GE Vernova CEO: AI is 20% of our backlog, nuclear SMRs coming by 2031-2032, commissioning 10GW for TSMC in Taiwan

GE Vernova: Powering the AI Build-Out — and Everything Else

GE Vernova's equipment generates roughly a third of the world's electricity each day, excluding China. Scott Strazik, the company's CEO and a GE lifer of 25 years, isn't running a narrowly focused data center play — AI represents about 20% of GE Vernova's backlog. The other 80% is the rest of the world trying to keep the lights on.

Scale and shape of the business

The 85,000-person company spans gas turbines, wind turbines, nuclear reactors, and grid electrification equipment. Strazik describes the electrification and grid segment as the fastest-growing part of the business, in part because efficiency gains on the existing grid can deliver new electrons faster than building a new plant. The company added 1,800 production workers in the last 15 months and is adding 400 machines this year to automate manufacturing.

AI demand — real, but not the whole story

The concern when the US AI build-out accelerated was that it would price out other markets. That hasn't happened. In Q1, GE Vernova saw strong demand from Vietnam, Mexico, Canada, Saudi Arabia, and Kuwait alongside US hyperscaler orders. The international signal matters because GE Vernova is already at scale globally — so even a large AI-driven acceleration shows up as a minority of total backlog.

Hyperscalers aren't simply buying electrons, either. Strazik says customers have shifted from purchasing power purchase agreements to asking GE Vernova to design integrated systems tailored to how each AI factory needs to run. That's a different commercial relationship, and it's pushing GE Vernova to co-create with customers rather than just deliver equipment.

“AI specific, it's about 20% of our backlog today. New incremental nuclear plants — 2032, 2031, that's about right. We're commissioning almost 10 gigawatts of power primarily for TSMC in Taiwan. We added about 1,800 production workers in the last fifteen months.”

Taiwan and TSMC

One figure that lands harder than the data center narrative: GE Vernova is commissioning almost 10 gigawatts of power in Taiwan, primarily for TSMC's fabs. Chip fabrication, not just AI inference, is a major electricity consumer — and that demand is now replicating in the US as domestic fab construction accelerates.

Nuclear: first US SMR approval expected this summer

GE Vernova's small modular reactor is a 300-megawatt unit — enough to power 300,000 US homes — built within a footprint roughly the size of an American football field. The first plant is already under construction outside Toronto. Strazik expects NRC approval for the first US plant this summer, with construction starting immediately after. From groundbreaking to commissioning runs about four years, putting the first new nuclear electrons in the US around 2031–2032.

The economic logic is standardization. Every large nuclear plant built historically was a one-off, which is why the industry never came down the cost curve. SMRs are designed to be modular and repeatable — after the first unit, additional 300-megawatt blocks can be added roughly every two years. The upfront capital is heavy, but Strazik argues the lifecycle cost is compelling given the plant's longevity.

Separately, upgrading the 56 existing US nuclear plants can add 5 gigawatts without waiting for new construction — and that work is already underway.

Grid modernization

The US grid was engineered for coal plants pushing electrons in one direction toward homes and factories. It now has to handle variable generation sources, rooftop solar selling back into the grid, and fragmented, islanded regional networks that don't connect. Strazik argues a fully meshed national grid across the 48 states is technically achievable and would deliver incremental electrons faster than new plant construction — but the harder obstacle is regulatory fragmentation and misaligned market structures across regions.

China and the competitive boundary

On solar panels and stationary storage, China is ahead by a wide margin and Strazik doesn't expect that to change given the manufacturing investment already in place. The boundary where Western manufacturers still hold a meaningful lead is rotating equipment — gas turbines, wind turbines, and nuclear — where high-temperature material science remains the differentiating factor.

Craft labor as the binding constraint

GE Vernova can produce the equipment. The constraint is field construction — getting skilled craft labor to often-remote sites where data centers and grid infrastructure are being built. Every gas turbine or wind turbine GE Vernova makes uses the same factory workforce. The site readiness problem is different and harder to industrialize, and Strazik describes getting equipment delivery and site readiness synchronized as one of the company's biggest operational challenges.

M&A and Silicon Valley

GE Vernova is leaning into energy ventures and wants to be seen as a partner of choice for startups that reach industrialization scale but can't get there alone. Strazik frames the relationship as reciprocal: hyperscalers, now 20% of the backlog, are pushing GE Vernova to think about innovation timelines differently. The company is hiring 69 MIT graduates starting in July to complement its growing production workforce.

The core bet is that electricity demand growth, suppressed for most of the past 25 years, has re-accelerated structurally — driven by AI, chip fabs, and energy security concerns globally. GE Vernova is spending to meet that demand, with the SMR program, grid modernization, and manufacturing automation all running in parallel. The first test of the nuclear timeline comes this summer with the NRC ruling.