Catalyst N° 059 of 125 17 Apr 2025

Serving data center load with carbon capture

with Julio Friedmann, chief scientist, Carbon Direct

In this note

01The question
02The answer
03The argument
04What you need to know first
05Details worth keeping
06Claims worth citing
07Where it’s contested

The question

Hyperscalers are driving a large gas buildout while holding real decarbonization commitments. Can carbon capture square that circle, or is “capture ready” the same empty promise “hydrogen ready” turned out to be?

The answer

Friedmann’s case is that the economics now work, the technology is mature, and the gas plants are being built regardless, so capture is the difference between controlled and uncontrolled emissions rather than a choice between gas and something cleaner. But he does not claim it is happening. No gas power plant with capture is operating anywhere, no data center project has announced one, and whether this becomes real turns on storage geography and on US policy questions that were open when he spoke.

03The argument

Start with the priority list, because it governs everything downstream. What data center developers want, in Friedmann’s ordering, is speed, speed, speed, cost and carbon. Carbon is fifth, and speed is what is driving multi-gigawatt gas announcements even as turbine shortages and cancellations start to bite. That sets up the reframe the episode rests on: the gas is being built either way, so the comparison is not gas against renewables or nuclear but the same plant emitting or not emitting. Friedmann credits the hyperscalers for keeping carbon on the list when nothing forces them to, and says that where renewables are cheaper and faster, build renewables.

Then comes the counterintuitive part. Gas flue gas is only 4% to 7% carbon dioxide when a plant runs well, and dilute streams are expensive to capture from. So in dollars per ton, gas is the most expensive place in the power sector to do carbon capture, which is the figure most people have absorbed. In dollars per megawatt-hour it inverts and gas becomes the cheapest, because gas emits far less carbon dioxide per unit of energy than coal, so there is less to capture per unit of electricity sold. Which denominator you pick decides whether gas capture looks costly or cheap, and a power buyer cares about the megawatt-hour. His numbers put a new build with everything included at $70 to $100 per megawatt-hour, competing with wind, solar and batteries in many markets and beating nuclear in many. The figure carries a hard condition: it assumes $3 per million British thermal units gas, which is why he rules Europe out and points to the Middle East and Australia.

Kann puts the obvious objection directly, borrowing David Hawkins’ line that if a power plant is capture-ready then his driveway is Ferrari-ready. Friedmann calls the concern reasonable and offers three reasons this moment differs: the plants are being built anyway, the 45Q tax credit means capture can now get paid for, and hyperscalers will pay a premium for clean power as they already do for nuclear, except gas with capture builds in roughly half the time. Then he concedes the substance of the objection anyway. Nobody has said of a specific plant that they will definitely capture. His own example is Meta’s 2.6 gigawatt, $10 billion Louisiana project, where he expects capture because he knows people working on it and where nothing has been announced. The industry, he says, has to move from capture ready to capture committed, which is an admission that it has not.

What decides where this can happen is geology rather than technology. No carbon dioxide transport and storage means no project, full stop. New England has nowhere to put it. Virginia, where the data centers actually are, has poor storage. Texas, Louisiana, central Illinois, the Rockies and parts of California are good. That adds another circle to a siting Venn diagram already crowded with power, fiber, land and community consent, though the aperture is wider than people assume, because carbon dioxide moves by rail and barge as well as by pipeline. Kann presses that all of this works against priority number one, since more permitting and infrastructure means slower. Friedmann pushes back with sequencing: build the gas plant fast, start earning revenue, and retrofit capture over the following couple of years, since generation and capture can be decoupled. He then immediately grants that this recreates exactly the risk the Ferrari line was about, because the second step has to actually happen.

04What you need to know first

45Q: The US tax credit paid per ton of carbon dioxide captured and stored. It is what makes the revenue side of these projects exist, and its survival in pending legislation is an open question throughout the episode.
Flue gas concentration: How much carbon dioxide is in the exhaust stream. A dilute stream costs more per ton to separate, which is the whole basis of the per-ton versus per-megawatt-hour inversion.
Class VI well: The US permit category for wells injecting carbon dioxide into deep formations for permanent storage. These permits have been the long pole for any US storage project. Some states hold primacy, meaning they run the program themselves rather than the federal Environmental Protection Agency.
Saline storage versus enhanced oil recovery: The first puts carbon dioxide into deep salt-water rock purely for disposal; the second uses it to push out more oil. Most historical projects did the second; most now do the first.

