2,354 TWh by 2035: Data Centers Turn to Gas-Plus-Capture Power

Global data center electricity consumption is projected to reach 2,354 terawatt-hours by 2035 — roughly 7.4% of world power generation — and the industry's answer to that surge, laid out at a dedicated COP31 panel, is natural gas generation paired with carbon capture. The Bloomberg projection, cited during the panel on decarbonizing data centers, frames the arithmetic that operators cannot escape. AI facilities need uninterrupted baseload power while their computational loads swing sharply, and many operators are building dedicated gas plants to secure rapid "speed to power." The material consequence is a CO2 stream at roughly 4% concentration from gas turbine exhaust — dilute, voluminous and expensive to capture unless the plant was designed for it.

What did the panelists actually propose?

Ilaria Perena, Global Sales Director for Climate Technology Solutions at Baker Hughes, said integrating carbon capture from the outset, rather than as an afterthought, is essential for hyperscalers seeking rapid, low-carbon deployment. Baker Hughes is already supporting a 1-gigawatt data center power project designed for CCUS integration. The design pairs turbine exhaust capture with facility cooling and digital reservoir monitoring. Mona Mølnvik, Vice President of Research at SINTEF Energy Research and Director of GigaCCS, added two conditions:

• Capture at 4% CO2 concentration demands standardized, repeatable turbine designs, not bespoke one-offs.
• Developers cannot site data centers without planning the full capture, transport and storage chain; operators should co-locate near industrial CCUS hubs rather than risk stranded assets.

The takeaway is simple: capture belongs on the drawing board, not in the retrofit budget.

Where does the money come from?

A back-to-back COP31 session on financing global carbon management put the bottleneck in plain terms. The technology works; the business models do not yet scale. Jared Daniels, CEO of the Global CCS Institute, moderated that discussion and supplied the operating numbers: 113 commercial CCS facilities now run worldwide, and capacity will surpass 100 million tons per year once projects under construction come online. Scaling toward gigatons, he argued, requires market-driven capital. "Capital needs confidence," Daniels said. "There needs to be confidence in the entire value chain that you're going to get an appropriate return on that capital investment." Panelists pointed to templates that already attract institutional money:

• Olivia Powis, CEO of the Carbon Capture and Storage Association, cited UK revenue-support mechanisms that unlocked £10 billion in non-recourse debt and drew major infrastructure equity into the HyNet cluster.
• Tim Ashby, Global Head of Sustainability Strategy at HSBC, said the shift from "first-of-a-kind" to "nth-of-a-kind" projects depends on predictable, long-term revenue contracts — backed by stable carbon pricing, tax credits such as US 45Q, or corporate off-take demand from AI and cloud computing.

The takeaway: subsidies built the first hundred plants; contracts must build the next thousand.

Who buys the CO2, and who pays if they do not?

The financing session's structural argument deserves attention from anyone tracking the sector. Corporate off-take from AI and cloud operators was named alongside carbon pricing and 45Q as a revenue pillar, which effectively makes hyperscalers the anchor demand for capture infrastructure. That reverses the usual order of things. Data centers were the load problem; now they are the proposed offtake solution. Whether they sign long-term capture contracts at prices that clear institutional hurdle rates will decide whether the 2,354 TWh projection produces a matched capture build-out or a gas fleet with nowhere to send its carbon. Google's announced $17 billion Wyoming data center with carbon capture, reported alongside the panel discussions, signals the direction at least one hyperscaler has chosen. The milestone to watch is contractual, not technical: the first repeatable, long-term offtake agreements between data center operators and CCUS hub developers. Those signatures, not the capture chemistry, will set the pace.