Vantage Data Centers’ plan to build “Frontier,” a $25B, 1.4GW hyperscale AI campus in Shackelford County, Texas is a clear signal that the AI compute build‑out is migrating toward mega‑sites engineered for 250kW+ rack densities, liquid cooling, and industrial‑scale power interconnects. The company says the 1,200‑acre campus will host 10 data centers totaling 3.7M sq ft, with construction underway and first delivery targeted for 2H26, and it expects the program to employ 5,000+ people across construction and operations. The project’s stated use of liquid cooling to support next‑gen GPU loads and its density targets are consistent with the shift from air‑cooled 10‑30kW racks to AI racks that commonly exceed 100‑250kW, which is increasingly mainstream across hyperscale roadmaps.
The siting choice is strategic. Shackelford County sits near ERCOT’s 345kV Competitive Renewable Energy Zone transmission corridors, including the West Shackelford 345kV substation built under Texas’ CREZ program, which gives Vantage a path to pull bulk power if interconnection is granted. Texas has become the most aggressive growth market for large flexible loads, with ERCOT documenting a surge of large‑load interconnection requests and a dedicated interconnection process for 10MW+ loads; ERCOT’s queue and planning materials show data centers as a central driver of forecast load growth through 2030. The state has simultaneously moved to regulate reliability impacts from these loads, including new authority to curtail or disconnect data centers during grid emergencies, which materially changes the risk calculus for developers and tenants.
In practical terms, 1.4GW of campus capacity is comparable to output from 1 large nuclear reactor, underscoring the magnitude of the program. If Vantage ultimately fills 1.4GW primarily with AI clusters built around Nvidia GB200 NVL72‑class systems that require roughly 120kW per rack, the site could support on the order of 10,000‑12,000 such racks at full build, depending on PUE and non‑IT loads, which would imply hundreds of thousands of GPUs on a single campus. While exact tenant mix is unknown, the density profile and time‑to‑power constraints argue for 1‑2 anchor hyperscalers absorbing large contiguous blocks.
Texas’ policy and incentive stack materially improves the project’s economics. The state’s Qualified Data Center program offers a 10‑15 year state sales‑tax exemption on eligible IT and infrastructure purchases for qualifying facilities that invest $200M‑$250M+, and the newer Chapter 403 “JETI” program provides property‑tax valuation reductions via school district agreements for eligible large capital projects. These benefits, layered on low land costs and scalable transmission, are a major reason AI campuses are clustering in Texas rather than higher‑cost coastal metros.
The risk side is equally clear. The newly enacted state rules enabling ERCOT to curtail data centers during emergencies will push developers to secure redundancy beyond traditional diesel backup. Vantage has already signaled a strategy to deploy prime‑power microgrids at scale via a partnership with VoltaGrid, targeting more than 1GW of gas‑fired on‑site generation across new campuses. This reduces interconnection timing risk and outage exposure but adds fuel supply, emissions permitting, and capex complexity; it also shifts the vendor stack toward large gas engines and turbines, high‑voltage switchgear, and grid‑forming controls.
Translating the plan into dollars, Vantage is telegraphing a capex intensity of roughly $17.9M per MW when dividing $25B by 1,400MW. That multiple is elevated vs pre‑AI builds but directionally consistent with high‑density, liquid‑cooled campuses with heavy‑duty electrical, mechanical, and cooling plants. On revenue potential, wholesale AI colocation pricing has risen toward or above $200 per kW‑month for 250‑500kW blocks in primary markets as vacancy tightened; scenarios at 70%‑90% campus utilization imply $2.4B‑$3.4B of annual recurring revenue before power pass‑throughs at $200‑$225 per kW‑month. Actual realized pricing will vary by lease tenor, power procurement structures, and whether tenants take powered shell vs turnkey. Preleasing across North America remains unusually high, and pricing trends remain constructive due to constrained near‑term power availability.
Financing capacity does not appear to be a gating factor for Vantage. The company secured more than $13B of incremental debt and equity in 2024 and early 2025 and previously closed a $9.2B equity round led by DigitalBridge and Silver Lake, with investors including AustralianSuper. Debt club deals and green‑loan structures have been used to fully fund other North American campuses. Given typical data‑center development financing at 65%‑80% loan‑to‑cost, the sponsor group’s access to both private credit and infrastructure LP capital is a competitive advantage.
The most immediate public‑market beneficiaries sit in the power and thermal stack. Vertiv has reported accelerating orders, record backlog, and raised guidance on AI‑driven demand for power distribution, switchgear integration, and liquid cooling, precisely the kit a 250kW+ rack campus consumes in volume. Modine has publicly disclosed large AI cooling awards and is investing to expand capacity. Johnson Controls highlighted data‑center revenue that doubled year over year with hyperscaler capex as a tailwind. These companies provide CDUs, rear‑door heat exchangers, dry coolers, heat‑rejection systems, and controls that liquid‑cooled campuses standardize on.
The high‑voltage electrical balance‑of‑plant remains a chokepoint and a second‑order winner set. Powell Industries has reported record backlog with improved visibility, and its specialty in MV/LV switchgear and e‑houses maps tightly to mega‑campus needs; SPX Technologies has emphasized strong transformer demand and data‑center‑oriented cooling solutions, and transformer supply remains structurally tight. Lead times for transformers, switchgear, and gensets are still a pacing item for delivery schedules, which gives pricing power to best‑positioned suppliers and supports sustained margin strength through 2026‑2027.
