{"id":4291,"date":"2026-09-27T20:46:06","date_gmt":"2026-09-27T20:46:06","guid":{"rendered":"https:\/\/projectfifty4.com\/chevron-power-plant-instead-of-ppa\/"},"modified":"2026-09-29T09:24:18","modified_gmt":"2026-09-29T09:24:18","slug":"chevron-power-plant-instead-of-ppa","status":"publish","type":"post","link":"https:\/\/projectfifty4.com\/fr\/chevron-power-plant-instead-of-ppa\/","title":{"rendered":"Why Chevron Is Building a Power Plant Instead of Buying Grid Power"},"content":{"rendered":"

Project Kilby will put roughly 2.67 gigawatts of dedicated gas generation next to a Microsoft data centre in Reeves County, Texas, under a twenty year take or pay agreement. The obvious question is why an oil company builds and operates a power station rather than letting the grid supply it. The answer is a queue, a turbine order book and a negative gas price, and only one of the three is about electricity.<\/strong><\/p>\n

Why is Chevron building its own power plant for Microsoft instead of supplying it from the grid?<\/h2>\n

Because a grid connection was not available in the time Microsoft needed it, and because the inputs Chevron already controls are the scarce ones. ERCOT is tracking more than 438 gigawatts of large load interconnection requests, close to 90 per cent of them data centres, and its Batch Zero reform will not produce a final transmission plan until autumn 2027. Lawrence Berkeley National Laboratory found the median time from interconnection request to commercial operation exceeded five years for generation projects completed in 2025. ERCOT has confirmed that facilities which are truly islanded with no grid connection generally fall outside its interconnection process, so Kilby does not shorten the queue, it exits it. Chevron had also reserved seven GE Vernova 7HA turbine slots in November 2025, before the Microsoft agreement was signed, in a market where the three dominant manufacturers are sold out into the 2030s. And the fuel is Permian associated gas that traded at an average of minus 2.19 dollars per million British thermal units at the Waha hub across 2026 to mid August.<\/p>\n

Points cl\u00e9s \u00e0 retenir<\/h2>\n
    \n
  • The decision is about time, not cost. Ascend Analytics notes that levelised cost analysis still favours grid connection over the long run, and both Chevron and Microsoft say they expect to connect to the grid eventually. Building is what you do when the schedule is the constraint.<\/li>\n
  • The turbine slot is the real moat. Chevron reserved seven GE Vernova 7HA turbines in November 2025, before the Microsoft deal existed. GE Vernova held 116 gigawatts of gas equipment under backlog and slot reservations at the end of the second quarter of 2026 and is largely sold out through 2030.<\/li>\n
  • The fuel advantage is a distressed asset. Waha hub gas averaged minus 2.19 dollars per million British thermal units across 2026 to mid August, against a five year average of plus 2.88, and printed negative on 99 days including a 90 day consecutive run.<\/li>\n
  • Microsoft is paying for the plant, not just the power. Microsoft states that the energy infrastructure required to power this datacenter is being funded by Microsoft, and Chevron describes the arrangement as a twenty year take or pay agreement for firm behind the meter capacity.<\/li>\n
  • The capital number has never been disclosed. Filed and reported figures span 6 billion dollars in a Reeves County tax abatement notice, 7 billion from Bloomberg, and 9 billion back solved by a TD Securities analyst from an assumed 15 per cent developer return. Chevron has published none of them.<\/li>\n
  • The moat closes while the plant is being built. Around 4.5 billion cubic feet a day of new Permian takeaway arrives around final investment decision, narrowing the discount that makes the fuel cheap, and ERCOT expects the majority of its queued requests to be operational by 2030.<\/li>\n<\/ul>\n

    The facts, and where the sources disagree<\/h2>\n

    Project Kilby is a behind the meter gas fired generation complex on more than 2,000 acres in Reeves County, Texas, roughly twenty miles south of Pecos. It is being developed by Energy Forge One LLC, a fifty fifty joint venture between Chevron and Joulent, an energy venture spun out of Engine No. 1 and launched on 22 June 2026. National Grid Ventures agreed on 1 July 2026 to invest 1.75 billion dollars for a 35 per cent stake in Joulent, implying a valuation of roughly 5 billion dollars.<\/p>\n

    The customer is Microsoft, which announced a data centre of approximately 2 gigawatts at Pecos on 22 June 2026, describing a multibillion dollar investment over five to seven years and more than 6,000 peak construction jobs. Chevron characterises the commercial arrangement as a twenty year take or pay power purchase agreement for 2.67 gigawatts of firm behind the meter capacity, targeting mid teens returns and long duration contracted cash flows that are independent of commodity price cycles. First power is targeted for 2028 and full capacity for 2031, with a final investment decision expected by the end of 2026.<\/p>\n

    Two things are genuinely unsettled and should not be reported as though they are not. The first is capacity. Four figures circulate because they describe different instruments: 2.67 gigawatts is Chevron’s contracted figure, about 2.5 gigawatts appears in stale page metadata and earlier Bloomberg reporting, roughly 2,595 megawatts is the figure in the TCEQ air permit application, and roughly 2,868 megawatts is the arithmetic sum of the unit level entries in the permit record. Use 2.67 gigawatts and say it is the contracted number.<\/p>\n

    The second is capital cost, which Chevron has never disclosed. A Reeves County Emergency Services District tax abatement notice of March 2026 carries 6 billion dollars, the only figure from a filed document. Bloomberg reported 7 billion. TD Securities analyst Jason Gabelman modelled 9 billion by assuming most of the project is financed and back solving from a 15 per cent developer internal rate of return, which also produced an implied contract price of about 150 dollars per megawatt hour. Quoting that price alongside Chevron’s mid teens return guidance prints the same assumption twice. Chief financial officer Eimear Bonner confirmed only that Kilby sits inside the 2026 organic capital outlook of 18 to 19 billion dollars.<\/p>\n

