Infrastructure · August 17, 2026
Flexibility is becoming a site asset.
The market has spent years asking one question about data center sites: how many megawatts can you deliver? We think the next question is: how rigid does that demand have to be?
The old model treats load as binary.
A 100 MW project enters a market as a request for 100 MW of firm service. If the grid cannot serve it, the answer is usually a larger upgrade, a longer queue, a private generation plan, or a different site.
That model makes sense for a factory that cannot stop a production line. It is less obvious for every kind of compute. Training, batch processing, crypto, storage, inference, and distributed workloads have different operating constraints. Some can pause. Some can move. Some can reduce load for short periods. Some can pair with batteries or on-site generation.
Flexibility can change the power path.
ERCOT's Batch Zero framework now includes a pathway for large customers willing to accept curtailment in response to local transmission constraints. That is a meaningful shift: controllability can become part of how capacity is allocated, not just an operational feature after the site is built.
Vish Tella's analysis on pricing grid flexibility in Texas pushed the same idea into economic terms. Once flexibility has value, it should be captured as data and used during site selection.
So we are adding it to Huluku.
Every compute requirement we collect can now carry a flexibility profile. Every site can carry information about its power status, storage, backup, behind-the-meter generation, and potential for flexible service.
The first Huluku Flex Score is intentionally simple: curtailment, response speed, duration, workload shiftability, and on-site support. The point is not to declare a universal truth with a 0-100 number. The point is to make a previously fuzzy attribute comparable enough to influence matching and diligence.
The marketplace gets more interesting from here.
A powered site should not be represented by megawatts and price alone. A compute buyer should not be represented by MW demand alone. The real match includes timing, workload, geography, fiber, density, firmness, and flexibility.
If we capture those variables at intake, Huluku can become more than a list of sites. It can become a matching layer between compute demand and the actual ways power can be delivered.