The Gigawatt Era: Inside the 5 Largest Data Center Projects in America in 2026 and the Contractors Pouring the Concrete
A few years ago, a 100-megawatt data center was considered massive. Today, developers are planning campuses measured in gigawatts. These projects require thousands of skilled tradespeople, billions in infrastructure, and enough electricity to power entire cities.
According to the American Edge Project, as of March 2026 the U.S. had an estimated 4,150 active data centers, with roughly 2,788 more announced or under construction. Texas and Virginia are each pushing past 100 active builds. According to McKinsey's estimates, global data center investment could approach $7 trillion by 2030, with more than 40% of it in the U.S. and roughly 70% spent by hyperscalers.
This post walks through the five biggest builds in the country right now, the engineering and logistical knots each one has to untie, and where it's public the general contractors and key subcontractors actually doing the work.
1. PORTS Technology Campus - Piketon, Ohio (SB Energy / SoftBank)
Scale: Up to 10 GW · ~2,700 acres · reported $67.2B total program
If one project captures the sheer audacity of the moment, it's this one. SB Energy, a SoftBank Group company, broke ground in March 2026 on what U.S. officials have described as potentially the world's largest AI data center, built on leased land adjacent to the U.S. Department of Energy's former Portsmouth Gaseous Diffusion Plant, a Cold War-era uranium enrichment site near Piketon. Commerce Secretary Howard Lutnick described it at the groundbreaking as potentially the largest construction project in the country.
The numbers are staggering. The campus is engineered to draw up to 10 GW of power. Phase 1 alone is an 800-MW data center expected to cost $30 to $40 billion and finish in early 2028. SB Energy is committing $4.2 billion through AEP Ohio to build new 765-kilovolt transmission lines and grid infrastructure, plus a plan for roughly 10 GW of new generation, the overwhelming majority natural gas. The project is projected to support up to an estimated 35,000 construction jobs cumulatively over the life of the build (not all on-site at the same time), plus roughly 2,500 permanent operations roles. In April 2026 it became the first project ever granted FAST-41 status under the high-performance-computing sector, a federal designation that forces participating agencies to synchronize permitting timelines.
The complexity: This is a brownfield remediation project, a merchant power plant, a transmission megaproject, and a hyperscale data center stacked on top of one another. Legacy enrichment buildings are still being demolished as new construction starts. The site was chosen partly because of the density of existing transmission infrastructure left over from its nuclear past, but that backbone still needs major reinforcement to carry 10 GW into the PJM grid.
Contractor picture: The development is led by SB Energy with AEP Ohio handling transmission. As of this writing, SB Energy has not publicly named a single prime GC for the data center build, unsurprising for a phased program this early. Given the layered civil, electrical, mechanical, and pipeline scopes, expect a coalition of mission-critical GCs and specialty trades rather than one prime, similar to the Meta model below.
2. Hyperion - Richland Parish, Louisiana (Meta + Blue Owl)
Scale: Targeting up to 5 GW · ~2,250 acres · 4M sq ft · ~$27B JV (Meta has signaled $200B+ all-in over time)
Hyperion is the build everyone in the industry is watching, both for its size and for how it's structured. Meta is erecting its largest-ever data center campus on the former Franklin Farm mega site in northeast Louisiana.
Contractor picture (the clearest of any megaproject):
General contractors: A three-firm joint venture of Turner Construction, DPR Construction, and Mortenson is building the campus. Per DPR, the program includes one network data center, four support structures, and extensive modular electrical and mechanical systems.
Power partner: Entergy Louisiana is building new gas-fired generation, transmission, and renewables, with energy infrastructure costs reported north of $5 to $6 billion. Meta reached a March 2026 agreement with Entergy for seven additional plants (5.2 GW) on top of an initial three. Meta is paying for the construction of all 10.
Workforce: roughly 3,700 workers on site as of late 2025, expected to peak around 5,000 by mid-2026.
Local spend: Meta reported contracting more than $875 million with Louisiana-based businesses in the first year.
The complexity: Two big ones. First, the financing is a milestone in its own right, Meta and Blue Owl Capital formed a joint venture (an SPV nicknamed "Beignet") in which Blue Owl owns 80% and Meta 20%; Meta operates and leases the campus with a built-in exit ramp, and BlackRock bought more than $3 billion of the venture's bonds in a sale arranged by Morgan Stanley. Second, community friction: Entergy's "pay-for-your-own-power" model is designed to shield other ratepayers, but locals have dealt with housing-cost spikes, construction-debris damage, and a rise in road accidents, a reminder that gigawatt builds land hardest on small towns.