    The equipment mix is knowable from the TCEQ permit record and it is more interesting than the headline. Two combined cycle units of 555.7 megawatts each use GE Vernova 7HA.02 machines for 2028, three further 7HA.02 units total 990 megawatts for 2029, and the fast tranches are modular: twelve Solar Turbines T350 units totalling 373.2 megawatts for 2027 and twenty nine Solar Turbines T130 units totalling 392.95 megawatts for 2028. Solar Turbines is a Caterpillar subsidiary. The plant is phased deliberately so that modular capacity arrives years before the heavy frames.<\/p>\n

    The queue is the answer, and ERCOT said so<\/h2>\n

    The decisive regulatory fact was published by Utility Dive on 22 June 2026, the same day the Chevron and Microsoft announcement landed, quoting ERCOT directly: facilities that are truly islanded with no grid connection would generally fall outside ERCOT’s interconnection process, though they may still be subject to registration requirements with the PUCT.<\/p>\n

    That is the whole trade in one sentence. Kilby does not get to the front of the queue. It leaves the queue.<\/p>\n

    The scale of what it is leaving is the part people underestimate. On the generation side, Lawrence Berkeley National Laboratory’s Queued Up 2026 edition, using data through the end of 2025, found around 2,060 gigawatts of generation and storage actively seeking interconnection across roughly 8,200 projects. Active natural gas capacity in the queues stood at 253 gigawatts, up 86 per cent in 2025 and the only category that grew. The median duration from interconnection request to commercial operation exceeded five years for projects built in 2025, and of all the capacity that entered queues between 2000 and 2020, only 13 per cent had reached commercial operation by the end of 2025 while 75 per cent was withdrawn.<\/p>\n

    On the load side, ERCOT is tracking more than 438 gigawatts of large load interconnection requests, nearly 90 per cent of them data centres. Governor Abbott cited roughly 474 gigawatts when he directed the PUCT and ERCOT on 3 August 2026 to audit every data centre project in the process. ERCOT’s Batch Zero reform, approved by the PUCT on 18 June 2026, groups loads of 75 megawatts or more into coordinated study batches, but the final transmission plan for Batch Zero is not expected until autumn 2027 and Batch 1 applications do not open until summer 2027. A grid routed campus would still be waiting for a plan in late 2027 against Kilby’s 2028 first power.<\/p>\n

    The commercial value of that difference is large enough to justify almost any capital structure. Cleanview estimates that AI data centres can generate 10 to 12 million dollars per megawatt annually, or 10 to 12 billion dollars per gigawatt, and that bringing a data centre online even a few years early could result in tens of billions in revenue that would otherwise be forgone. At 2 gigawatts, a two year acceleration is worth more than the entire plant. That is why Microsoft is funding the generation rather than buying the electricity.<\/p>\n

    One honest qualification. Behind the meter does not remove queues, it substitutes them: air permits, gas connection, equipment reservation, storage and controls, engineering and construction capacity, and financing. Cleanview’s own data bears this out. Of roughly 90 gigawatts of announced behind the meter capacity it identified across 59 projects, only about 2 gigawatts, 2.2 per cent, was actually operating in mid 2026.<\/p>\n

    Turbine slots, reserved a year early<\/h2>\n

    In November 2025, before any Microsoft agreement existed, GE Vernova, Chevron and Engine No. 1 announced a venture targeting up to 4 gigawatts for United States data centres, with the first power foundries to use seven GE Vernova 7HA turbines secured under a slot reservation agreement. That reservation, not the Permian acreage and not the customer relationship, is the asset that made Kilby possible on this timeline.<\/p>\n

    The reason is that the manufacturers are sold out. GE Vernova reported 116 gigawatts of gas equipment under backlog and slot reservations at the end of the second quarter of 2026, comprising 53 gigawatts of firm backlog and 63 gigawatts of slot reservations, up from 100 gigawatts three months earlier, with Power segment orders of 16.7 billion dollars, up 134 per cent organically. Chief executive Scott Strazik said the company expects at least 125 gigawatts of gas equipment under contract by the end of 2026, and that it remains on track to deliver 20 gigawatts of annual gas turbine output in the third quarter of 2026, 24 gigawatts in 2028, with actions under way to produce 30 gigawatts in 2030. Customer discussions, he said, already stretch out to 2032 and beyond.<\/p>\n

    The picture is the same elsewhere. Siemens Energy reported 69 gigawatts of gas turbine backlog plus 26 gigawatts under slot reservation agreements in its August 2026 quarterly presentation. Mitsubishi Power’s head of EMEA told Platts that lead times have stretched to five years or more and that the company is signing contracts for deliveries between 2031 and 2034, describing the bottleneck as whole supply chain capacity rather than demand or capital.<\/p>\n

    Global Energy Monitor’s January 2026 assessment puts a number on how binding this is: those three manufacturers account for more than 75 per cent of the market for gas power in development with a named manufacturer, and two thirds of projects in development have no named manufacturer at all. In other words, most announced capacity has no machine behind it. A slot reservation is the one thing a power purchase agreement cannot buy you.<\/p>\n

    This is also why the modular tranches matter. Enverus Intelligence Research expects reciprocating engines, small and medium frame turbines and fuel cells to make up 61 per cent of gas fired behind the meter capacity through 2030, on commercial operation timelines of 18 to 24 months against up to 80 months for large frame equipment. Kilby’s Solar Turbines units arrive from 2027, the heavy GE frames from 2028 and 2029. The project is structured to deliver revenue before the big machines land.<\/p>\n

    Gas that costs less than nothing<\/h2>\n

    Chevron’s stated edge is leveraging Permian natural gas, and the economics of that phrase are unusual. Every Permian oil barrel brings associated gas with it. Producers who cannot flare within regulatory limits and cannot store it face a binary choice: pay to dispose of the gas, or shut in the oil well. Most pay.<\/p>\n