Backup and bridge‑power vendors also screen positively. Cummins reported power‑systems growth driven by data centers, and Caterpillar’s energy and transportation segment is increasingly levered to data‑center generator demand. If Vantage deploys gas‑fired prime power to mitigate curtailment risk, it not only lifts the genset and controls vendors but also pulls in gas‑handling, emissions, and black‑start systems integrators. This dynamic also creates incremental optionality for midstream operators with West Texas footprints that can deliver firm gas supply to on‑site assets.
Grid and transmission EPCs should continue to enjoy backlog expansion. Quanta Services has been explicit about Texas and data‑center load growth underpinning multi‑year utility capex, and MYR Group has shown accelerating T&D revenues with favorable book‑to‑bill, while MasTec and Primoris have signaled active pursuit of data‑center infrastructure work across power delivery and heavy civil. A 1.4GW campus requires new or expanded substations, dedicated 345kV feeders, and substantial HV buildouts, all of which feed multi‑year programs for these contractors.
Network vendors and optical suppliers are a quieter but material beneficiary set. AI clusters at this scale drive demand for 800G‑1.6T optics, high‑radix switching, and new dark‑fiber routes. Arista, Broadcom, and Coherent have all pointed to AI‑driven step‑ups in 800G and future 1.6T shipments, while fiber carriers such as Zayo and Lumen are actively expanding long‑haul and metro capacity into AI corridors. A rural mega‑site west of DFW will compel new diverse fiber laterals and likely metro rings to Dallas and beyond.
The most credible risks cluster around power, water, and regulatory execution. ERCOT has codified the ability to curtail large loads, which will make firm, redundant power a non‑negotiable contractual term; on‑site generation reduces but does not eliminate exposure to fuel supply and policy changes. Water consumption for AI data centers is under intense scrutiny statewide as Texas cycles through drought; while liquid cooling can be deployed in closed‑loop configurations that minimize consumption, heat rejection often still requires evaporative systems without alternative designs, and a number of Texas jurisdictions have imposed water‑use restrictions during 2025. Stakeholder pressure will intensify if local reservoirs tighten.
Execution risk also includes supply‑chain constraints for transformers and switchgear, competition for skilled labor in West Texas, and schedule risk tied to interconnection studies under ERCOT’s large‑load process. Finally, macro demand risk exists: if AI infrastructure spending rephases because of model‑efficiency gains or changes in GPU roadmaps, lease signings could stagger. That said, near‑term supply remains the binding constraint across U.S. primary markets, with record preleasing and rental‑rate support.
Portfolio implications. First, treat Frontier as confirmatory evidence that AI campuses will concentrate in a handful of power‑rich U.S. geographies where incentives and 345kV access intersect. This supports a multi‑year overweight to the high‑voltage electrical and thermal stack: Vertiv for liquid cooling and power distribution, Powell Industries for switchgear/e‑houses, SPX Technologies for transformers and heat‑exchanger systems, and Modine for data‑center cooling. Cummins and Caterpillar screen as durable beneficiaries of both traditional backup and prime‑power needs. Quanta, MYR, Primoris, and MasTec benefit from the surrounding utility and campus buildouts that must precede IT commissioning.
Second, recognize the capital‑formation angle. DigitalBridge, which leads the Vantage investor group, should continue to scale fee‑earning AUM and deployment pace as hyperscaler demand outstrips power availability. DBRG’s ability to syndicate mega‑projects and to raise hybrid credit/infra capital at scale is a competitive differentiator in an environment where project tickets are climbing into multi‑billion‑dollar tranches.
Third, neutral to slightly positive read‑throughs for the public data‑center REITs. Digital Realty and Equinix compete with Vantage in wholesale and hyperscale segments, but the market is severely power‑constrained, vacancy in primary U.S. markets is near record lows, and pricing is rising. The supply response from private platforms like Vantage, QTS, and others extends the AI up‑cycle rather than cannibalizing near‑term demand for the REITs’ differentiated interconnection, retail colo, and metro campuses; the main caveat is longer‑term return pressure if cost of power and SB‑6‑style curtailment frameworks force higher capex per MW or lower availability SLAs.
Fourth, stress‑test the bear case. If ERCOT’s reliability challenges intensify and curtailment events occur during peak seasons, tenants will demand costlier redundancy packages or push builds toward self‑powered sites, raising delivered $/MW and extending timelines. Water constraints could further complicate local permits unless designs emphasize water‑sipping closed‑loop or air‑side heat rejection. Transformer and switchgear bottlenecks could shift first‑power dates to the right. However, policy and private‑sector responses are already forming, including Texas’ new funds to accelerate firm generation, evolving large‑load interconnection frameworks, and the proliferation of on‑site power partnerships.
Bottom line, Frontier’s 1.4GW scale, density targets, and liquid‑cooling design are in line with where AI infrastructure is actually going, not where legacy data‑center templates have been. The immediate investable conclusions are to stay overweight power and thermal suppliers with credible liquid‑cooling portfolios and HV gear backlogs, own select EPCs tied to transmission and substation work in ERCOT, maintain exposure to generator vendors leveraged to both backup and prime‑power, and consider sponsors with advantaged capital formation. The main diligence items we will track are interconnection milestones and substation scope in Shackelford County, on‑site power architecture and fuel contracting, preleasing cadence versus 2H26 delivery, and any local water‑use mitigation commitments. If the campus executes on time and on budget, the medium‑term revenue run‑rate potential is measured in low‑single‑digit billions at reasonable utilization and market rents, and the beneficiary set across the public markets is broad and durable.