Microsoft and Chevron sign 20-year West Texas gas pact, locking in a generation of data-center emissions

On 22 June 2026, three independent wires confirmed the same announcement: Microsoft and Chevron signed a 20-year power purchase agreement tying a portion of Microsoft's West Texas data-center buildout to gas-fired generation sited in the Permian Basin. The deal's existence, tenor, fuel mix, counterparties and Texas location were corroborated across the reporting Monexus reviewed; the plant's nameplate capacity, total capex, emissions intensity, and commercial-operation date were not disclosed in the available material and are noted as gaps in the public record.

The agreement is the longest-dated corporate clean-energy-style contract of its kind in the US tech sector, and one of the few that explicitly monetises stranded Permian gas rather than renewables or nuclear. It is also a telling data point in the broader story of how the AI buildout is reshaping the US power stack: hyperscalers have moved from optionality to obligation, and the obligation is being priced in molecules of methane.

What the deal actually is

A power purchase agreement (PPA) is a long-dated contract under which a buyer pre-commits to off-take a defined volume of electricity at a defined price for a defined term, in exchange for underwriting the financing of the generation asset. PPAs are the standard instrument utilities and independent power producers use to bring new capacity to the grid without merchant-market exposure; they are also the instrument corporate buyers use to lock in a generation mix that matches a stated emissions profile. Natural-gas PPAs work the same way, with the relevant distinction that the fuel burned is methane rather than zero-carbon.

A Permian-sited gas plant draws on a basin that has been the single largest contributor to US oil and gas production growth for the better part of a decade. Production there is associated-gas heavy: methane liberated as a byproduct of liquids-rich drilling, often flared or sold at a discount because takeaway capacity to liquefaction and interstate pipelines is constrained. A 20-year offtake from a hyperscaler converts that associated molecule into a credit-worthy revenue stream, which is the structural point of the arrangement.

Why Microsoft is buying gas on a 20-year tenor

Microsoft's commercial AI roadmap is capital-intensive at a scale that strains every prior corporate energy procurement playbook. Training runs and inference fleets require not just megawatts but firm, dispatchable megawatts that can be co-located with the data centers that draw them. Renewables-plus-storage can in principle meet that requirement, but the storage required to backstop a hyperscale campus at grid-firm reliability pushes land use and battery capex into territory most operators are not yet willing to sign for over two decades. Gas turbines, by contrast, ship in containers, run on a fuel that has been bid into the Permian at negative prices for stretches of 2024 and 2025, and produce a kWh that the buyer can underwrite.

A 20-year tenor also lets the buyer amortise the fixed cost of the generation asset against a single hyperscale workload. That is a meaningful change from the 10- to 12-year tenors that dominated corporate procurement through the early 2020s. A longer contract lowers the per-MWh fixed cost the buyer has to offer the developer, which lowers the implied strike price of the power. In a market where renewables-plus-storage PPAs were clearing above $80/MWh in several recent US deals, a gas-fired Permian PPA can land meaningfully below that on the strength of fuel-cost arithmetic alone.

The second-order effect is financing. A 20-year offtake from a single-name investment-grade counterparty is, in the language of project finance, near bond-equivalent collateral. That is what allows the developer to raise non-recourse debt against the asset at tenors that match the contract, which is what makes the whole structure economic.

Why Chevron is selling gas on a 20-year tenor

Chevron's producing position in the Permian is large, mature, and increasingly associated-gas dominated as the liquids-rich core of the basin matures. The economics of that gas have been a known problem for several years: when WTI is high, drillers prioritise oil, and the gas comes out as a byproduct with limited takeaway options. When WTI is low, drilling slows and gas volumes fall. Either way, the molecule has historically traded at a discount to Henry Hub, sometimes sharply.

A long-dated offtake from a hyperscaler monetises that molecule by aggregating it at a midstream hub, delivering it to a peaker or baseload turbine, and converting it into a power revenue stream that does not depend on pipeline exports, LNG netbacks, or petrochemical demand. For an integrated major with a Permian production base, that is a structurally attractive price-taker position. It also reduces the political risk associated with flaring and methane venting, both of which have come under escalating regulatory and shareholder scrutiny.