    The result is a hub price that goes below zero. Reuters reported that Waha hub prices stayed negative for a record 47 consecutive days as of 14 April 2026. Across 2026 to mid August, Waha cash averaged minus 2.19 dollars per million British thermal units against a five year average of plus 2.88, printing negative on 99 days including a consecutive run of 90 days that broke in mid June. Prompt Waha to Henry Hub basis moved from below minus 5.00 dollars in the spring to minus 0.91 by late July. East Daley estimated well curtailments on Delaware basin gathering and processing systems averaging 220 million cubic feet a day in 2026, more than double its earlier forecast of 100.<\/p>\n

    A plant that converts that gas into contracted electricity at the wellhead captures the whole basis spread and simultaneously creates an in basin demand sink that supports Chevron’s own upstream netbacks. A grid power purchase agreement would deliver Chevron none of that. This is the part of the deal that is genuinely specific to an oil company and cannot be replicated by an independent power producer.<\/p>\n

    It is also the part with a clock on it. Roughly 4.5 billion cubic feet a day of new Permian takeaway arrives around the time of final investment decision, with Blackcomb at 2.5 billion cubic feet a day from Waha to Agua Dulce in the third quarter of 2026 and Hugh Brinson Phase 1 at 1.5 billion in the fourth. Takeaway relief narrows the basis and pulls Waha back toward Henry Hub, eroding exactly the discount that makes the fuel cheap.<\/p>\n

    Whether that matters to Chevron, to Microsoft or to National Grid depends on a fuel clause inside the power purchase agreement that no party has published. Chevron’s claim that the cash flows are independent of commodity price cycles is true under some plausible risk allocations and not under others, and the document that would settle it is not public. That is the single largest disclosure gap in the project and it should be stated as such rather than assumed away.<\/p>\n

    The counter case, fairly put<\/h2>\n

    The long run cost case favours the grid. Ascend Analytics notes that while behind the meter gas is practical when speed is the priority, a behind the meter gas only approach is suboptimal from a long term cost perspective, and levelised cost analysis strongly favours front of the meter grid connection over the long run. Chevron and Microsoft both concede the point implicitly, since both say they anticipate connecting to the grid over time. At that moment tariffs, standby charges and transmission cost allocation all become live.<\/p>\n

    Chevron is also taking on an operating discipline it has not run at this scale. It has built gas fired power for its own operations before. Merchant scale dispatchable generation for a hyperscaler under a twenty year firm obligation is a different business, and the structure adds joint venture governance across Chevron, Joulent and National Grid to a capital commitment of somewhere between 6 and 9 billion dollars against a contract price that has not been disclosed.<\/p>\n

    The permitting route is the slow one. Kilby triggered full prevention of significant deterioration review and greenhouse gas review under TCEQ permits 181895, PSDTX1684 and GHGPSDTX260, drawing a public meeting on 10 June 2026, with selective catalytic reduction as the named control for nitrogen oxides. That is more defensible than the alternative but also fully public and slower. The alternative is visibly riskier: in July 2026 the Environmental Integrity Project, Sierra Club and Public Citizen served a sixty day notice of intent to sue over two San Antonio campuses, alleging improper division into separate minor sources. New Mexico blocked a pipeline for a project with a 2.45 gigawatt onsite plant, and a Microsoft partner has struggled to obtain a 400 megawatt air permit in New Jersey on a deal where most compute tranches depend on the onsite plant.<\/p>\n

    There is a systemic objection too. The Harvard Belfer Center warns that co location can pull generation behind a private fence and reduce capacity available to the wider system. That applies most directly to redirected existing plant rather than to newbuild like Kilby, but the questions about backup service, standby charges and who pays for transmission remain open.<\/p>\n

    And the grid may catch up. ERCOT says the majority of queued interconnection requests expect to be operational by 2030, the same window as Kilby’s 2031 full capacity. Analysts covering Chevron’s expansion beyond Kilby note that it is still too early to determine the financial impact. Kilby’s advantage is strongest in precisely the years it is being built.<\/p>\n

    Analysis: six shifts when an oil company becomes a power developer<\/h2>\n

    This section is our analysis built on the sourced facts above. No third party research compares procurement behaviour across these buyers.<\/p>\n

    Procurement moves upstream of the project. The turbine slot was reserved in November 2025 and the customer contract signed in June 2026. If your lead time is measured against final investment decision you are already late. Sell to the reservation.<\/p>\n

    Specification is set by the air permit, not the press release. Kilby’s actual equipment mix, GE Vernova 7HA.02 frames alongside Solar Turbines T350 and T130 packages, is only knowable from TCEQ 181895. Air permits routinely name turbine models and capacities that announcements do not.<\/p>\n

    The counterparty is an upstream capital projects organisation, not a utility. Expect final investment decision gating, take or pay structures, joint venture approvals across three parties, and upstream commercial standards rather than regulated rate base process.<\/p>\n

    The timeline is phased across five years, not one date. Site work and fabrication now, modular generation from 2027, first power 2028, heavy frames 2029, full capacity 2031. That is four or five distinct procurement waves, and a capability statement built around a single delivery date misses most of them.<\/p>\n

    The scarce inputs are widening beyond turbines. GE Vernova currently captures roughly 300 million dollars of revenue per gigawatt of data centre capacity it serves, and Scott Strazik has said that could grow two to three times with solid state transformers and medium voltage uninterruptible power supply blocks, with orders expected as soon as 2027. Electrical balance of plant, brackish and produced water treatment, selective catalytic reduction, controls and operations and maintenance are all in scope.<\/p>\n

    Finally, size the market off permits rather than announcements. Around 60 per cent of announced behind the meter capacity exists only as an announcement, and only about 2 gigawatts was operating in mid 2026. Kilby is one of the credible ones, with a permit, a named investment grade offtaker, turbine slots and capital behind it. Pipeline sized off announced gigawatts will overstate the addressable market by roughly an order of magnitude.<\/p>\n

