hyd.eco — Site Intelligence for AI data centres

Critical infrastructure siting · United States · AI data centres first

Know what kills the site
before you pay for the option.

We screen a site against the four things that end projects — power, water, permits and local politics — and we tell you which of them nobody can answer yet. One site, three working days, no charge for the first three.

Send us a site See a real brief Built first for independent developers, and for miners turning power into AI.

Evidence · what a brief actually says

Sweetwater County, Wyoming. Our own screen, unedited.

This is not a mock-up. It is the output for the county we are studying first — including the part where our own answer disagrees with the people who live there.

Site Brief 41.66, −108.88 · 300 MW IT · 5 Oct 2026
Verdict
Conditional
Go requires the utility in writing
County rank
141 of 3,100 · first in Wyoming
Time to power
42–66 mo estimate, not a queue position
Nearest heavy grid
500 kV Anticline, ~15 km
Generation in county
2,441 MW coal — and coal is retiring
Stop-factors found
None at county level
Unknowns we will not guess
4 of 9 capacity, water rights, fibre, zoning
Sources behind the facts
15 each with a response time
Contradiction flagged. In September 2026 the county commissioner said publicly that power for industry in south-west Wyoming is already exhausted. Our screen ranks this county in the top 5% of the country. Both cannot be true. An energy auditor is resolving which substation the answer depends on — and until that is settled, this county is Conditional, not Go.
Ranked by the same model for all 3,100 US counties · no paid placement · corrections welcome and published.

Limits · what we will not tell you

Four things you will never read in our report.

Screening is only worth buying if you know exactly where it stops. Most reports leave that boundary to the reader. We put it on the first page.

  • We will not call capacity available. Proximity to a substation is not headroom. Only the utility knows what is free. We give you the owner, the voltage, the distance and the exact question to send them — and we keep the verdict Conditional until they answer in writing.
  • We will not give an engineering or legal opinion. No foundation design, no title opinion, no permitting guarantee. We tell you which licensed professional to hire and what to ask them.
  • We will not hide an unknown inside a score. Every brief lists what we could not verify. A tidy number with silent gaps is worse than no report, because it gets believed.
  • We will not take a site that is already under option. If you have signed, the decision is made. We are built for the stage before that.

Why sites die · 2026

Land is no longer the scarce input.

Power was promised, then queued

Texas ordered an audit of every data centre in the interconnection queue — roughly 474 GW of requests, about 90% of them data centres. New approvals wait for the outcome.

Cost of finding out late: years, and a deposit. ERCOT queue audit, Aug 2026

The council voted it down

A 430-acre technology park in Hanover County, Virginia was refused. In a neighbouring county a 900 MW project failed by a single vote. The land was fine. The politics were not.

Cost of finding out late: the site, publicly. Virginia Mercury / WTVR, May 2026

The state froze new load

New York paused new projects above 50 MW for up to a year. Fourteen states have moratorium bills in play. A county can be perfect and still be closed.

Cost of finding out late: a stalled deal. Executive order, 14 Jul 2026

The neighbours organised first

In Dinwiddie County the developer changed the cooling design and moved buildings off a Civil War battlefield core — after the public meeting, not before.

Cost of finding out late: redesign under scrutiny. Local coverage, Sep 2026

Method · four moves, three days

From a load to a decision.

01

Tell us the load

Megawatts, timeline, latency needs, budget, and what you already know. These become hard filters and weights — not a wish list.

02

We screen the ground

Grid, water, fibre, land and zoning, permitting, incentives, hazards, on-site energy. Fifteen public services answer in parallel; each fact carries its source and its response time.

03

We publish the disagreements

Where our data contradicts local officials, a utility filing or your own information, it goes in the brief as a flagged contradiction — not smoothed over.

04

You decide, with the gaps named

Verdict, stop-factors, time to power, the unknowns, and the exact questions to send the utility, the county and the water authority.

The engine · eight layers, labelled honestly

What is automatic, and what still needs a human.

Every layer below is either computed for you or explicitly verified with you. No layer is silently half-done.

PowerSubstations, voltage, distance, ownership, gasAutomatic
WaterStress, drought, rights, cooling optionsPartly manual
ConnectivityLong-haul fibre, distance to exchange pointsManual today
Land & zoningArea, zoning class, buildable areaOwner is manual
PermittingWhich bodies, which processTimelines manual
IncentivesState and local programmesChecked by hand
Risk15 hazards now, and to 2050; sentimentAutomatic
On-site energyGas, renewables, storage, natural hydrogenAutomatic

Who it is for

Independent developers with power in hand

You are holding several sites and need the weak ones gone before you spend on options.

Miners and generators turning to AI

You already own land, a substation and an interconnection position. Which of them actually works as a campus?

Site selection and engineering firms

We do the upstream filter, white-label if you prefer, so your own work starts on sites worth the fee.

Who it is not for

Teams with an in-house siting group

At hyperscaler scale you already have better data than we do. We would be a step backwards.

Sites already under option

The decision is behind you. Screening now is confirmation, and confirmation is not what we sell.

Anyone who wants a number without caveats

Our briefs name what is unknown. If a clean score is the deliverable, a different supplier will serve you better.

Origin

Built for hydrogen. Rebuilt for power.

hyd.eco began as a geospatial method for natural hydrogen exploration: finding where geology and existing infrastructure meet, so a resource could actually reach a market. The research was presented at EHEC 2026 in Seville and at PowerConnect in Gdańsk.

Data centres ask the same question from the other side — where do energy, water and connectivity meet a building? The screening engine carries over. Hydrogen stays inside it.

Founder

EK

Evgeny Korneev

Twenty years between companies, governments and the people who live with what those companies build: construction approvals and a site on public contracts, operations at an oil export pipeline consortium, partnerships across Russia, Kazakhstan, Argentina and Europe.

For the past two years: infrastructure siting — first natural hydrogen, now data centres. Most of the open questions turn out to be about power, water and land use, and about who in a given place needs to say yes.

LinkedIn · e@hyd.eco

Pilot · three slots

$0 for the first three, in exchange for one thing

What you get

  • A screen of one site you are already considering — grid, water, permits, politics, hazards
  • A ranked shortlist if you have several, with the reasoning behind each rank
  • A decision memo where every automatic fact carries its source and response time
  • A named list of what we could not verify, and who to ask
  • Three working days, directly with the founder

What we ask

  • One 30-minute call afterwards: what was useful, what was wrong
  • Permission to use the case anonymously
  • Patience with a product that is early and will say so

Send a site

Tell us what you are planning and where.

The founder reads every message and replies personally — usually within one working day, and with a straight answer if we are not the right fit.

We use this to prepare, not to send you a sequence of emails. No newsletter, no CRM drip.