One audited model, five rungs, one named invention per rung. The commercial band is $30–45/MWh; today's best coupled result is $61.65. This is the arithmetic of the gap — and the company that closes it rung by rung.
The ladder below is five project definitions inside one recomputed, audited cost model — not additive savings stacked on one plant. Model rules are held constant across every row: 25-year life, 90% availability, 8% real discount, 5% gross auxiliaries, 2%/yr fixed O&M of capital, $8/MWh variable. The arithmetic was independently recomputed from the stored physical trajectories and published as a QA record model. Geology and costs remain unvalidated — that is what field runs are for — but no number here is hopeful. Each row is a requirement with a bench gate, and each transition names the invention that must deliver it.
Read the rungs as a work order. Row 2 is E1 Shift-Latch — commandable reservoir reconfiguration puts more of each well pair's rock to work and takes subsurface cost from $37m to $25m per pair. Row 3 is access and survival at 300 °C: E4 Ember-BHCT drilling, E5 Flex-Seal integrity through cycling, E2 Clear-Chemistry keeping the hot branch conductive. Row 4 is E3 Conversion — a factory-built modular flash+binary block at the $757–$1,012/gross-kW installed allowances; the corrected invention target at ladder duty is $890–1,145/gross-kW (installed scope). Row 5 is all five, with Relay controls holding the margins — and it stays a spec: at real conversion cost the 320 °C row is the $45.72/MWh R5C band product (red-team repair R4).
The margin view is honest but thin: the spec row survives 0.90% of lifetime-electricity shortfall or 0.91% of capital overrun (~$0.68m headroom); the 2026-09-30 cushion rule reads $27/MWh nominal at ≈$831/gross-kW. The 2026-10-02 red team confirmed that fragility (R5) and replaced margin discipline with box discipline. Fractions of a declared stress box are shares of declared input space — never probabilities.
And the stress is quantified, not asserted: 20,000 cases across a declared
uncertainty box under pre-registered falsifiers simulated.
The reinforced run reproduces the $29.80 row exactly, then re-prices at real conversion
cost: R5C, the 320 °C row, is $45.72/MWh nominal and holds $30 in
0% of both boxes; R6S, superhot at today's conversion class, is $37.19/MWh
and holds the $45 band across 77–99.9% of declared stress boxes; R6D is the $30 design
spec — $24.58/MWh nominal at ≈$970/gross-kW, holding $30 in 80% of the
declared box, 99.9% of the contract box (fixed-price EPC + long-tenor debt, an
assumption-narrowed box, not a guarantee). The two falsifiers the 2026-09-30 campaign
breached — the thin cushion and the discount-rate lever — stay published as redesign
levers; de-risked long-tenor capital sits inside the engineering plan
(qa/ladder-reinforced-2026-10-02.json, qa/ladder-stress-2026-09-30.json).
The size of the required invention is audited — honest about what optimization alone cannot do. At the coupled target's electricity, discount and operating rules, $30/MWh needs total capital of $40.68m where today's design carries $99.21m — a 59% reduction. Make the subsurface work free and the model still prints $41.64/MWh. Eliminate all circulation pumping in a thought experiment and it prints $53.45/MWh. With today's $1,800/gross-kW surface cost the arithmetic floor is $34.48/MWh even with free wells and zero pumping. The gap therefore closes only through the named inventions — more electricity per well pair, cheaper access to hotter rock, and conversion equipment at half the installed cost — exactly the series we have designed and virtually qualified.
Nothing in this model is measured plant performance; the spec rows are requirements, and each one expires unless its bench gate closes. If a gate fails, the published failure re-prices the row; the ladder re-computes.
Three demand signals point the same way. Capital arrived at the superhot frontier in 2026: Quaise $180M (with Nabors), Mazama $135M (Khosla), Hephae $17.8M Series A, Fervo's Cape Station first power — all verified press. Load arrived behind it: data-center electricity grew 17% in 2025 toward ~945 TWh/yr by 2030, and hyperscalers are contracting firm geothermal directly (Google–Fervo framework up to 3 GW through 2033; Meta–Sage up to 150 MW — frameworks and partnerships, cited as such). And the public target arrived: $45/MWh by 2035 (DOE Enhanced Geothermal Shot).
Every one of those programs drills wells whose transient and hot-section behavior nobody can currently measure. Demand for instrumentation and qualification scales with wells drilled regardless of which developer wins — the picks-and-shovels position in a category being dug by four well-funded teams. The layer those teams share is unfunded: $100M+ rounds price plants; the tools and standards layer is exactly where a defensible component-and-service business lives.
| Line | What sells | When it turns on |
|---|---|---|
| Instruments | Cool-Brain profilers and the series — the E1 latch arrays, E5-rated well-integrity engineering, the Ember-BHCT control system | on the bench record (R3) |
| Witnessed qualification | duty-basis testing and witnessed records at 450 °C-class — the service every tool maker in the segment needs and nobody offers | on the first record, Q1 2027 |
| Data & basis | the qualification dataset and the standard itself; duty-cycle rating services (E5's product regardless of any patent posture) | as records accumulate |
Product lines in the field: Relay (heat routing and control — the opening move that holds auxiliaries down), Shift (reversible flow treatment — the reservoir repair and redirect), Flex (thermal movement — the moving pressure boundary). The gigascale ladder converts lines into manufacturing: G1 bench proof → G2 a three-well repeatable-service program (spec row $55.05) → G3 the first 45–50 MW block at the band (R6S $37.19/MWh nominal), built from factory flash+binary power-block modules → G4 the fleet at the $30 design spec (R6D $24.58 nominal), where 25-year operating records make the position undroppable.
| Rung | Size | What it buys |
|---|---|---|
| Sprint (open now) | $10–30k | prototype materials, Tier-0/1 fixtures — terms memo ready |
| Pre-seed (this round) | $2–3M lead $1–1.5M | the R1–R3 numbers, the Rev B field string, design-partner wells, qualification-service stand-up |
| Non-dilutive (parallel) | $300k–$4.5M class | DOE UTR via a university PI, the next ARPA-E FOA (its QA-infrastructure category is our lane), EU SHiFT via the Iceland labs |
| Series A | $10M+ class | repeatable service scaled: field pilots, the manufacturing line |
| Project finance | $100M+ / block | fleet rungs at measured LCOE |
Use of proceeds is named milestones, never runway: (1) chain proof → first hot data, (2) the 500-hour qualification record at 450 °C, (3) design-partner wells with the Shift latch and Cool-Brain strings, (4) first commercial subsystems, (5) the witnessed-qualification service in revenue. The schedule: R1 within 30 days, R2 within 90, R3 in Q1 — SGW week (Feb 8–10) is the coming-out venue for the record and the recruiting.
The honest status line, printed because diligence will find it anyway and it is better as our sentence than theirs: zero measured duty hours and zero meters today — design, software, digital qualification and audited economics are done, and that is precisely what this round converts into measurements. Every bench run judges a frozen prediction registered before it. Bench number, publication and named attachments land inside the schedule above. Money conversations open on the demo, the audited economics and this results calendar — and every number in the room keeps its label.
The ask: lead $1–1.5M of a $2–3M pre-seed. First step is thirty minutes — we will bring the model, the gates and the falsifiers, and you can attack all three.