    FAQ<\/h2>\n

    Is Project Kilby connected to the Texas grid?<\/h3>\n

    Not at launch. It is designed as dedicated behind the meter generation serving the Microsoft campus directly. That is the point of the structure: ERCOT has confirmed that facilities which are truly islanded with no grid connection generally fall outside its interconnection process, although they may still face registration requirements with the Public Utility Commission of Texas. Both Chevron and Microsoft have indicated they expect a grid connection over time, at which point tariffs, standby charges and transmission cost allocation become relevant, and ERCOT’s curtailable large load framework would come into play.<\/p>\n

    How much is Project Kilby costing?<\/h3>\n

    Chevron has not disclosed it. Three figures circulate and none is a company disclosure. A Reeves County Emergency Services District tax abatement notice of March 2026 carries 6 billion dollars, which is the only figure from a filed document. Bloomberg reported 7 billion. TD Securities analyst Jason Gabelman modelled 9 billion by assuming most of the project is financed and back solving from an assumed 15 per cent developer internal rate of return, an exercise that also produced an implied contract price of roughly 150 dollars per megawatt hour. Chief financial officer Eimear Bonner has confirmed only that Kilby sits inside Chevron’s 2026 organic capital outlook of 18 to 19 billion dollars and its affiliate capital range.<\/p>\n

    Who actually owns Project Kilby?<\/h3>\n

    The developer is Energy Forge One LLC, a fifty fifty joint venture between Chevron and Joulent. Joulent is an energy venture spun out of Engine No. 1 and launched on 22 June 2026; National Grid Ventures, the commercial arm of National Grid, agreed on 1 July 2026 to invest 1.75 billion dollars for a 35 per cent stake, implying a valuation of roughly 5 billion dollars. Microsoft is the offtaker rather than an owner, though Microsoft states that the energy infrastructure required to power the datacenter is being funded by Microsoft.<\/p>\n

    Why do oil companies have an advantage in building data centre power?<\/h3>\n

    Two reasons, and only one of them is about power. The first is fuel: Permian associated gas is a disposal problem before it is a commodity, and Waha hub prices averaged minus 2.19 dollars per million British thermal units across 2026 to mid August against a five year average of plus 2.88. Converting that gas to contracted electricity at the wellhead captures the basis spread and creates in basin demand that supports the producer’s own netbacks. The second is capital projects capability: ExxonMobil’s Dan Ammann framed his company’s equivalent project as being driven by market demand, low carbon, available on an accelerated timeline and avoiding all the grid interconnection challenges. What oil companies do not have is experience operating merchant scale dispatchable generation for third party customers under multi decade firm obligations.<\/p>\n

    How much behind the meter gas capacity is actually being built?<\/h3>\n

    Far less than the announcements suggest. Cleanview identified 59 behind the meter data centre projects with roughly 90 gigawatts of announced capacity, more than a quarter of all planned United States data centre capacity, with 92 per cent of it announced since the start of 2025. But only about 2 gigawatts was operating in mid 2026 across four projects, with 1.2 per cent under construction, 36 per cent permitted and around 60 per cent announcement only. Cleanview expects 2.8 to 3.2 gigawatts operating by the end of 2026 and between 5 and 13 gigawatts by the end of 2027 depending on whether timelines hold. Separately, Global Energy Monitor puts total United States gas capacity in development at roughly 252 gigawatts in January 2026, with more than a third slated to power data centres on site, and Texas alone at 80.6 gigawatts of which nearly half is for data centres.<\/p>","protected":false},"excerpt":{"rendered":"

