Deployments · 324 computed cases · parametric design-study estimates

Configure a HELIX site for your facility.

HELIX sites number up from a single ~5 MWe sealed module to a hyperscale campus, and every configuration on this page is computed, not quoted. Our engineering register runs 324 cases, six facility styles across six environments through nine operating scenarios, and each case logs the verification evidence it produces. Pick a facility style and an environment below and the page sizes the site live from the same parametric models, showing the honest derates along with the capacity. Every figure is a pre-application design target, not field data.

The discipline here is the same as everywhere else on this site: the model is printed, the inputs are visible, and the numbers follow from them. Net output per module falls on a hot afternoon and at altitude, so the configurator shows that fall instead of quoting a nameplate. Module counts carry an N+1 reserve because sealed-core swaps and single-module trips are planned events, not surprises. And every scenario in the register is paired with the attested record an operating site would actually produce, because a deployment you cannot verify is just a promise.

Which deployment fits your facility?

Choose a facility style and an environment. The results update from the parametric model; without JavaScript the page shows the mid-size colocation case in a Northern Virginia environment.

Model, exactly as computed: net-cycle efficiency η = 0.395 − 0.0009 × max(0, T − 15); gross = 12.3 × η × 0.98; fan parasitics = 0.25 + 0.006 × (T − 25) MWe; net = (gross − fans − 0.10) × (1 − 0.011) × (1 − 0.01 × altitude_m / 1000). Modules required = ⌈IT ÷ net⌉, plus the reserve for the facility style. Annual attested output assumes a 95% capacity factor against the IT load. Heat rejected is 7.1 MWth per installed module.

How does this configuration behave when things go wrong?

The register runs every configuration through nine operating scenarios, from an ordinary Tuesday to an extended station blackout. Across all 324 cases in the register, every capacity scenario passes: the N+1 reserve carries a swap or a trip, and the peak-ambient derate never takes a site below its IT load. That is a property of how the sizing rule works, sized against the honest derate with reserve on top, and it is a screening-level claim on the frozen design basis, not credited analysis. The table below is the selected configuration's row of the register, with the verification evidence each scenario produces.

Capacity checks: swap and trip scenarios use (total modules − 1) × net; the peak-ambient scenario recomputes net output at design ambient + 5 °C. Extended station blackout is a safety scenario, decay heat leaves by passive air cooling with zero operator action, so it is never a capacity question.

How does net output move with the environment?

This is the curve doing the work in the configurator. Output falls as ambient rises, because the dry cycle rejects heat to hotter air and the cooler fans work harder, and it falls about one percent per thousand meters of altitude. Both lines are computed at build time from the same model printed above; nothing on this chart is drawn by hand.

How does the verification work here?

Every scenario in the register ends the same way: with evidence. That is the point of deploying HELIX rather than a reactor you have to take on faith, and it works on three layers.

Protection. Each module's identity begins with a witnessed genesis at the factory, and its telemetry crosses a hardware one-way path, an observe-only boundary that physically cannot carry a command toward a safety system. Because custody and configuration are attested from the factory floor onward, diversion or tampering is detectable rather than deniable: a module that is moved, opened, or altered stops matching its own signed history.

Efficiency. The same signed stream that proves a site's output also polices it. An expected-behavior model, the same parametric model on this page, normalized to the site's own environment, flags a module that drifts from what its conditions predict. Cooler fouling, instrument drift, and tampering all surface as the same signal: attested reality diverging from the model. One detector serves both the maintenance plan and the security case.

Independence. A tenant, an insurer, or a regulator verifies a site against the append-only log and its independent off-site witnesses, not against our word. The attested MWh in the table above are checkable by the party buying them. Read how the reactor works and where the program stands with the NRC.