodity risks. Hyperscalers face potential fuel volatility, with some analyst forecasts anticipating regional natural gas prices could more than triple before 2030. These ongoing operational expenses could alter the margins of tech firms running captive power assets.
Grid delays force direct energy generation
Lengthy connection backlogs are forcing software labs to become power plant developers to keep data centers running.
Exposure to natural gas price swings
Running dedicated gas plants leaves AI developers vulnerable to projected fuel cost increases by 2030.
- Gigawatt (GW)
- A unit of electric power equal to one billion watts, commonly used to describe large-scale power plant capacity.
- Hyperscalers
- Large cloud providers and tech firms that build and run massive data centers at scale.
My take
From my perspective, heavy equipment manufacturers like Caterpillar are the undeniable winners in this dynamic.
They are effectively capitalizing on AI panic and severe infrastructure backlogs to lock in massive equipment deals. Meanwhile, tech companies like Anthropic are taking on all the long-term fuel and operational risks by running captive power plants. It leaves software labs vulnerable to volatile commodity markets while hardware makers walk away with guaranteed windfalls.
Aviv Hagag· Editor, Hey Future
Companies that may be affected
Not investment advice. Companies are listed for context only. Hey Future summarizes public reporting about technology; nothing here is a recommendation to buy, sell or hold any security.
Sources & credits
Every fact above comes from these sources.
- 01TechCrunch · Tim De Chant · Nov 9, 2026Anthropic joins AI’s natural gas frenzy, buys $3.4B worth of turbinestechcrunch.com
- 02Azeem Azhar (Exponential View) · Azeem Azhar · Nov 9, 2026📈 AI-exposed workers wait a month longer for a jobexponentialview.co
Anthropic logo: a trademark of Anthropic, shown for identification only.
xAI logo: a trademark of xAI, shown for identification only.
How this was written: AI-assisted summary, checked and edited by Aviv Hagag. “My take” was written by Aviv Hagag.
How we credit sourcesHey Future channel
Want the next stories straight to your phone?
New posts, with sources and the editor’s take.