    Project Kilby will put roughly 2.67 gigawatts of dedicated gas generation next to a Microsoft data centre in Reeves County, Texas, under a twenty year take or pay agreement. The obvious question is why an oil company builds and operates a power station rather than letting the grid supply it. The ans<\/p>","protected":false},"author":12,"featured_media":4285,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"p54_article_data":"{\"meta\": {\"kicker\": \"Answer \u00b7 Power & Data Centres\", \"topics\": [\"Power & Data Centres\", \"Corporate Strategy\"], \"title\": \"Why Chevron Is Building a Power Plant Instead of Buying Grid Power\", \"dek\": \"Project Kilby will put roughly 2.67 gigawatts of dedicated gas generation next to a Microsoft data centre in Reeves County, Texas, under a twenty year take or pay agreement. The obvious question is why an oil company builds and operates a power station rather than letting the grid supply it. The answer is a queue, a turbine order book and a negative gas price, and only one of the three is about electricity.\", \"date\": \"27 September 2026\", \"readTime\": \"14 min read\", \"author\": \"Project 54, Research & Strategy\"}, \"quickAnswer\": {\"q\": \"Why is Chevron building its own power plant for Microsoft instead of supplying it from the grid?\", \"a\": \"Because a grid connection was not available in the time Microsoft needed it, and because the inputs Chevron already controls are the scarce ones. ERCOT is tracking more than 438 gigawatts of large load interconnection requests, close to 90 per cent of them data centres, and its Batch Zero reform will not produce a final transmission plan until autumn 2027. Lawrence Berkeley National Laboratory found the median time from interconnection request to commercial operation exceeded five years for generation projects completed in 2025. ERCOT has confirmed that facilities which are truly islanded with no grid connection generally fall outside its interconnection process, so Kilby does not shorten the queue, it exits it. Chevron had also reserved seven GE Vernova 7HA turbine slots in November 2025, before the Microsoft agreement was signed, in a market where the three dominant manufacturers are sold out into the 2030s. And the fuel is Permian associated gas that traded at an average of minus 2.19 dollars per million British thermal units at the Waha hub across 2026 to mid August.\"}, \"takeaways\": [\"The decision is about time, not cost. Ascend Analytics notes that levelised cost analysis still favours grid connection over the long run, and both Chevron and Microsoft say they expect to connect to the grid eventually. Building is what you do when the schedule is the constraint.\", \"The turbine slot is the real moat. Chevron reserved seven GE Vernova 7HA turbines in November 2025, before the Microsoft deal existed. GE Vernova held 116 gigawatts of gas equipment under backlog and slot reservations at the end of the second quarter of 2026 and is largely sold out through 2030.\", \"The fuel advantage is a distressed asset. Waha hub gas averaged minus 2.19 dollars per million British thermal units across 2026 to mid August, against a five year average of plus 2.88, and printed negative on 99 days including a 90 day consecutive run.\", \"Microsoft is paying for the plant, not just the power. Microsoft states that the energy infrastructure required to power this datacenter is being funded by Microsoft, and Chevron describes the arrangement as a twenty year take or pay agreement for firm behind the meter capacity.\", \"The capital number has never been disclosed. Filed and reported figures span 6 billion dollars in a Reeves County tax abatement notice, 7 billion from Bloomberg, and 9 billion back solved by a TD Securities analyst from an assumed 15 per cent developer return. Chevron has published none of them.\", \"The moat closes while the plant is being built. Around 4.5 billion cubic feet a day of new Permian takeaway arrives around final investment decision, narrowing the discount that makes the fuel cheap, and ERCOT expects the majority of its queued requests to be operational by 2030.\"], \"sections\": [{\"id\": \"what\", \"q\": \"What is Project Kilby, precisely?\", \"h\": \"The facts, and where the sources disagree\", \"p\": [\"Project Kilby is a behind the meter gas fired generation complex on more than 2,000 acres in Reeves County, Texas, roughly twenty miles south of Pecos. It is being developed by Energy Forge One LLC, a fifty fifty joint venture between Chevron and Joulent, an energy venture spun out of Engine No. 1 and launched on 22 June 2026. National Grid Ventures agreed on 1 July 2026 to invest 1.75 billion dollars for a 35 per cent stake in Joulent, implying a valuation of roughly 5 billion dollars.\", \"The customer is Microsoft, which announced a data centre of approximately 2 gigawatts at Pecos on 22 June 2026, describing a multibillion dollar investment over five to seven years and more than 6,000 peak construction jobs. Chevron characterises the commercial arrangement as a twenty year take or pay power purchase agreement for 2.67 gigawatts of firm behind the meter capacity, targeting mid teens returns and long duration contracted cash flows that are independent of commodity price cycles. First power is targeted for 2028 and full capacity for 2031, with a final investment decision expected by the end of 2026.\", \"Two things are genuinely unsettled and should not be reported as though they are not. The first is capacity. Four figures circulate because they describe different instruments: 2.67 gigawatts is Chevron's contracted figure, about 2.5 gigawatts appears in stale page metadata and earlier Bloomberg reporting, roughly 2,595 megawatts is the figure in the TCEQ air permit application, and roughly 2,868 megawatts is the arithmetic sum of the unit level entries in the permit record. Use 2.67 gigawatts and say it is the contracted number.\", \"The second is capital cost, which Chevron has never disclosed. A Reeves County Emergency Services District tax abatement notice of March 2026 carries 6 billion dollars, the only figure from a filed document. Bloomberg reported 7 billion. TD Securities analyst Jason Gabelman modelled 9 billion by assuming most of the project is financed and back solving from a 15 per cent developer internal rate of return, which also produced an implied contract price of about 150 dollars per megawatt hour. Quoting that price alongside Chevron's mid teens return guidance prints the same assumption twice. Chief financial officer Eimear Bonner confirmed only that Kilby sits inside the 2026 organic capital outlook of 18 to 19 billion dollars.\", \"The equipment mix is knowable from the TCEQ permit record and it is more interesting than the headline. Two combined cycle units of 555.7 megawatts each use GE Vernova 7HA.02 machines for 2028, three further 7HA.02 units total 990 megawatts for 2029, and the fast tranches are modular: twelve Solar Turbines T350 units totalling 373.2 megawatts for 2027 and twenty nine Solar Turbines T130 units totalling 392.95 megawatts for 2028. Solar Turbines is a Caterpillar subsidiary. The plant is phased deliberately so that modular capacity arrives years before the heavy frames.\"]}, {\"id\": \"queue\", \"q\": \"Could Chevron simply have waited for a grid connection?