Data Center Power Demand Sparks a New U.S. Gas Buildout

Planned U.S. gas-fired power capacity tied to data centers nearly doubled in the first half of 2026, according to OilPrice, with Amazon building a gas plant in Texas that would become the single-biggest source of power-related emissions in the country if developed as planned.

The artificial intelligence build-out has stopped being an abstract story about chips and started being a story about combustion. Planned gas-fired power capacity tied to U.S. data centers nearly doubled in the first half of 2026 alone, according to reporting by OilPrice, as hyperscale operators concluded that the grid as it exists cannot deliver firm electricity to their campuses on the schedule their models demand.

The most striking single project in that pipeline belongs to Amazon.com Inc. (AMZN), which is building a gas-fired power plant in Texas. If the plant is developed as currently planned, it is set to become the single-biggest source of power-related emissions in the entire United States. That is not a marginal footnote to the AI capital expenditure cycle. It is a statement about what the cycle actually runs on.

Why hyperscalers stopped waiting for the grid

Data centers are unusual electricity customers. They want power that is large, constant and available on a fixed date. Utility interconnection queues offer none of those three reliably. Renewables plus storage can supply volume and cheap marginal energy, but they do not by themselves guarantee that a training cluster stays energised at 3 a.m. on a windless night in August. Nuclear supplies firmness but not speed; new reactors do not arrive on an AI product roadmap's timeline.

That leaves gas turbines as the default answer for capacity that has to exist within a few years rather than a decade. Building generation on or beside the campus — behind the meter, in industry language, meaning the plant serves the facility directly rather than selling into the public grid — also sidesteps the interconnection wait and the transmission upgrades that come with it. The trade-off is that the emissions land on the operator's own ledger, which is why a corporate data center project can end up as the largest power-related emitting site in the country.

The doubling of planned capacity in six months is the number worth holding onto. Pipelines are not plants, and plenty of announced megawatts never get financed, permitted or connected. But a pipeline that expands that fast in half a year tells you that procurement teams across Big Tech reached the same conclusion at roughly the same time, and that they reached it after the initial round of AI power commitments had already been made.

Where the money lands

A gas buildout of this shape distributes spending along a long chain, and the constraints are not evenly spread.

• Turbine manufacturers. Heavy-duty gas turbines are made by a very small group of global suppliers with fixed factory throughput. When demand jumps, the scarce good is not the equipment price but the delivery slot. Order backlogs stretch, and pricing power moves to the manufacturer.
• Engineering and construction contractors. Someone has to pour the foundations, run the interconnects and commission the units. Skilled craft labour for power construction is already competing with the data center shells themselves.
• Gas producers and midstream operators. New generation means new firm gas demand, which means pipeline capacity to the site and long-dated supply contracts. Producers in basins with spare takeaway capacity and proximity to data center clusters are structurally better placed than those without.
• Utilities and independent power producers. Some hyperscalers will build and own; many will contract with a developer or regulated utility instead, converting a corporate obligation into a rate-based or contracted asset with a decades-long life.

The pattern to watch is who absorbs the risk. If hyperscalers own the plants outright, they carry fuel-price exposure and emissions liability directly. If they sign long-term power purchase agreements, that risk sits with a developer or utility and, potentially, with ratepayers who share the transmission and reliability costs.

Texas as the test case

Texas is the natural home for this activity: abundant gas, a permitting environment that moves faster than most, an independent grid, and land at scale. It is also where the political economy of AI power becomes visible. A single facility that ranks as the country's largest source of power-related emissions will attract regulatory attention, litigation risk and pressure from the same corporate sustainability commitments the operator has publicised for years.

Investors should treat that as a live variable rather than a settled one. Emissions-heavy projects can be redesigned mid-course — smaller, phased, paired with carbon capture, or partially converted to grid supply — and each of those changes moves the cost and the timeline. The phrase "if it's developed according to plan" is doing real work in the reporting.