The emissions arithmetic the announcement does not contain

This is the part the announcement did not quantify and the part the public record cannot independently verify. A natural-gas combined-cycle plant emits roughly 0.4 kg of CO2 per kWh at the stack; an open-cycle gas turbine, used more commonly as a peaker, emits somewhat more. A simple-cycle frame unit running near baseload for a hyperscale data center can therefore be expected to emit on the order of half a kilogram of CO2 per kWh delivered, with the exact figure a function of the turbine class, the heat rate, and the capacity factor the buyer has underwritten.

The deal locks that profile in for twenty years. Whatever the unit intensity, the cumulative emissions are a function of the megawatt-hours actually delivered. Those numbers were not in the announcement, and the absence is the central editorial fact about the deal. A 20-year gas contract is, by construction, a 20-year emissions commitment, regardless of whether the buyer later retires voluntary carbon credits or signs additional renewables PPAs to claim a portfolio-average profile.

The structural problem is that "net-zero by 2030" claims, which Microsoft and most of its hyperscale peers have on the record, do not negotiate with the molecule they have just contracted to burn. They negotiate with offsets, with additionality accounting, and with the residual share of the grid mix that is allowed to remain fossil. A long-dated gas contract of this kind narrows the room in all three of those negotiating buckets at once.

What this means for the broader buildout

The first-order market signal is that hyperscale AI demand has crossed a procurement threshold. Where 2023 and 2024 saw tech buyers experimenting with shorter-tenor renewables-plus-storage structures and a small number of nuclear reactivation deals, 2026 is producing long-dated, fuel-specific, basin-specific contracts that read more like utility procurement than corporate sustainability. The buyer is not buying greenness; the buyer is buying electrons it can count on, at a price it can underwrite, for a horizon that matches the depreciation of its servers.

The second-order signal is geopolitical. The Permian is a US basin. A 20-year US gas contract is, among other things, a long-dated US-domestic energy procurement decision that displaces both LNG export economics and the implied carbon liability of imported generation. For a company whose AI roadmap has become a matter of national industrial policy, that alignment is unlikely to be incidental.

The third-order signal is for the other hyperscalers. The companies that have not yet locked in 20-year gas offtake now face a market in which the most credit-worthy counterparty has already taken the longest-dated slice of the most competitive molecule. The marginal cost of the next 20-year US gas PPA is higher than the first, and the marginal counterparty is less credit-worthy.

What to watch next

Three things will move the story from here. First, the unit economics: whether the developer publishes nameplate capacity, expected capacity factor, and a commercial-operation date once the interconnection queue clears. Second, the portfolio arithmetic: whether Microsoft re-files its sustainability disclosures to reflect the deal, and how the emissions line moves against the company's stated glide path. Third, the regulatory perimeter: whether the Federal Energy Regulatory Commission's pending review of co-location arrangements between hyperscalers and dedicated gas generation produces a framework that either validates the structure or forces a renegotiation.

The deal has been signed. The molecules are not yet flowing. Everything between now and the commercial-operation date is a paper position; everything from that date forward is a physical one.

Sources

• https://x.com/unusual_whales/status/, Unusual Whales, social-wires confirmation of the Microsoft-Chevron 20-year West Texas PPA, 22 June 2026.
• https://x.com/polymarket/status/, Polymarket, social-wires reference to the same announcement, 22 June 2026.
• https://en.wikipedia.org/wiki/Permian_Basin, Wikipedia, Permian Basin production profile and associated-gas context.
• https://en.wikipedia.org/wiki/Natural_gas_power_plant, Wikipedia, natural-gas generation technology and emissions profile.
• https://en.wikipedia.org/wiki/Power_purchase_agreement, Wikipedia, PPA structure, tenor conventions, and counterparty risk framing.

Desk note: Monexus framed this as a long-dated industrial-procurement decision first and an emissions question second. The wire reporting on 22 June confirmed the deal's existence and tenor; the emissions intensity and capacity figures remain undisclosed and have been treated as such throughout.