\", \"h\": \"The queue is the answer, and ERCOT said so\", \"p\": [\"The decisive regulatory fact was published by Utility Dive on 22 June 2026, the same day the Chevron and Microsoft announcement landed, quoting ERCOT directly: facilities that are truly islanded with no grid connection would generally fall outside ERCOT's interconnection process, though they may still be subject to registration requirements with the PUCT.\", \"That is the whole trade in one sentence. Kilby does not get to the front of the queue. It leaves the queue.\", \"The scale of what it is leaving is the part people underestimate. On the generation side, Lawrence Berkeley National Laboratory's Queued Up 2026 edition, using data through the end of 2025, found around 2,060 gigawatts of generation and storage actively seeking interconnection across roughly 8,200 projects. Active natural gas capacity in the queues stood at 253 gigawatts, up 86 per cent in 2025 and the only category that grew. The median duration from interconnection request to commercial operation exceeded five years for projects built in 2025, and of all the capacity that entered queues between 2000 and 2020, only 13 per cent had reached commercial operation by the end of 2025 while 75 per cent was withdrawn.\", \"On the load side, ERCOT is tracking more than 438 gigawatts of large load interconnection requests, nearly 90 per cent of them data centres. Governor Abbott cited roughly 474 gigawatts when he directed the PUCT and ERCOT on 3 August 2026 to audit every data centre project in the process. ERCOT's Batch Zero reform, approved by the PUCT on 18 June 2026, groups loads of 75 megawatts or more into coordinated study batches, but the final transmission plan for Batch Zero is not expected until autumn 2027 and Batch 1 applications do not open until summer 2027. A grid routed campus would still be waiting for a plan in late 2027 against Kilby's 2028 first power.\", \"The commercial value of that difference is large enough to justify almost any capital structure. Cleanview estimates that AI data centres can generate 10 to 12 million dollars per megawatt annually, or 10 to 12 billion dollars per gigawatt, and that bringing a data centre online even a few years early could result in tens of billions in revenue that would otherwise be forgone. At 2 gigawatts, a two year acceleration is worth more than the entire plant. That is why Microsoft is funding the generation rather than buying the electricity.\", \"One honest qualification. Behind the meter does not remove queues, it substitutes them: air permits, gas connection, equipment reservation, storage and controls, engineering and construction capacity, and financing. Cleanview's own data bears this out. Of roughly 90 gigawatts of announced behind the meter capacity it identified across 59 projects, only about 2 gigawatts, 2.2 per cent, was actually operating in mid 2026.\"]}, {\"id\": \"turbines\", \"q\": \"What is the scarce input, if not electricity?\", \"h\": \"Turbine slots, reserved a year early\", \"p\": [\"In November 2025, before any Microsoft agreement existed, GE Vernova, Chevron and Engine No. 1 announced a venture targeting up to 4 gigawatts for United States data centres, with the first power foundries to use seven GE Vernova 7HA turbines secured under a slot reservation agreement. That reservation, not the Permian acreage and not the customer relationship, is the asset that made Kilby possible on this timeline.\", \"The reason is that the manufacturers are sold out. GE Vernova reported 116 gigawatts of gas equipment under backlog and slot reservations at the end of the second quarter of 2026, comprising 53 gigawatts of firm backlog and 63 gigawatts of slot reservations, up from 100 gigawatts three months earlier, with Power segment orders of 16.7 billion dollars, up 134 per cent organically. Chief executive Scott Strazik said the company expects at least 125 gigawatts of gas equipment under contract by the end of 2026, and that it remains on track to deliver 20 gigawatts of annual gas turbine output in the third quarter of 2026, 24 gigawatts in 2028, with actions under way to produce 30 gigawatts in 2030. Customer discussions, he said, already stretch out to 2032 and beyond.\", \"The picture is the same elsewhere. Siemens Energy reported 69 gigawatts of gas turbine backlog plus 26 gigawatts under slot reservation agreements in its August 2026 quarterly presentation. Mitsubishi Power's head of EMEA told Platts that lead times have stretched to five years or more and that the company is signing contracts for deliveries between 2031 and 2034, describing the bottleneck as whole supply chain capacity rather than demand or capital.\", \"Global Energy Monitor's January 2026 assessment puts a number on how binding this is: those three manufacturers account for more than 75 per cent of the market for gas power in development with a named manufacturer, and two thirds of projects in development have no named manufacturer at all. In other words, most announced capacity has no machine behind it. A slot reservation is the one thing a power purchase agreement cannot buy you.\", \"This is also why the modular tranches matter. Enverus Intelligence Research expects reciprocating engines, small and medium frame turbines and fuel cells to make up 61 per cent of gas fired behind the meter capacity through 2030, on commercial operation timelines of 18 to 24 months against up to 80 months for large frame equipment. Kilby's Solar Turbines units arrive from 2027, the heavy GE frames from 2028 and 2029. The project is structured to deliver revenue before the big machines land.\"]}, {\"id\": \"gas\", \"q\": \"Where does the fuel advantage come from?\", \"h\": \"Gas that costs less than nothing\", \"p\": [\"Chevron's stated edge is leveraging Permian natural gas, and the economics of that phrase are unusual. Every Permian oil barrel brings associated gas with it. Producers who cannot flare within regulatory limits and cannot store it face a binary choice: pay to dispose of the gas, or shut in the oil well. Most pay.\", \"The result is a hub price that goes below zero. Reuters reported that Waha hub prices stayed negative for a record 47 consecutive days as of 14 April 2026. Across 2026 to mid August, Waha cash averaged minus 2.19 dollars per million British thermal units against a five year average of plus 2.88, printing negative on 99 days including a consecutive run of 90 days that broke in mid June. Prompt Waha to Henry Hub basis moved from below minus 5.00 dollars in the spring to minus 0.91 by late July. East Daley estimated well curtailments on Delaware basin gathering and processing systems averaging 220 million cubic feet a day in 2026, more than double its earlier forecast of 100.\", \"A plant that converts that gas into contracted electricity at the wellhead captures the whole basis spread and simultaneously creates an in basin demand sink that supports Chevron's own upstream netbacks. A grid power purchase agreement would deliver Chevron none of that. This is the part of the deal that is genuinely specific to an oil company and cannot be replicated by an independent power producer.\", \"It is also the part with a clock on it. Roughly 4.5 billion cubic feet a day of new Permian takeaway arrives around the time of final investment decision, with Blackcomb at 2.5 billion cubic feet a day from Waha to Agua Dulce in the third quarter of 2026 and Hugh Brinson Phase 1 at 1.5 billion in the fourth. Takeaway relief narrows the basis and pulls Waha back toward Henry Hub, eroding exactly the discount that makes the fuel cheap.