The market read on Amazon

Amazon shares last traded at 266.43 in the most recent session, up 3.97% on the day from a prior close of 256.26, with a session range of 257.78 to 267.56. That advance ran against a soft tape: the S&P 500 tracker (SPY) closed at $769.35, down 0.23%, the Nasdaq 100 proxy (QQQ) at $716.43, down 0.65%, and the Dow tracker (DIA) at $535.06, down 0.03%, all as of 20:00 GMT on Friday, 28 August 2026.

The read-through is straightforward. Equity markets are currently rewarding demonstrated AI capacity, not penalising the fuel it takes to run it. Power procurement is being treated as an enabler of cloud revenue growth rather than as a cost or a liability. That framing holds until something forces a reprice — a state regulator, a shareholder resolution, a fuel-price shock, or a turbine delivery slot that slips two years and pushes a campus launch with it.

What to watch next

Three things will tell you whether the pipeline converts into steel in the ground. First, turbine order books and stated lead times, which are the hardest physical constraint in the chain. Second, firm gas transportation commitments to data center regions, because generation without dedicated supply is a paper plant. Third, the regulatory treatment of behind-the-meter generation and cost allocation for shared grid upgrades — the question of who pays for reliability when a single customer arrives with the appetite of a mid-sized city.

For anyone modelling the AI trade, the useful adjustment is to stop treating electricity as an input assumption and start treating it as a separate capital cycle with its own winners, its own bottlenecks and its own political risk.

Key facts

• AMZN last price: 266.43, +3.97% (as of 20:00 GMT, Fri 28 Aug 2026 close)
• Planned gas capacity tied to data centers: Nearly doubled in the first half of 2026
• Amazon Texas plant: Would be the largest single U.S. source of power-related emissions if built as planned
• Benchmarks at the close: SPY $769.35 (-0.23%), QQQ $716.43 (-0.65%), DIA $535.06 (-0.03%)

Frequently asked questions

How much new gas-fired capacity is tied to data centers?

OilPrice reports that planned gas-fired power capacity linked to U.S. data centers nearly doubled during the first half of 2026. That figure covers announced and planned projects rather than operating plants, so some of it may never be financed, permitted or built, but the pace of the increase over six months signals a broad shift in hyperscaler power procurement.

What is Amazon building in Texas?

Amazon is developing a gas-fired power plant in Texas to serve its data center operations. According to the reporting, if the facility is built as currently planned it would become the single-biggest source of power-related emissions anywhere in the United States. No completion date or capacity figure was disclosed in the material available.

Why are tech companies choosing gas over renewables or nuclear?

Data centers need very large, constant, firm power on a fixed timeline. Renewables plus storage supply cheap energy but not guaranteed round-the-clock output at scale, and new nuclear cannot be delivered fast enough. Gas turbines can be built in a few years and sited near the campus, avoiding long grid interconnection queues, at the cost of higher emissions.

Which industries benefit from a data center gas buildout?

Spending flows to heavy-duty gas turbine manufacturers, whose factory slots become the binding constraint; engineering and construction contractors; natural gas producers and pipeline operators supplying firm fuel; and utilities or independent power producers that build and contract the plants. Turbine lead times and gas transportation commitments are the practical bottlenecks to watch.

How did Amazon stock perform in the latest session?

Amazon shares last traded at 266.43, a gain of 3.97% from the previous close of 256.26, with a day range of 257.78 to 267.56, as of the close at 20:00 GMT on Friday, 28 August 2026. That rise came while the S&P 500, Nasdaq 100 and Dow trackers all finished slightly lower on the day.

What does 'behind the meter' generation mean?

Behind-the-meter generation is a power plant that supplies electricity directly to a specific facility rather than selling into the public grid. For data centers it removes the wait for a grid interconnection and the associated transmission upgrades, but it places the fuel cost, operating risk and emissions squarely on the site owner's own books.

Sources

• Data Centers Are Driving a New U.S. Natural Gas Buildout — OilPrice

Photo: Brett Sayles · Pexels Licence — source