Firm clean power × AI compute
AI is running out of power before it runs out of chips. Archegon builds data centres on top of geothermal heat — firm, 24/7, behind the meter, and clear of a grid queue measured in years.1
The thesis is written and the numbers are on the table. What it needs now are the people who have done this before.
The thesis
A new large grid connection takes years. That queue, not chip supply, is what now sets the pace of the AI buildout — and whoever controls firm power controls the pace. Inference is a flat, around-the-clock load, which is exactly the shape geothermal delivers best.34
Meanwhile some of the hottest rock on Earth sits idle — not because the engineering failed, but because nobody put the demand next to it. Transmission is the expensive part. Archegon moves the data centre to the heat instead of moving the electricity to the data centre.56
Capacity-factor and market statements are indicative and drawn from public benchmarks. They are not financial projections.
The projects
A preliminary 100 MW Phase 1 concept anchored by firm geothermal power. Candidate routes include Central North Island geothermal proximity and colder-climate Southland infrastructure paths, subject to site, power, fibre, consent, and partner diligence.1112
Download New Zealand summaryIllustrative: Wairakei geothermal infrastructure in the Taupo volcanic zone; not an Archegon site or partner asset.
A frontier Cooper Basin route combining enhanced geothermal systems, solar, storage, dry cooling, and resilient backhaul. The central diligence question is whether the subsurface, water, cooling, and customer gates can be de-risked before any larger campus build.513
Download Australia summaryIllustrative: Cooper Creek landscape context for remote Australian infrastructure; not the Habanero plant.
Project summaries are generalised from source business plans. Detailed financial models, capital stacks, and performance assumptions are not published here and belong only in a qualified diligence process. Illustrative only; built from public benchmarks; subject to due diligence.
The opportunity
AI compute demand is rising into a power-constrained grid. Speed-to-power is now a strategic advantage for hyperscalers, AI labs, and infrastructure investors.23
The model is under active diligence: capex per MW, contracted revenue, cooling, fibre, water, and power-stack ownership all matter. No return figures are offered on this website.
Year 1
Subsurface review, temperature-at-depth mapping, land and consent pathway, grid and fibre routing, offtake conversations.
Gate A defensible resource case and a site with a credible consent and connection path, or the route is stopped.
FeasibilityPre-drill
Year 2
Drill the initial doublet, stimulate, and measure. This is where subsurface uncertainty is converted into flow-rate and temperature data, or is not.
Gate Measured flow and temperature sufficient to support a commercial power block. The single largest technical risk in the programme.
DrillingCapital intensive
Year 3
Commission an initial power block and a small compute hall against it. Prove the closed loop end to end at modest scale.
Gate Sustained generation and a live customer load, with contracted revenue rather than pilot goodwill.
Pilot buildPlant + hall
Years 4–5
Repeat the well pattern, expand the power block, and build out the campus toward the preliminary 100 MW Phase 1 concept.
Gate Repeatable well cost and schedule. Cost per well falling with each pad is what makes the programme financeable at scale.
Project financeDebt-led
Capital profile Equity-led through drilling; debt-led once flow is proven
Indicative phasing only. Capital requirements, capital stack, and any return figures are not published here and belong in a controlled diligence process with appropriately qualified parties. Benchmarks for data-centre and geothermal cost structure are set out in the evidence base. Image: data centre interior, used illustratively. Nothing on this website constitutes an offer or invitation to invest, financial advice, or a financial promotion.
About Archegon
Archegon began with a stubborn question: if power is the real limit on AI, why are we shipping electricity across continents instead of putting the computers where the clean energy already is?
That question became two years of work — geothermal resource, data-centre economics, water, cooling, and the places on Earth where hot rock and cheap land meet. It is no longer a thought experiment. There are two candidate routes, a phased plan, and a published evidence base anyone is welcome to attack.
What Archegon does not have yet is a team. I am one founder with a thesis, and a thesis is not an asset until people who have drilled wells, financed plants, and run halls put their hands on it. That is the gap I am trying to close, and I would rather say so plainly than pretend otherwise.
Who I am looking for
Research and consulting
Why the interconnection queue, not chip supply, now sets the pace — and why compute will move to the energy rather than the reverse.
Read noteArchegon also keeps a research and advisory thread on agentic AI: workflow design, evaluation, handoffs, and governance for practical deployment.
Read noteNo site, no wells, no revenue. What exists is a thesis, two candidate routes, and a sourced evidence base. Judge it on that.
Get in touchGet involved
Sceptics welcome — particularly on the subsurface. If you think the wells will not deliver, that is the most useful conversation I can have. Tell me who you are and which part you would take on.