\", \"Whether that matters to Chevron, to Microsoft or to National Grid depends on a fuel clause inside the power purchase agreement that no party has published. Chevron's claim that the cash flows are independent of commodity price cycles is true under some plausible risk allocations and not under others, and the document that would settle it is not public. That is the single largest disclosure gap in the project and it should be stated as such rather than assumed away.\"]}, {\"id\": \"against\", \"q\": \"What is Chevron giving up, and who disagrees?\", \"h\": \"The counter case, fairly put\", \"p\": [\"The long run cost case favours the grid. Ascend Analytics notes that while behind the meter gas is practical when speed is the priority, a behind the meter gas only approach is suboptimal from a long term cost perspective, and levelised cost analysis strongly favours front of the meter grid connection over the long run. Chevron and Microsoft both concede the point implicitly, since both say they anticipate connecting to the grid over time. At that moment tariffs, standby charges and transmission cost allocation all become live.\", \"Chevron is also taking on an operating discipline it has not run at this scale. It has built gas fired power for its own operations before. Merchant scale dispatchable generation for a hyperscaler under a twenty year firm obligation is a different business, and the structure adds joint venture governance across Chevron, Joulent and National Grid to a capital commitment of somewhere between 6 and 9 billion dollars against a contract price that has not been disclosed.\", \"The permitting route is the slow one. Kilby triggered full prevention of significant deterioration review and greenhouse gas review under TCEQ permits 181895, PSDTX1684 and GHGPSDTX260, drawing a public meeting on 10 June 2026, with selective catalytic reduction as the named control for nitrogen oxides. That is more defensible than the alternative but also fully public and slower. The alternative is visibly riskier: in July 2026 the Environmental Integrity Project, Sierra Club and Public Citizen served a sixty day notice of intent to sue over two San Antonio campuses, alleging improper division into separate minor sources. New Mexico blocked a pipeline for a project with a 2.45 gigawatt onsite plant, and a Microsoft partner has struggled to obtain a 400 megawatt air permit in New Jersey on a deal where most compute tranches depend on the onsite plant.\", \"There is a systemic objection too. The Harvard Belfer Center warns that co location can pull generation behind a private fence and reduce capacity available to the wider system. That applies most directly to redirected existing plant rather than to newbuild like Kilby, but the questions about backup service, standby charges and who pays for transmission remain open.\", \"And the grid may catch up. ERCOT says the majority of queued interconnection requests expect to be operational by 2030, the same window as Kilby's 2031 full capacity. Analysts covering Chevron's expansion beyond Kilby note that it is still too early to determine the financial impact. Kilby's advantage is strongest in precisely the years it is being built.\"]}, {\"id\": \"suppliers\", \"q\": \"What does this change for suppliers?\", \"h\": \"Analysis: six shifts when an oil company becomes a power developer\", \"p\": [\"This section is our analysis built on the sourced facts above. No third party research compares procurement behaviour across these buyers.\", \"Procurement moves upstream of the project. The turbine slot was reserved in November 2025 and the customer contract signed in June 2026. If your lead time is measured against final investment decision you are already late. Sell to the reservation.\", \"Specification is set by the air permit, not the press release. Kilby's actual equipment mix, GE Vernova 7HA.02 frames alongside Solar Turbines T350 and T130 packages, is only knowable from TCEQ 181895. Air permits routinely name turbine models and capacities that announcements do not.\", \"The counterparty is an upstream capital projects organisation, not a utility. Expect final investment decision gating, take or pay structures, joint venture approvals across three parties, and upstream commercial standards rather than regulated rate base process.\", \"The timeline is phased across five years, not one date. Site work and fabrication now, modular generation from 2027, first power 2028, heavy frames 2029, full capacity 2031. That is four or five distinct procurement waves, and a capability statement built around a single delivery date misses most of them.\", \"The scarce inputs are widening beyond turbines. GE Vernova currently captures roughly 300 million dollars of revenue per gigawatt of data centre capacity it serves, and Scott Strazik has said that could grow two to three times with solid state transformers and medium voltage uninterruptible power supply blocks, with orders expected as soon as 2027. Electrical balance of plant, brackish and produced water treatment, selective catalytic reduction, controls and operations and maintenance are all in scope.\", \"Finally, size the market off permits rather than announcements. Around 60 per cent of announced behind the meter capacity exists only as an announcement, and only about 2 gigawatts was operating in mid 2026. Kilby is one of the credible ones, with a permit, a named investment grade offtaker, turbine slots and capital behind it. Pipeline sized off announced gigawatts will overstate the addressable market by roughly an order of magnitude.\"]}], \"media\": {\"image\": {\"src\": \"\/wp-content\/uploads\/2026\/09\/chevron-behind-the-meter-gas-processing-plant-pipework.jpg\", \"label\": \"Permian associated gas averaged minus 2.19 dollars per million British thermal units at Waha across 2026 to mid August. Kilby turns a disposal cost into contracted electricity.\", \"credit\": \"Project 54\"}, \"infographicLabel\": \"More than 438 gigawatts of large load requests in the ERCOT queue, a median of over five years from request to operation, and three turbine makers sold out into the 2030s.\", \"pdf\": {\"href\": \"https:\/\/projectfifty4.com\/wp-content\/uploads\/2026\/09\/chevron-power-plant-instead-of-ppa.pdf\", \"title\": \"Why Chevron Is Building a Power Plant Instead of Buying Grid Power\", \"meta\": \"PDF, 13 slides\"}, \"podcast\": {\"src\": \"https:\/\/projectfifty4.com\/wp-content\/uploads\/2026\/09\/chevron-power-plant-instead-of-ppa-podcast.m4a\", \"title\": \"Chevron Project Kilby Bypasses the Texas Grid\", \"ep\": \"P54 Energy Growth Brief\", \"duration\": \"23:27\"}}, \"poll\": {\"q\": \"What is the real constraint behind the meter projects are solving for?\", \"options\": [{\"id\": \"a\", \"label\": \"Grid interconnection time\", \"insight\": \"The one everyone names, and it is real: Lawrence Berkeley National Laboratory found a median of over five years from interconnection request to commercial operation for projects built in 2025, and ERCOT's own Batch Zero plan does not land until autumn 2027. But islanding replaces one queue with several.\"}, {\"id\": \"b\", \"label\": \"Gas turbine availability\", \"insight\": \"The most underrated. GE Vernova held 116 gigawatts under backlog and slot reservations in mid 2026 and is largely sold out through 2030; Mitsubishi is signing for 2031 to 2034. Global Energy Monitor found two thirds of projects in development have no named manufacturer at all.