3. Stargate Site 1 (Lancium Clean Campus) - Abilene, Texas (OpenAI / Oracle / SoftBank / Crusoe)
Scale: 1.2 GW · ~1,100 acres · ~4M sq ft across 8 buildings · ~$3.4B build-to-suit JV (within the $500B Stargate program)
This is the flagship of the headline-grabbing Stargate initiative. The first two buildings (980,000 sq ft, 200+ MW) went live on Oracle Cloud Infrastructure for OpenAI in September 2025, and Crusoe topped out the final building of the eight-building plan in May 2026.
Contractor picture (well documented):
General contractor: DPR Construction, which broke ground in 2022.
Developer/design engineer/operator: Crusoe Energy Systems, building on land owned by Lancium (the "Lancium Clean Campus").
Electrical EPC / substation: Mortenson served as the EPC contractor for the electrical infrastructure, building the 200-MW, 138-kV substation and associated AEP transmission work.
On-site power: A roughly 360-MW (scaling toward 1+ GW) natural gas turbine fleet, with turbines reported from GE Vernova, Solar Turbines, and a major supply agreement with Parker Hannifin.
Capital: A $3.4B build-to-suit JV among Crusoe, Blue Owl Capital, and Primary Digital Infrastructure.
Workforce: Crusoe has cited around 7,000 electricians, construction pros, and tradespeople on site daily at peak.
The complexity: Speed and power. Crusoe is positioning Abilene as a new "blueprint" for velocity, vertically integrating development, build, and operation to compress timelines. The site leans on West Texas's stranded/low-cost energy and a largely self-provided power strategy (a simple-cycle gas plant filed to run as off-grid backup) to sidestep grid-interconnection queues that stall most projects. It's worth noting ENR has reported the broader Stargate program scaled back some of its initial pace and ambition even as Abilene itself raced ahead, construction here runs into early 2027.
4. Project Rainier - New Carlisle, Indiana (Amazon Web Services, for Anthropic)
Scale: $11B Phase 1 (operational) + $15B Phase 2 announced · 1,200 acres · up to ~22 buildings
Rainier is Amazon's answer to Stargate, and, notably, it was built exclusively to train and run models for Anthropic, on Amazon's own Trainium chips. A year ago the New Carlisle site was farmland; by late 2025 it was one of the largest operational AI clusters in the world. In November 2025 Amazon announced an additional $15 billion for new campuses across Northwest Indiana, pushing committed investment past $26 billion and making it the largest single construction project in Indiana history.
Contractor picture: Amazon did something unusual, site lead Josh Sallabedra (a 14-year Amazon data center veteran) brought on four general contractors simultaneously to compress the schedule. Amazon hasn't publicly broken out which firms hold which scopes. Power comes via AEP's Indiana Michigan Power subsidiary for Phase 1 and NIPSCO for the expansion, with Amazon partnering with both utilities on rate protections for local customers.
The complexity: Amazon deliberately chose a more permanent build path than some rivals, updating its facility design midway through construction to accelerate deployment rather than throwing up temporary structures. The project also hit a real regulatory wall: the Indiana Department of Environmental Management issued a temporary halt order over unauthorized impacts to wetlands at the site, and Amazon spent early 2026 working through revised applications and mitigation. It's a clean illustration of how environmental permitting, not engineering, is increasingly the binding constraint.
5. Prometheus - New Albany, Ohio (Meta)
Scale: ~1 GW (billed as the world's first gigawatt-capable cluster) · multiple buildings · "tent-style" construction
Prometheus is smaller in dollar terms than the others here but matters because of how it's being built and powered. Meta describes it as what would be the world's first gigawatt-capable data center, coming online in 2026, made up of at least five buildings working together as a supercluster.
Contractor picture: Meta hasn't named a prime GC publicly, but a recurring name in the public record is Sidecat LLC, listed as a Meta data-infrastructure sub-processor/subcontractor entity that has been assembling land (including a $42M, 72-acre parcel on Clover Valley Road in late 2025) and is tied to a 200-MW on-site natural gas project. Corning is supplying fiber optic cable under a multi-year, up-to-$6B agreement spanning Meta's data centers.
The complexity: Two themes. First, speed-over-polish construction, Meta is famously using giant tent-style structures to stand the cluster up faster than conventional shells would allow. Second, an all-of-the-above power scramble: Meta has signed nuclear agreements with TerraPower, Oklo, and Vistra (including 2.1+ GW from Vistra's Ohio nuclear fleet and a 1.2-GW Oklo campus in Pike County targeted as early as 2030) on top of on-site gas, all structured so Meta covers the energy costs rather than passing them to consumers.
The Subcontractor Ecosystem Behind the Headlines
The hyperscalers and developers get the press, but gigawatt campuses are won and lost in the specialty trades. A few names that recur across these and similar 2026 builds:
Electrical: Rosendin, the nation's largest employee-owned electrical contractor and arguably the leading hyperscale electrical specialist, is a fixture on mega data center work, increasingly delivering integrated modular power skids and enclosures rather than just field labor.
