AI build-outs are running into a growing constraint that investors are increasingly focusing on: power availability. According to a policy paper from Anthropic, training a future frontier AI model could require gigawatts of electricity, and the U.S. AI sector may need 50 gigawatts of power by 2028—an outlook that is pushing hyperscalers to look beyond long, grid-connection timelines. Within the power supply chain, GE Vernova has attracted attention for its ability to deliver both near-term generation and grid solutions, supported by a surge in demand for heavy-duty gas turbines.
Key takeaways
-
Order momentum: GE Vernova reported a fourfold jump in heavy-duty gas equipment orders in the second quarter, booking 52 heavy-duty gas units.
-
Catalyst: Anthropic’s “Build AI in America” highlights projected power needs that industry investors are interpreting as a multi-year buildout opportunity.
-
Implication for capacity: GE Vernova’s focus on heavy-duty and aeroderivative gas turbines is aligned with hyperscalers seeking bridge power while grid expansion catches up.
-
Scale and backlog: The company said its backlog reached $176 billion by the end of the quarter, reinforcing visibility into future revenue opportunities.
-
Execution focus: The shift toward alternative power solutions underscores the importance of delivery speed and permitting navigation for developers and regulators.
What is driving the shift to power solutions
While investors have widely discussed memory chips and other hardware bottlenecks tied to AI demand, the supply chain challenge is increasingly dominated by electricity. According to Anthropic’s policy paper “Build AI in America,” the power requirements of frontier-model training are expected to scale dramatically, with gigawatt-level needs for a single training run in the future. The paper also projects that by 2028, the U.S. AI sector could require 50 gigawatts of electricity.
That scale collides with the realities of power infrastructure. Building additional transmission lines, securing substation approvals, and completing grid interconnections can take years and often face community resistance. The company also pointed out that these delays create pressure to find capacity that can come online faster without shifting the cost burden to broader households.
Why GE Vernova is tied to hyperscalers’ near-term power needs
GE Vernova provides power and grid infrastructure, including grid solutions, energy management systems, wind turbines, and gas turbines. The company also highlights a global footprint, with an installed base spanning more than 100 countries that it says generates about one-quarter of the world’s electricity.
In the current cycle, the emphasis is on gas turbines as hyperscalers seek practical ways to secure power capacity while long-term network upgrades proceed. The article highlights two categories that can serve different timelines:
-
Aeroderivative turbines, which the company says can be shipped, installed, and commissioned in as little as six months, potentially providing “bridge power” during grid buildouts.
-
Heavy-duty gas turbines, which the company positions as efficient baseload power as demand increases over time.
Demand appears to be showing up in orders. In the second quarter, GE Vernova’s heavy-duty gas equipment orders jumped fourfold, and the company booked 52 heavy-duty gas units during the period. The company also said its backlog reached $176 billion by the end of the quarter, a key metric investors typically use to gauge future manufacturing and project execution momentum.
Project relevance: on-site generation and bypassing grid friction
A core theme in the AI power discussion is that hyperscalers want reliability and speed, even if it means building closer to demand centers. The article points to GE Vernova’s involvement in Project Kilby in Texas, where the company is working with Chevron and Microsoft to build a 2.67-GW co-located power facility for Microsoft’s AI data center. The project is described as using GE Vernova gas turbines and electrical infrastructure to deliver power directly to Microsoft without burdening the regional power grid.
For investors, the implication is straightforward: projects that reduce dependence on the regional transmission network can help shorten timelines and mitigate the risk that grid constraints become a gating factor for AI expansion. At the same time, co-located capacity can introduce its own permitting and construction risks, making order growth and backlog visibility particularly relevant.
Management outlook and capacity targets
GE Vernova’s leadership framed the opportunity as a multi-decade growth cycle for electric power infrastructure. During the company’s second-quarter earnings call, CEO Scott Strazik characterized the long-cycle electric power industry as being in the early stages of a multi-decade growth opportunity.
Operationally, the company said it expects to have 125 GW of gas equipment orders under contract by the end of this year. It also outlined capacity expansion to provide 30 GW of gas equipment by 2030, up from the 20 GW annual output projected for the third quarter. Those targets suggest management views demand as durable enough to justify scaling production.
Market reaction and what investors may focus on next
Power availability is increasingly shaping how investors evaluate the AI supply chain. As constraints shift from components to energy delivery, companies positioned to supply generation equipment and grid-related infrastructure may see their relevance rise. For GE Vernova specifically, the reported heavy-duty gas order surge and large backlog are supportive, but investors will still likely watch how quickly contracted equipment moves from orders to revenue, and whether execution timelines align with hyperscalers’ build schedules.
Next on the radar: GE Vernova’s continued order conversion and any updates on gas equipment capacity, alongside broader U.S. policy and infrastructure developments that influence permitting and grid expansion. Investors will also be watching the next round of AI demand signals from major data center operators, as well as forthcoming earnings and macro data that can affect interest rates and project financing conditions.
