On October 21, Kinder Morgan formally committed to building one of the largest natural gas pipelines sanctioned in years — and the customer that made it bankable was not a utility, not a gas producer, and not an LNG exporter. It was a data center operator.
The project is Permian Link, a roughly $4.2 billion pipeline to be built through Kinder Morgan’s Natural Gas Pipeline Company of America (NGPL) joint venture. It will carry about 2.8 billion cubic feet of gas per day across roughly 434 miles, from the Permian Basin near Pecos County in West Texas north to NGPL’s Midcontinent system, serving power demand in the Texas Panhandle. Kinder Morgan’s share of the cost is about $1.6 billion. The line is expected to enter service in late 2030.
The anchor customer is power providers serving an unnamed “A-list hyperscaler,” under long-term, take-or-pay transportation agreements running at least 20 years. The pipeline will supply gas for new power generation serving that operator’s facilities. Kinder Morgan has not named the hyperscaler, has not disclosed the generation technology the gas will feed, and has not disclosed any price terms — and those silences are themselves informative. What matters for the industry is the structure: a single data-center customer, creditworthy enough and committed enough, underwrote a multi-billion-dollar pipeline the way anchor tenants have underwritten generation deals for the past two years.
The anchor-tenant model comes to pipelines
Data-center power deals have followed a recognizable pattern since the AI build-out accelerated. A hyperscaler signs a long, large, credit-backed commitment — a power purchase agreement, a colocation arrangement, a capacity reservation — and that commitment converts a speculative infrastructure project into a financeable one. Generation developers, utilities, and now pipeline operators have all learned to build on the strength of that signature.
Permian Link extends the pattern one layer deeper into the energy supply chain. Until now, the anchor-tenant logic mostly stopped at the power plant gate: the hyperscaler contracted for electricity, and somebody else worried about fuel. Here, the commitment reaches upstream to the fuel artery itself. The pipeline exists because a data-center customer, through its power providers, promised to pay for capacity for two decades whether or not it uses every molecule. That is what a take-or-pay contract means, and it is what allowed Kinder Morgan to take a final investment decision on a $4.2 billion project in a single board action.
Kinder Morgan’s natural gas pipelines president, Sital Mody, put the dynamic plainly on the company’s earnings call: the unnamed hyperscaler is the anchor tenant and is “obviously driving the need for speed.” Speed is the scarce commodity in data-center development, and it is now shaping pipeline development too. Chief executive Kim Dang described the project as capturing a “compelling opportunity” by moving Permian gas north to emerging data-center clusters in the Panhandle.
The timing is worth noting. Permian Link was announced alongside Kinder Morgan’s third-quarter results, which beat expectations — adjusted earnings of 35 cents per share against a 32-cent consensus, on revenue of $4.53 billion, up 9.3 percent year over year. The two pipeline approvals announced that day lifted Kinder Morgan’s project backlog to $14.4 billion, including a record $13.2 billion of natural gas projects. This is not a company stretching to fund a speculative bet. It is a company whose backlog has become, in large part, a data-center and large-load backlog.
Why the Panhandle, and why gas
The geography of Permian Link tells the siting story. Data-center development has clustered where power, land, fiber, and community consent align, and the Texas Panhandle offers an unusual combination: abundant land, relatively permissive development conditions, and proximity to some of the cheapest gas in the world. What it has historically lacked is the infrastructure to turn that gas into reliable power at data-center scale. A 2.8 Bcf/d pipeline is a deliberate attempt to close that gap.
That matters because of how data centers actually use gas. The industry’s bring-your-own-power trend — on-site turbines, and in some cases dedicated off-site generation — depends entirely on fuel delivery. A data center can install turbines in 18 to 30 months, far faster than a grid interconnection that may take years, but turbines without a pipeline are ornaments. Fuel supply is the binding constraint that rarely makes the headlines, and Permian Link is what removing that constraint looks like: not a turbine order, but a 434-mile steel commitment to deliver the molecules.
There is a second, subtler reason the project matters for reliability. Kinder Morgan’s chief financial officer, David Michels, emphasized on the earnings call that Permian Link’s link to storage sets it apart — storage provides the flexibility and reliability that hyperscalers require. Data centers need power around the clock, and gas-fired generation paired with storage can firm up output in ways that a bare pipeline connection cannot. The storage link is the difference between a fuel supply and a reliability product, and reliability is what hyperscalers are actually buying.
For illustration, consider what 2.8 Bcf/d represents. At typical combined-cycle efficiencies, that volume of gas could fuel on the order of 15 to 20 gigawatts of generation — several times the demand of a single hyperscale campus. Permian Link is therefore not a pipeline for one data center. It is a pipeline anchored by one customer and sized for a cluster, with Kinder Morgan executives noting that other customers have already shown interest beyond the anchor. The anchor tenant opens the corridor; the corridor then invites everyone else.
Pipelines are now data-center infrastructure
The deeper shift here is definitional. Two years ago, a data center’s infrastructure stack ended at the substation: the facility bought power, and the power system was somebody else’s problem. Then came on-site generation, which pushed the boundary to the fence line. Permian Link pushes it further still — to the wellhead, in effect. When a hyperscaler anchors a pipeline, fuel transportation becomes part of the data-center development stack, alongside land, water, fiber, and permits.
This has consequences for how the industry should think about siting. Our siting coverage has stressed that power is only one variable: land, water, fiber, and community consent decide where facilities actually get built. Permian Link adds a variable to that checklist, or rather elevates one that was always there. A site near abundant gas is not the same as a site with deliverable gas. The difference is a pipeline, a compression network, storage access, and a firm transportation contract — infrastructure that takes four years to build and must be planned on the same horizon as the data center itself.