05Details worth keeping

The technology is old: first capture device fielded in 1938, first carbon dioxide injections in 1972, integrated systems for climate since 1997.
Every capture technology gets tested on natural gas, at the National Carbon Capture Center in the US and Technology Centre Mongstad in Norway, because gas exhaust lacks the sulfur, particulates and other contaminants that complicate coal and industrial streams.
The Teesside project in the UK, under construction, planned to start in 2026 and storing offshore under the North Sea, would be the first gas power plant built with integrated capture. A contract for differences supports it, which Friedmann calls the kind of policy that acts as a magnet.
Scale is opening a lane for newer vendors. Established suppliers (Fluor, Schlumberger, Shell, CANSOLV, Mitsubishi) sell facilities capturing up to 10 million tons a year with a performance guarantee, but hyperscalers are talking about starting at 50 to 100 megawatts, where those are the wrong size. That favors next-generation solvents, physical sorbents, cryogenic approaches and membranes, each with a different sweet spot by duty cycle, heat rate and climate.
Carbon dioxide moves by rail and barge, not only pipeline. His illustration, offered explicitly without endorsement, is barging it from Minnesota down to Louisiana, which their analysis finds more cost competitive than people assume.
Vertically integrated offerings are emerging fast: Chevron with GE Vernova and Engine No. 1, ExxonMobil offering to build, run and store, Baker Hughes, and Mitsubishi and Kawasaki supplying turbine, capture technology and engineering contract together.
Canada’s tax credit is strong enough that Friedmann thinks the first integrated gas-plus-capture data center project could land in Saskatchewan, Manitoba or Alberta. Gulf states are courting hyperscalers, and Japanese and Korean firms are selling the whole package into northwestern Australia and Mexico.
Communities have generally not stopped the gas plants and want the revenue and jobs; capture also cuts conventional pollutants. He calls it a retail conversation, one community at a time.

06Claims worth citing

All figures as stated on 2025-04-17 and attributed to the speaker, not verified independently. The policy figures in particular were describing live legislative fights.

Natural gas flue gas is 4% to 7% carbon dioxide when the plant is running well. Friedmann
New build, all costs included (capture, compression, transport, storage, permitting), assuming $3 per million British thermal units gas: $70 to $100 per megawatt-hour total. Carbon Direct report published that March, cited by Friedmann
Retrofit of a fully depreciated existing plant: $40 to $70 per megawatt-hour, with an adder of about $25 to $30 per megawatt-hour in most markets, and 45Q covering roughly half of that. He runs the retrofit range and the adder together in one sentence, so which case the adder attaches to is not stated cleanly. Friedmann
About 40 carbon capture facilities worldwide capturing roughly 60 million tons per year, now mostly for saline storage rather than enhanced oil recovery. Friedmann
Gas with capture can be built in roughly half the time of a nuclear plant. Friedmann
Meta’s Louisiana project: 2.6 gigawatts and $10 billion. Friedmann
Canada offers a 60% investment tax credit for carbon capture. Friedmann
Roughly $8 billion of projects, including Office of Clean Energy Demonstrations awards, hydrogen hubs and direct air capture hubs, are exposed to proposed rescissions. Friedmann
On 45Q’s survival: 21 House members wrote to Speaker Johnson and four senior senators including Murkowski have come out for preserving the credits, numbers he says are large enough to be blocking in both chambers. Kann notes the letters cover a basket of energy credits rather than 45Q by name. Friedmann and Kann
California permitted four Class VI wells in Bakersfield despite not holding primacy; the bipartisan infrastructure law gave the Environmental Protection Agency $50 million to train permitting staff. Friedmann
Newer capture technologies are at technology readiness level six or seven. Friedmann

07Where it’s contested

The capture-ready objection is granted, not answered. Friedmann calls it a reasonable concern, sets his own bar at “capture committed,” and confirms nobody has met it. The strongest case against his thesis is his own standard.
Speed versus complexity. Kann argues carbon capture pushes directly against the top three priorities. Friedmann’s rebuttal is the two-step build, and he immediately concedes that the second step becoming optional is the risk.
The Meta example is an expectation, not a fact. He expects capture there because of who is working on it, and says no announcement exists and that skepticism is reasonable until one does.
The cost figures carry conditions. The $70 to $100 per megawatt-hour number requires cheap gas, which he flags as excluding Europe, and it comes from his own firm’s report. Carbon Direct is commercially active here, advising nearly all the hyperscalers plus utilities and energy park developers on these projects.
He does not claim this beats the alternatives generally. He says renewables will be cheaper and faster in a bunch of markets and that you should build them there, and that there is “hair on all of this.” His argument is scoped to places where gas is being built anyway and storage exists nearby.
Community benefit is unresolved. Most communities do not understand what carbon capture is, and he describes the benefits to them as opaque at best and complicated at worst.
The policy outlook is explicitly unstable. He declines to detail the administration’s moves as too fast-changing to be useful, gives only “high odds” on 45Q surviving, and identifies the uncertainty itself, rather than cost or technology, as the thing killing investment decisions.