Realtime AI News
Anthropic Buys $3.4B of Natural Gas Turbines as AI Power Race Widens
Anthropic has bought roughly $3.4 billion worth of natural gas turbines, TechCrunch reported on November 9, joining a widening scramble among AI companies for dedicated power generation. The report also warns the deal could give the buyer remorse, since it locks fuel-price and emissions exposure into a long-lived equipment purchase.
Anthropic has bought roughly $3.4 billion worth of natural gas turbines, TechCrunch reported on November 9, making it the latest AI company to reach for dedicated generation as datacenter demand outruns what grids can readily supply.
The backdrop is a power bottleneck. AI buildouts are now constrained less by chips than by access to reliable electricity, and gas turbines have become the short-term default because they can be ordered, sited and brought online faster than nuclear plants or large renewable projects.
TechCrunch's own framing is skeptical. The deal, it writes, could give the buyer remorse. That phrase points at the risks built into fossil generation: fuel-price exposure over the life of the equipment, a mismatch between a multi-decade asset and a hardware refresh cycle measured in a couple of years, and tension with corporate emissions commitments.
The scale of the spending is the signal. A $3.4 billion equipment order is not a pilot unit or a backup generator for a single site; it implies a generation plan sized for sustained, high-density load.
It also changes how AI labs relate to infrastructure. Renting capacity from cloud providers keeps power off the balance sheet; buying turbines puts it on. That trade is worth making only if the compute demand behind it is expected to persist long enough to amortize the equipment.
For the wider industry, the move extends a pattern already visible this year: model developers are pushing into energy, land and construction, the unglamorous layers that determine whether a training run can happen at all.
The report does not say where the turbines will be sited, when they will be delivered, or how the generation squares with Anthropic's emissions targets. Those details, and whether other labs place similar orders, are the things to watch.
Why it matters
AI companies moving from renting compute to owning generation shifts the infrastructure race toward electricity and pushes fuel-price and carbon risk directly onto AI balance sheets. If supply constraints persist, self-generation could move from exception to default.
Nearby Updates
All11/10, 01:05
Pelosi's successor in Congress wants to slow AI down and give it a kill switch
After winning Nancy Pelosi's San Francisco seat, California state senator Scott Wiener told MIT Technology Review that the US needs to slow AI progress, citing a METR report as evidence of recursive self-improvement. He backs a strong federal regulator, auditing and kill-switch requirements, and opposes broad federal preemption of state AI laws.
11/10, 00:43
Google Adds TPU 8tx and 8i Max to Its Gen-8 Lineup, Aimed at Agent Workloads
Google has expanded its eighth-generation TPU lineup with two new parts, the TPU 8tx and TPU 8i Max, according to a November 9 report from NDTV Profit that frames both chips as targeting AI agent workloads. The report gives model names and a workload target but no specifications, pricing or availability.
11/10, 00:12
Medium's New TK App Puts AI Assistance Inside the Writing Process
Medium has launched a new writing app called TK that brings AI assistance directly into the writing process, TechCrunch reported on November 9. The report notes that Medium is embracing AI-assisted writing even as controversy over the technology's role in publishing keeps growing.
11/10, 00:03
Northern Trust and Microsoft Deploy AI Agent, Cutting Client Onboarding Time by 83%
Northern Trust has deployed an AI agent built with Microsoft to handle client onboarding, cutting the time the process takes by 83%, according to StreetInsider. The project is a concrete case of a major custodian bank putting agentic AI directly into middle- and back-office operations with a quantified result.