Power-plant EPC: Kiewit is the design-builder of choice for the on-site generation now bolted onto these campuses, for example, it's building the Homer City Energy Campus in Pennsylvania, powered by GE Vernova turbines, a 4.6-GW gas plant explicitly sized to feed a data center campus.
GCs and A/E firms broadly: Industry market reports consistently list DPR, Turner, Mortenson, Holder Construction, Clayco, HITT Contracting, Jacobs, and architect-engineer Corgan as the firms generating the most revenue from complex data center work.
Turbines and prime power: GE Vernova, Solar Turbines (Caterpillar), and Parker Hannifin are supplying the on-site gas generation that's become standard as grid queues stretch for years.
Cooling: As racks cross 50 to 100 kW, direct-to-chip liquid cooling is going mainstream, pulling specialists like Motivair (Schneider Electric) and Vertiv deep into the mechanical scope.
Fiber/connectivity: Corning, via its multibillion-dollar Meta agreement, anchors the optical backbone.
By the Numbers: The Workforce Crunch
If power is the first constraint on these builds, skilled labor is the second, and it's catching up fast. A single hyperscale campus can absorb anywhere from roughly 1,500 workers at peak on a conventional build to an estimated 35,000 on the very largest. Here's where the projects in this post land at peak construction:
SoftBank PORTS Technology Campus (Ohio): up to an estimated 35,000 (projected cumulatively over the life of the project, not concurrent on-site workers at any single point)
Stargate Site 1 / Abilene (Texas): ~7,000 daily
Meta Hyperion (Louisiana): ~5,000 at peak (mid-2026), up from ~3,700
Amazon Project Rainier (Indiana): 1,000+ roles
Typical (non-mega) hyperscale facility: up to ~1,500
But the headline numbers hide the real story: these aren't generic labor counts; they're a specialty-trades bottleneck. Electrical work is the spine of a data center, the IBEW estimates electrical systems account for 45% to 70% of total construction cost (switchgear, paralleling switchboards, UPS, PDUs, battery storage), and every other trade on the job depends on the electrical scope landing right. Alongside electricians, the chronically short roles are mechanical contractors, HVAC technicians, controls specialists, high-voltage field talent, and commissioning teams.
The macro shortage behind it is severe:
The U.S. construction industry entered 2026 short roughly 439,000 workers, the majority in skilled positions like electricians and pipe layers.
Associated Builders and Contractors estimates the industry needs to hire up to 349,000 additional workers in 2026, rising toward half a million in 2027.
On electricians specifically: 300,000+ needed against only ~20,000 retiring per year, with 41% of the existing workforce set to retire by 2031.
That scarcity is reshaping economics and site selection. Data center work now carries a wage premium of up to 30% over typical construction, with Northern Virginia electricians clearing $120K+. Demand has been intense enough that a DMV-area electricians' union doubled its membership to 14,700 between 2018 and early 2026. And the cost of getting it wrong is brutal: a delay on even a modest 60-MW facility can run an estimated $14.2 million per month in lost revenue, which is exactly why labor availability has become a site-selection criterion, not just a staffing line item.
For specialty contractors, this labor shortage creates a different challenge. Finding electricians is only part of the equation. Coordinating multiple staffing suppliers, tracking compliance, onboarding workers quickly, and maintaining workforce visibility across massive jobsites has become just as critical. That's where workforce management platforms like Smart Labor Management fit into the equation.
The Through-Lines
Step back from the individual projects and the same five complexities show up everywhere in 2026:
The land war is really a power war. Every project on this list is gated by electricity, not real estate. The winning move is bringing your own generation, whether gas turbines, nuclear PPAs, or 765-kV transmission, to skip multi-year interconnection queues.
Brownfield reuse is back. A decommissioned uranium plant (PORTS), a retired coal station (Homer City), a stalled Foxconn site (Microsoft's Mount Pleasant), a former power plant (New York's Cayuga): legacy industrial sites come with the one scarce thing: existing grid infrastructure.
Speed is the product. Tents, modular skids, mid-build design changes, vertically integrated developers, FAST-41 permitting: every choice optimizes for time-to-energization.
The capital stack is novel. Blue Owl SPVs, BlackRock bond sales, build-to-suit JVs, "pay-for-your-own-power" utility deals. These are project-finance structures, not real estate plays.
Communities push back. Wetlands halt orders, cracked windshields, housing squeezes, ratepayer fights. The gigawatt era is colliding with small towns, and the social license to build is now part of the critical path.
The companies that figure out how to move fast and keep the locals, the regulators, and the grid operators on side are the ones who'll define the next wave. For the contractors and trades, electricians, mechanical crews, civil firms, turbine OEMs, it may be the busiest decade the industry has ever seen.