Test report DSG-5263 · Rev D · tested November 18, 2026
AI Datacenter InfrastructureDevice under test
Giga Energy Signs $1B, 10-Year Lease for 60MW Texas AI Campus
Giga Energy's $1B, 10-year lease covers Project Clay, a 60MW Texas AI campus with spring 2027 service and an unnamed U.S. cloud tenant.
- Read
- 4 min
- Words
- 745
- Node
- 5nm
- Operator
- Elena Vasquez
Spec summary
- Giga Energy signed a ~$1 billion, 10-year lease on November 17, 2026 for a 60MW AI data center in Hunt County, Texas.
- Service at Project Clay is expected in spring 2027 after a July 2026 groundbreaking and a claimed sub-eight-month factory build.
- The tenant is an undisclosed U.S.-based AI cloud provider; no GPU count, power supplier or payment schedule was revealed.
- A flat lease split works out to about $1.67 million per megawatt per year.
- The site will use closed-loop cooling with no municipal or groundwater draw.

Giga Energy signed a roughly $1 billion, 10-year lease on November 17, 2026 for Project Clay, a 60MW AI data center campus in Hunt County, Texas — and it won't say who the tenant is. The counterparty is described only as an undisclosed U.S.-based AI cloud provider. Service is expected in spring 2027.
Giga Energy will develop, own and operate the roughly 30-acre campus, making it a landlord and operator rather than a contractor handing over keys. Groundbreaking happened in July 2026, and the company claims its factory-built GigaBase approach will finish the facility in under eight months. Count forward from July and eight months lands in March 2027, so the dates line up — but that build claim is the company's own, not an independent audit.
What does the lease math work out to?
Giga Energy published no payment schedule, so per-year figures are arithmetic, not disclosure. A flat $1 billion over 10 years is about $100 million a year. Spread across 60MW, that's roughly $1.67 million per megawatt per year, or about $139,000 per megawatt per month. At full load, 60MW running 8,760 hours would consume about 525,600 MWh a year — a ceiling, not a forecast, since no facility runs flat-out every hour and Giga hasn't defined whether the 60MW is grid draw or critical IT load.
The structure matters more than the headline. A 10-year term is long by cloud standards, where customers traditionally kept contracts short and workloads portable. A provider signing for a decade is betting AI demand outlasts several hardware generations.
Why the closed-loop cooling pledge?
The site will use closed-loop cooling and will not draw on municipal or groundwater supplies. Water use has become a flashpoint — local fights have blocked 75 data center projects worth $130 billion in the first quarter alone. The pledge reads as reassurance aimed as much at local officials as at the tenant. The trade-off is energy: systems that skip evaporative cooling can draw more power for the same heat load, and Giga has published no efficiency target for a Texas summer.
Power remains the open question. The announcement names no supplier and doesn't say whether the campus uses the grid, on-site generation, or a mix. It also gives no GPU count, no server configuration, no chip vendor, no construction cost and no payment schedule.
How does 60MW compare?
Sixty megawatts is mid-sized by 2026 standards — enough for a serious GPU cluster, small enough to build fast. A recent Texas order for 42 Wärtsilä gas engines delivering 790 MW off-grid is roughly 13 times larger by nameplate power. The 2.7 MW TECO HULK 3.0 modular unit shown in Warsaw is about 22 times smaller. Project Clay sits in the middle, where much of the practical demand for AI capacity lives.
For context on the money: AI infrastructure outlays are estimated near $700 billion while tech layoffs have reached 206,817 workers. Capital is flowing to buildings, chips and power while payrolls shrink elsewhere.
Why lease instead of build?
A cloud provider's scarce resources are capital and time. Every dollar spent on concrete and chillers is a dollar not spent on accelerators, and every month waiting for a building is a month of lost rental revenue. For the landlord, a 10-year contract anchors financing — lenders underwrite construction debt against long leases. The counterparty risk runs the other way: a $1 billion obligation is only as sturdy as the unnamed company that signed it.
Timing risk exists too. SK hynix has locked up more than 50% of HBM4 capacity for 2027 and 2028, so if the tenant's hardware arrives late, the lease clock and revenue clock can drift apart. Giga hasn't said whether rent starts at building delivery or at customer operations.
What happens next?
Our predictions: the tenant gets named by mid-2027; the schedule slips by weeks, not months; Giga announces a second site; more leases carry flexible-load clauses; and water commitments become a standard line in announcements. If AI rental pricing falls sharply, a decade-long lease at today's rates could squeeze the tenant — if capacity stays scarce, it may look cheap in hindsight.
The takeaway: a confirmed deal with confirmed size, location and timeline, wrapped around a customer nobody can yet name. We'll update if Giga Energy or the customer discloses more.
via shattered.io (Source)
More from Elena Vasquez
Show full bio
Senior reporter covering people, launches and day-to-day practice in hardware layer of artificial intelligence at Die Signal.
675 articles
Same lot · LOT-362E
- DSG-861428nmGiga Energy lands $1bn lease for 60MW Texas AI data center
- DSG-222410nmSpaceX's $24B Starbase Gas Plant Becomes Texas Cloud Policy Test
- DSG-84453nmFoxconn Prices NVIDIA Vera Rubin Datacenter at $47B per Gigawatt
- DSG-188065nmConsortium Plans 400MW AI Data Center Complex in Oman
- DSG-68893nmWärtsilä Books 790 MW Off-Gas Power Plant for Texas Data Center