Carbon Capture, Integration, and the Strategic Bet

The oil-and-gas majors are not entering behind-the-fence power naively; they are entering with a specific answer to the objection their entry immediately raises. Burning natural gas at gigawatt scale to power AI produces a great deal of carbon dioxide, and in an era of climate commitments that is both a reputational and, increasingly, a regulatory liability. The majors' proposed answer is carbon capture — capturing the CO2 produced by generation before it reaches the atmosphere and sequestering it underground. Whether that answer holds up technically and economically is one of the pivotal questions of this whole market, and it deserves a clear-eyed look.

Start with why carbon capture is the majors' natural move rather than an afterthought. An integrated major already possesses much of the capture value chain. It has the subsurface geological expertise to identify and manage sequestration sites — the same skill set used to find and produce hydrocarbons, run in reverse to inject CO2 into stable formations. It often has depleted reservoirs and saline formations suitable for storage, and in some cases existing CO2 pipeline infrastructure. So when Exxon proposes to pair gas-fired data-center power with carbon capture, it is not bolting on an unfamiliar technology; it is deploying a capability adjacent to its core business. The integrated pitch becomes: we supply the gas, generate the power, capture the carbon, and store it in formations we understand — a full-stack, lower-carbon power service that a standalone developer could not easily assemble.

But the practitioner must hold the enthusiasm at arm's length, because carbon capture carries real and unresolved challenges. Capturing CO2 from combustion exhaust is energy-intensive — it imposes a parasitic load that reduces the net power available and raises the effective cost per delivered megawatt-hour. Capture rates in practice have often fallen short of headline promises, meaning a meaningful fraction of the carbon can still escape. The economics depend heavily on policy support such as tax credits for sequestered carbon, and on the cost of building capture and transport infrastructure. So while carbon capture is genuinely credible as a strategy, especially in the hands of players with the geological and project capabilities, it is not a solved problem, and a sober analyst treats capture-equipped gas power as lower-carbon rather than zero-carbon, and prices in execution risk. That is Faraday's read: the capture story is plausible and strategically coherent, but it is a bet, not a settled fact, and its success is contingent on capture rates, cost, and durable policy.

Why do the majors make the bet anyway? Because the alternative — powering data centers with unmitigated gas — invites exactly the emissions and permitting friction that could constrain the market and their own license to operate. If AI's power demand is going to be met substantially by natural gas, the majors would rather be the ones defining how that gas is burned and how its carbon is managed than cede the ground or be cast as the problem. Carbon capture lets them square their gas-growth ambition with the climate commitments they and their customers have made. Hyperscalers, many of which carry aggressive corporate decarbonization targets, cannot be seen to power their AI on unabated fossil generation without eroding their own credibility. A capture-equipped gas plant offers them a story they can defend to their own stakeholders. The two sides' needs align: the major gets a gas-demand outlet, the hyperscaler gets firm power with a defensible carbon narrative, and carbon capture is the bridge that makes the deal palatable to both.

Step back to the strategic level and the shape of the bet is clear. The majors are wagering that behind-the-fence power for AI is a large, durable market; that natural gas will be a dominant fuel for it because of its firmness, availability, and speed relative to the grid; that their gas-supply and project-execution capabilities give them a decisive edge; and that carbon capture can manage the emissions well enough to keep the whole enterprise licensed to operate. If those bets pay off, the majors transform a century-old molecule business into an electron business tied to the fastest-growing demand story of the age, and they anchor gas demand for decades. If carbon capture disappoints, or if cheaper firm-and-clean alternatives mature faster than expected, the bet looks more exposed — stranded capital in a decarbonizing world.

The practitioner's takeaway is to read every major-backed data-center power announcement on two levels at once. On the surface it is a power-supply deal. Underneath it is a claim about the future of gas, the viability of carbon capture, and the belief that AI's hunger for firm power will outrun the grid for years to come. All of it rests on the thesis this course keeps returning to: power is the first-order constraint on AI infrastructure. The majors are moving because they have concluded that whoever solves the power constraint — with gas, at scale, on the compute's timeline — captures one of the defining markets of the decade. That conviction is why the emissions and permitting trade-offs of the next module are not a footnote. They are the terrain on which this entire contest will be decided.