Construction AI BriefSubscribe →
Issue
№109
Pillar
Trend
Audience
GC ops
Dated
2026.07.24

OpenAI is designing and building its own data center — and says it will publish the blueprints

OpenAI announced a $20-30 billion Georgia data center campus it will design and build itself, breaking from its usual model of leasing capacity from developers — and its compute strategy VP says the resulting blueprints will be made public. That changes who a GC pitches for hyperscale data center work, and what they're bidding against.

ByConstruction AI BriefAbout this publication

OpenAI is designing and building its own data center for the first time, instead of leasing finished capacity the way it does everywhere else — and the executive running the project says the resulting design will be published for anyone in the industry to use. For the general contractors and construction managers who build hyperscale data centers, that's two changes in one announcement: a new client sourcing its own build team directly, and a design playbook that might stop being anyone's proprietary edge.

What did OpenAI actually announce?

On July 22, OpenAI unveiled Project Camellia, a data center campus on roughly 1,400 acres in Effingham County, Georgia, inside the Savannah Gateway Industrial Hub. The company has contracted with Georgia Power for up to 3.2 gigawatts of electricity, delivered in phases from 2028 through 2032. OpenAI's VP of compute strategy, Sachin Katti, put the fully built cost above $30 billion. OpenAI is also committing $80 million to local schools, public safety, health care, workforce training, and housing, plus up to $71 million in AI credits for students at Georgia colleges near the site.

None of that alone would be unusual — hyperscale data center campuses of this size have become routine news this year, from Meta's Alberta campus to the half-built, permit-stalled Project Jupiter in New Mexico. What's different here is who's holding the pen.

Why does it matter that OpenAI is building this one itself?

Everywhere else, OpenAI buys compute — it leases capacity inside data centers that Oracle, Amazon Web Services, CoreWeave, and other operators design, finance, and build. Project Camellia flips that: OpenAI is the principal designer and builder, and it's now searching for its own financing and construction partners to scale the project to full size. That means the GC or CM contract for this campus, and presumably for whatever OpenAI builds next under this model, gets sourced directly by OpenAI rather than by a data center developer acting as the middleman. If you're a firm with a hyperscale data center group — Turner, DPR, Holder, Mortenson, Clayco, JE Dunn, and the rest of that short list — the buyer you'd be pitching just changed, and the RFP may not run through the channels you're used to.

What does "public blueprints" mean for the GCs who build these things?

This is the part worth watching closest. Katti has said publicly that OpenAI intends to develop uniform data center blueprints and publish them for the broader industry, on the theory that shared design lessons cut both cost and build time. A hyperscale data center's design — power density, cooling topology, structural loading, electrical distribution — has historically been a proprietary asset that a handful of design-build GCs iterate on job after job, and it's a real part of why those firms win repeat work at premium fees. If OpenAI publishes a reference design that performs well, that specific edge gets easier to replicate.

Leased-capacity model (Oracle, AWS, most hyperscalers)OpenAI's self-build model (Project Camellia)
Who hires the GCData center developer/operatorOpenAI directly
Who owns the designDeveloper, often proprietaryOpenAI, explicitly to be published
Where the RFP surfacesDeveloper's existing GC relationshipsOpenAI's own construction-partner search
Design as competitive edgeYes — repeat-client design IPDiminished if the blueprint is public

Should a mid-size GC try to get in on this?

Realistically, Project Camellia's scale — $30 billion, four buildings, 3.2 GW — puts the general contract with the handful of firms that already run national hyperscale data center groups. But the labor and subcontractor side is more open: OpenAI has explicitly committed to prioritizing local contractors and suppliers during both construction and operations, and Effingham County will be competing for craft labor against other Georgia mega-projects already drawing from the same regional pool. A regional mechanical, electrical, or sitework sub within reach of Rincon has a real reason to get in front of whichever GC wins the work now, before subcontractor packages get let. Estimators tracking data center opportunities should also flag that the full $30 billion-plus build is still contingent on OpenAI landing its construction partners — the project isn't fully staffed yet, which means the RFP window on parts of it may still be open.

Construction AI Brief tracks which AI-industry moves actually change how a GC gets hired — new pieces most days at constructionaibrief.com.

FAQCommon questions
What is OpenAI's Project Camellia data center?
Project Camellia is a data center campus OpenAI announced on July 22, 2026, for Effingham County, Georgia, near Savannah. It's sited on roughly 1,400 acres inside the Savannah Gateway Industrial Hub, will draw up to 3.2 gigawatts of power from Georgia Power delivered in phases from 2028 to 2032, and carries a construction cost OpenAI's own compute strategy VP put at more than $30 billion at full build-out.
Is OpenAI building this data center itself, or hiring a developer to build it?
OpenAI is acting as its own designer and builder on this project — a first for the company. Its other facilities run on leased capacity from operators like Oracle and Amazon Web Services. For Project Camellia, OpenAI says it is now sourcing its own financing and construction partners directly.
Will OpenAI's data center designs be made public?
OpenAI's VP of compute strategy, Sachin Katti, has said the company plans to develop uniform data center blueprints and make them publicly available to the industry, arguing that shared design lessons can cut both cost and build time. No publication date or licensing terms have been announced.
How many construction jobs will Project Camellia create?
OpenAI says the project will generate thousands of construction jobs during the build, with local contractors and suppliers prioritized. Permanent on-site staffing is projected at 400 jobs when the first phase opens in 2028, growing to roughly 1,000 by 2032.
Where is Project Camellia and when will it be finished?
It's in Effingham County, Georgia, about 45 minutes from Savannah. Power comes online in phases starting in 2028, with construction and full build-out continuing through 2032.
End of sheet — issue №109
Published · 2026.07.24
Project
Construction AI Brief
Dated
2026.09.07
Sheet
1 / 1
Rev
A
Published independently · constructionaibrief.com · © 2026Facebook·Privacy·About