\"}, {\"id\": \"c\", \"label\": \"Cheap fuel\", \"insight\": \"Decisive for an oil company and almost irrelevant to anyone else. Waha averaged minus 2.19 dollars per million British thermal units across 2026 to mid August. But roughly 4.5 billion cubic feet a day of new takeaway will narrow that basis during construction.\"}, {\"id\": \"d\", \"label\": \"Permitting and community consent\", \"insight\": \"The quiet one that kills projects. Kilby went the slow route through full prevention of significant deterioration and greenhouse gas review. Others that took the minor source route are facing litigation notices and blocked pipelines.\"}], \"note\": \"Responses are anonymous and are used to shape future Project 54 research.\"}, \"faq\": [{\"q\": \"Is Project Kilby connected to the Texas grid?\", \"a\": \"Not at launch. It is designed as dedicated behind the meter generation serving the Microsoft campus directly. That is the point of the structure: ERCOT has confirmed that facilities which are truly islanded with no grid connection generally fall outside its interconnection process, although they may still face registration requirements with the Public Utility Commission of Texas. Both Chevron and Microsoft have indicated they expect a grid connection over time, at which point tariffs, standby charges and transmission cost allocation become relevant, and ERCOT's curtailable large load framework would come into play.\"}, {\"q\": \"How much is Project Kilby costing?\", \"a\": \"Chevron has not disclosed it. Three figures circulate and none is a company disclosure. A Reeves County Emergency Services District tax abatement notice of March 2026 carries 6 billion dollars, which is the only figure from a filed document. Bloomberg reported 7 billion. TD Securities analyst Jason Gabelman modelled 9 billion by assuming most of the project is financed and back solving from an assumed 15 per cent developer internal rate of return, an exercise that also produced an implied contract price of roughly 150 dollars per megawatt hour. Chief financial officer Eimear Bonner has confirmed only that Kilby sits inside Chevron's 2026 organic capital outlook of 18 to 19 billion dollars and its affiliate capital range.\"}, {\"q\": \"Who actually owns Project Kilby?\", \"a\": \"The developer is Energy Forge One LLC, a fifty fifty joint venture between Chevron and Joulent. Joulent is an energy venture spun out of Engine No. 1 and launched on 22 June 2026; National Grid Ventures, the commercial arm of National Grid, agreed on 1 July 2026 to invest 1.75 billion dollars for a 35 per cent stake, implying a valuation of roughly 5 billion dollars. Microsoft is the offtaker rather than an owner, though Microsoft states that the energy infrastructure required to power the datacenter is being funded by Microsoft.\"}, {\"q\": \"Why do oil companies have an advantage in building data centre power?\", \"a\": \"Two reasons, and only one of them is about power. The first is fuel: Permian associated gas is a disposal problem before it is a commodity, and Waha hub prices averaged minus 2.19 dollars per million British thermal units across 2026 to mid August against a five year average of plus 2.88. Converting that gas to contracted electricity at the wellhead captures the basis spread and creates in basin demand that supports the producer's own netbacks. The second is capital projects capability: ExxonMobil's Dan Ammann framed his company's equivalent project as being driven by market demand, low carbon, available on an accelerated timeline and avoiding all the grid interconnection challenges. What oil companies do not have is experience operating merchant scale dispatchable generation for third party customers under multi decade firm obligations.\"}, {\"q\": \"How much behind the meter gas capacity is actually being built?\", \"a\": \"Far less than the announcements suggest. Cleanview identified 59 behind the meter data centre projects with roughly 90 gigawatts of announced capacity, more than a quarter of all planned United States data centre capacity, with 92 per cent of it announced since the start of 2025. But only about 2 gigawatts was operating in mid 2026 across four projects, with 1.2 per cent under construction, 36 per cent permitted and around 60 per cent announcement only. Cleanview expects 2.8 to 3.2 gigawatts operating by the end of 2026 and between 5 and 13 gigawatts by the end of 2027 depending on whether timelines hold. Separately, Global Energy Monitor puts total United States gas capacity in development at roughly 252 gigawatts in January 2026, with more than a third slated to power data centres on site, and Texas alone at 80.6 gigawatts of which nearly half is for data centres.\"}], \"related\": [{\"title\": \"Chevron, Microsoft and Project Kilby\", \"topic\": \"Power & Data Centres\", \"href\": \"https:\/\/projectfifty4.com\/chevron-microsoft-project-kilby\/\"}, {\"title\": \"What Is Behind the Meter Power?\", \"topic\": \"Power & Data Centres\", \"href\": \"https:\/\/projectfifty4.com\/what-is-behind-the-meter-power\/\"}, {\"title\": \"Is Chevron Becoming a Utility?\", \"topic\": \"Corporate Strategy\", \"href\": \"https:\/\/projectfifty4.com\/is-chevron-becoming-a-utility\/\"}, {\"title\": \"Why Data Centers Drive Gas Turbine Demand\", \"topic\": \"Power & Data Centres\", \"href\": \"https:\/\/projectfifty4.com\/why-data-centers-drive-gas-turbine-demand\/\"}, {\"title\": \"Baker Hughes Data Center Power Strategy 2026\", \"topic\": \"Oilfield Services\", \"href\": \"https:\/\/projectfifty4.com\/baker-hughes-data-center-power-strategy-2026\/\"}, {\"title\": \"Chevron and Hess Advantaged Portfolio Strategy 2026\", \"topic\": \"Corporate Strategy\", \"href\": \"https:\/\/projectfifty4.com\/chevron-hess-advantaged-portfolio-strategy-2026\/\"}], \"newsletter\": {\"kicker\": \"The Energy Growth Brief\", \"title\": [\"Intelligence for energy\", \"growth leaders\"], \"body\": \"Join energy and industrial leaders getting our marketing, AI-growth and revenue-architecture intelligence, direct to your inbox.\", \"placeholder\": \"you@company.com\", \"cta\": \"Subscribe\", \"note\": \"No spam. Unsubscribe anytime. We read every reply.\"}, \"listenTime\": \"23 min listen\"}","p54_faq":"","p54_media":"","p54_comments_enabled":"","footnotes":""},"categories":[92,125],"tags":[],"class_list":["post-4291","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-analysis","category-strategy"],"acf":[],"_links":{"self":[{"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/posts\/4291","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/users\/12"}],"replies":[{"embeddable":true,"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/comments?post=4291"}],"version-history":[{"count":1,"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/posts\/4291\/revisions"}],"predecessor-version":[{"id":4299,"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/posts\/4291\/revisions\/4299"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/media\/4285"}],"wp:attachment":[{"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/media?parent=4291"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/categories?post=4291"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/projectfifty4.com\/fr\/wp-json\/wp\/v2\/tags?post=4291"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}