Goldman Sachs shares were little changed near $892 as the bank moved to co-lead financing for a 5-gigawatt buildout of private power campuses in Texas designed for artificial intelligence workloads. The deal marks one of the largest Wall Street bets yet on dedicated electricity for data centers, a market racing to secure firm, around-the-clock power outside the clogged grid queue. The size is eye-catching: 5 gigawatts is on par with several nuclear units and signals that private power for AI is shifting from pilot to industrial scale.
The financing gives developers a way to stand up “power islands” for hyperscalers and AI startups that cannot wait years for grid interconnections. Instead of relying on traditional utility procurement, these campuses bundle onsite generation, transmission, and data hall buildouts under one umbrella, often with long-term capacity contracts from tenants. For banks, the structures look familiar: asset-backed loans, project finance tranches, and revenue visibility tied to blue-chip offtakers. For AI players, it is a speed-to-energize calculation. Latency is money, and the fastest route today is often building your own power.
Texas is where this strategy is scaling. In recent weeks, capital has flooded into the state’s AI power stack: a $5 billion financing package at VoltaGrid to accelerate more than 4.3 GW through 2028, Blackstone’s more than $500 million into Lancium to build over 5 GW of sites in West Texas, and Alphabet’s reported $40 billion commitment to three Texas data centers. The logic is simple. Texas offers land, access to abundant wind and solar, and a market design that rewards flexible generation. Developers can site near fuel, transmission corridors, and fiber, then stitch together the power and compute blocks AI tenants demand.
For Goldman Sachs (GS), the fee pool spans origination, underwriting, syndication, and hedging. These deals come with commodity complexity—natural gas supply, power price exposure, and ancillary services—that suit a balance sheet bank with a trading engine. The repeatability matters. Once one 1- to 2-gigawatt campus is financed and operational, the blueprint can be cloned across Texas and other permissive jurisdictions, turning a headline transaction into a pipeline. With public markets rewarding AI infrastructure stories and private credit awash in cash, the syndicate base is deep. That combination can compress execution timelines, a selling point to tenants starved for megawatts.
ERCOT is dealing with an avalanche of gigawatt-scale requests that would have seemed implausible two years ago. Demand from AI, crypto, and industrial reshoring is testing Texas’ ability to connect load fast enough. Private power campuses are a workaround. By generating behind the meter or tapping lightly congested interties, developers cut years off interconnection queues and reduce curtailment risk. That does not eliminate dependencies. Gas pipelines, water permits, and air approvals still govern pace. But compared with waiting in the queue, the campus model offers a more controllable critical path—and that is the currency AI operators value most.
The rush has an environmental cost profile that investors must underwrite. Texas developers have filed plans for more than 100 new gas-fired plants over the next few years to meet industrial and data center loads, a slate that could add more than 100 million metric tons of greenhouse gases annually if fully built. Many AI campuses will pair gas turbines with batteries and co-locate near renewables to shave emissions and hedge costs, yet the near-term reality is simple: firm power in ERCOT still means gas. If federal or state rules tighten, financing costs could rise, timelines could stretch, and operating profiles may shift toward cleaner blends faster than pro formas assume. That uncertainty does not kill the deals—but it should price into covenants, step-in rights, and hedging packages.
AI tenants are signing multi-year, capacity-like contracts that function more like infrastructure offtakes than traditional utility bills. The structure pulls forward capital from lenders and sponsors, recouped through fixed and indexed payments tied to uptime commitments. It is a sweet spot for private equity and infrastructure funds seeking long-duration cash flows, and for banks like Goldman that can package debt, distribute tranches, then cross-sell commodity risk management. The result is a layered capital stack—term loans, asset-based revolvers, notes—secured by contracted megawatts and high-credit-quality counterparties. As more campuses hit commercial operation, expect refinancing waves that free up equity and recycle into the next build.
GS finished the session little moved, but the strategic direction is clear. After a volatile consumer push and a subdued M&A cycle, bankable megaprojects with hard assets and contracted revenues look like the right lane. Competitors from JPMorgan to Morgan Stanley will chase mandates as data center developers consolidate and hyperscalers standardize procurement. Capital charges under Basel rules can shape how much holds stay on balance sheet, but distribution appetite is ample with private credit and insurance capital hungry for duration. If Goldman converts this Texas mandate into a multi-gigawatt franchise, the revenue mix tilts back to classic Goldman: originate, structure, distribute, hedge.
Texas confers advantages—permissive siting, merchant-friendly markets, and mature energy logistics—that are hard to replicate at scale elsewhere. That is why Lancium’s sites sit in wind- and solar-rich West Texas and why Google and others are planting flags around existing transmission spines. But the model has limits. Water scarcity can constrain turbine cooling in arid counties. Local pushback can slow permits even in friendly jurisdictions. Supply chains for turbines, switchgear, and large power transformers remain tight. The first wave of projects will get built; the second wave will compete for scarce kit and skilled labor, pushing timelines and budgets.
Near-term catalysts will come fast. Watch for EPC contract awards and gas turbine orders that lock schedule, P95 availability guarantees in offtake term sheets, and signs that ERCOT fast-tracks interties around major campuses. Track any shift in emissions permitting or incentives for cleaner onsite generation, including tax credits that could tilt designs toward hybrid gas-renewables-battery stacks. Consolidation among developers is likely as capital picks perceived winners with shovel-ready acreage, interconnect proximity, and anchor tenants. For markets, the tell will be whether Goldman and peers announce repeat financings tied to the same campus platform—turning today’s headline into a serial engine. If that happens, the AI power land grab moves from story to steady state, with Wall Street underwriting the new base layer of the internet.