It also changes the competitive map. Regions that can offer firm, deliverable gas supply gain an edge in attracting the next wave of campuses, just as regions with available transmission capacity did in the last wave. The Panhandle’s pitch is now concrete: a named pipeline, a sanctioned budget, a 2030 in-service target, and an anchor customer whose credit underwrote it.
The four-year problem
Permian Link also exposes an uncomfortable timeline mismatch. The data-center industry operates on a clock measured in quarters: chips ordered, campuses announced, capacity promised to customers on aggressive schedules. Pipeline development operates on a clock measured in years. Permian Link was sanctioned in October 2026 and is not expected in service until late 2030. Four years will pass between the decision and the first molecules flowing north.
That gap is the central tension of the AI infrastructure build-out. Everything downstream of the pipeline — the generation, the campuses, the compute — can move faster than the fuel artery that feeds it. The anchor tenant’s 20-year commitment bridges the financing gap, but nothing bridges the construction gap except starting early. The hyperscaler that anchored Permian Link did what the industry’s most sophisticated players have learned to do: commit to infrastructure years before it is needed, because infrastructure committed late arrives too late.
The timeline also carries ordinary execution risk. Both newly sanctioned Kinder Morgan projects still await final regulatory approvals, and pipeline projects of this scale face permitting, land acquisition, and construction risks that can move schedules. The 2030 target is a plan, not a certainty, and readers should treat it as such. But the final investment decision itself is the meaningful event: it converts a proposal into a funded project with a budget, a route, and a customer.
The other pipeline in the announcement
Permian Link was not the only project sanctioned on October 21. Kinder Morgan also approved the Southeast Connector, a roughly 470-mile pipeline on its Southern Natural Gas system that will carry up to 2.6 Bcf/d to growing markets in the Southeast, with Kinder Morgan’s share at about $3.1 billion and phased in-service dates beginning in late 2030. Together, the two projects account for about $4.7 billion of Kinder Morgan investment.
The Southeast Connector is not a data-center pipeline in the same direct sense — its customers are investment-grade utilities — but it belongs to the same story. Kinder Morgan’s management cited more than 200 gigawatts of large-load interconnection requests across the Southeast, and Dang noted on the call that all of the projects the company is sanctioning today are “demand pull,” driven by buyers rather than producers. The pipeline industry spent a decade building supply-push projects to move shale gas out of producing basins. The current wave is the reverse: buyers of energy — utilities, power providers, and now hyperscalers — pulling infrastructure toward their demand.
What this means for the grid
There is a legitimate question about whether fueling data centers with new gas pipelines helps or hurts the broader power system. The honest answer is both, in different ways.
On one hand, gas-fired generation dedicated to data centers can relieve pressure on the grid. A campus that generates its own power, or buys from dedicated generation fed by Permian Link, is a campus that does not join an interconnection queue or demand grid upgrades that other customers would fund. In Texas, where the interconnection queue has become a central policy battleground, off-grid or dedicated-supply solutions can function as a safety valve, letting load growth proceed without waiting for system-wide upgrades.
On the other hand, dedicated gas supply can entrench a parallel power system. If the largest new loads increasingly self-supply through pipelines and on-site generation, the grid loses both their revenue and their flexibility. Regulators and grid planners will need to account for these loads explicitly, rather than treating them as invisible adjustments to the forecast.
There is also an emissions question that should not be waved away. Gas-fired generation serving data centers has the emissions profile of a gas-fired power plant, whatever building it sits next to. Kinder Morgan has disclosed nothing about the generation technology or any emissions commitments attached to this gas, and readers should not assume any. The climate debate over data-center power, previously fought over grid mix and renewable procurement, now extends to pipeline capacity as well.
The new siting checklist
For developers, utilities, and communities, Permian Link rewrites a piece of the siting playbook. The question for a prospective data-center site is no longer only whether transmission is available and how long interconnection will take. It is also whether firm fuel delivery is available, whether storage is accessible, and whether an anchor customer’s commitment can underwrite the infrastructure that does not yet exist.
That is a more complex negotiation than a power purchase agreement, and it will favor a particular kind of player: large, creditworthy, patient, and willing to sign 20-year commitments years before first power. Smaller developers without that balance sheet will find it harder to replicate the model, which may concentrate the next wave of development among the largest operators — the same concentration dynamic that already characterizes the industry.
For communities in the Panhandle and along the route, the project brings the familiar mix of construction employment, tax base, and infrastructure investment, alongside familiar concerns about land use, air quality, and the character of industrial development. A pipeline that serves data centers will face the same community scrutiny as the data centers themselves.
The artery and the stack
Permian Link is, at its core, a bet that the data-center boom is not a passing cycle but a structural shift in energy demand — durable enough to justify a $4.2 billion, 434-mile, 20-year commitment. Kinder Morgan is not the first company to make that bet, but it is among the first to make it in steel, in the ground, in the form of a pipeline whose anchor customer builds server racks rather than power plants.
The industry’s infrastructure stack has grown by one layer. Land, water, fiber, power, and now fuel delivery — each with its own lead time, its own permitting path, and its own anchor-tenant economics. The companies that understand the full stack, and commit to it early, will build the next wave of AI infrastructure. The ones that treat fuel as an afterthought will discover that turbines without pipelines are just expensive metal.
Four years from now, gas should be flowing north from Pecos County toward the Panhandle. Whether the data-center clusters it serves will have arrived on the same schedule is the open question — and the reason the anchor tenant signed for twenty years.