Power-First Data Centers in 2025: How Grid Constraints Are Repricing Land, Leases, and Revenue
- Ishan Bhattarai
- 2 hours ago
- 4 min read
In 2025, the U.S. data center boom isn’t being held back by demand; it’s being held back by power. Everyone wants more computing power, but projects hit the same wall of grid constraints: substation capacity, crowded interconnection queues, and transmission upgrades that can take years, shrinking the number of sites that can actually get energized on schedule. The successful developers aren’t just the ones with the biggest checks but they’re the ones who lock in land with real offtake potential and a clear, believable path to power. At LandGate, that “path to power” is exactly what we underwrite: Analytics at the parcel level provide deep insight into on-the-ground feasibility, defining the differences between a real project and a slide-deck project.
A key point we see repeatedly in market conversations is that “data center revenue” varies widely depending on what is included. A practical way to frame the topline is to anchor on established industry definitions. IBISWorld estimates the US Data Processing & Hosting Services industry at roughly $383.8B in 2025, while separately estimating Hyperscale Data Center Services at $111.2B (2025) and US Colocation Services at $17.1B (2025). The takeaway isn’t which taxonomy a reader prefers; it’s that the revenue pool tied to “digital infrastructure delivered in the US” is now large enough and accelerating fast enough that power delivery has become a first-order economic input. LandGate’s siting and feasibility work consistently shows that the difference between a “good” site and a “great” site is increasingly measured in months shaved off utility timelines, not in cents per square foot on land cost.
Colocation pricing is where the power story becomes impossible to ignore, and the figure below captures that shift with LandGate’s average market estimates for primary-market wholesale asking rates on 250–500 kW requirements. The typical pricing moved from roughly $120/kW-month (H2 2021) to about $138/kW-month (H2 2022), then stepped up again to around $165/kW-month (H2 2023) and approximately $184/kW-month (H2 2024), with H1 2025 trending higher based on a modest additional increase. The headline isn’t just that prices rose; it’s what the pricing is actually buying now: certainty of powered capacity. In constrained metros, colocation behaves less like a real estate product and more like a power access product, because the hardest thing to secure isn’t land, it’s deliverable megawatts on a timeline customers can underwrite. When substation headroom is limited, feeders are tapped out, and transmission or utility upgrades push out for years, the pool of sites that can support near-term energization shrinks and the market clears through higher $/kW-month pricing.
Now translate that pricing into project economics: the chart below shows LandGate estimated annual revenue for a 5 MW colocation facility delivering in 2026 across realistic utilization ramp scenarios. The takeaway is straightforward: at today’s power-constrained pricing levels, a 5 MW facility can land in roughly the $5–$12M revenue range in its first full year depending on leasing velocity, and then converge toward the $11–$14M run-rate at high utilization. This is why ‘time-to-power’ and ‘time-to-lease’ matter so much when interconnection delays push delivery to the right, you’re not just delaying a build, you’re delaying the revenue curve.
The AI layer is now turning into a revenue driver that directly shapes infrastructure behavior, rather than merely riding on top of cloud adoption. Reuters reported OpenAI’s annualized revenue run rate reached $10B as of June 2025, a scale that validates long-duration compute purchasing and materially impacts demand for dense, power-hungry infrastructure. For Perplexity, public reporting and statements have cited around $100M ARR in June 2025 and reports of exceeding $150M ARR by mid-2025 as fundraising accelerated. LandGate’s observations indicate that once AI application revenues reach this tier, these companies begin to behave less like “software users of cloud” and more like “economic anchors for capacity procurement.” Their priorities converge on the same physical truths: regional redundancy, predictable delivery schedules, and access to large contiguous power blocks, which is why we increasingly see AI-driven interest concentrating around substation-adjacent parcels, transmission-close industrial corridors, and markets that can credibly deliver power without multi-year slippage risk.
All of this is unfolding under an investment reality that is getting heavier, not lighter. Development economics remain compelling at the top line, but the capex and the schedule risk are rising. For AI-leaning designs, cooling and electrical distribution complexity can push total costs higher, which means the margin for schedule error narrows. When you combine this with tightening land dynamics, especially for the large parcels needed for campus-style builds; the message for investors becomes clear: “cheap land” is irrelevant if it cannot become “energized land” on a timeline that matches return hurdles. This is where LandGate’s approach becomes practical: parcel screening that integrates land characteristics with nearby grid realities helps stakeholders distinguish between theoretical develop-ability and investable deliverability.
Looking ahead, the market’s growth story remains intact, but it will be unevenly distributed across regions based on grid readiness and power deliverability. US data centers used roughly 4% of U.S. electricity in 2024, with projections indicating meaningful growth in consumption by 2030 as AI and broader digital demand expand. That macro reality implies that the best-positioned markets will be the ones that can upgrade transmission and distribution capacity, streamline interconnection, and bring generation online in step with load. Conversely, markets that cannot relieve bottlenecks will increasingly “price” their scarcity through higher colo rates, longer delivery timelines, and more frequent project deferrals. This is why LandGate’s grid-plus-parcel view matters: where power is deliverable, development follows; where it isn’t, pricing and timelines do the talking.
The core message for operators, hyperscalers, AI companies, landowners, utilities, and investors is straightforward: the competitive advantage in the next decade is not defined by who can build the biggest facility, but by who can secure the right land with a credible power path at the right time. From LandGate’s perspective, power is becoming the currency of digital expansion, and parcel-level feasibility is becoming the language that sophisticated stakeholders use to price that currency correctly. LandGate helps to make that feasibility legible so the market can move from hype-driven site selection to power-true decisions.
To learn more about the data & tools available to data center developers, book a demo with our dedicated infrastructure team.
