Tekcapital takes 51% stake in geothermal AI data center firm

Tekcapital formed Vesari Inc. to commercialize IP for geothermal-powered AI data centers, taking a 51% stake with no cash consideration while Clifford Gross retains 49% and will contribute 11 patent applications. The model targets behind-the-meter, off-grid AI infrastructure using geothermal power and satellite connectivity, addressing a data-center electricity market that already consumes about 4-5% of U.S. power and could rise to 9-17% by 2030. The deal was approved by independent directors and deemed fair and reasonable, but the immediate market impact is likely limited.

Analysis

This is less a revenue event than an embedded call option on the AI power bottleneck. The strategic edge is not the patents themselves but the combination of baseload geothermal, behind-the-meter siting, and non-fiber connectivity, which could create a template for compute deployment where grid interconnection queues and transmission constraints are the real gating item. If that model works, the first-order beneficiaries are not only the venture vehicle but also geothermal developers, low-latency connectivity providers, and the few infrastructure names with exposure to off-grid or microgrid-enabled data center builds.

The second-order effect is competitive pressure on traditional colocation and cloud operators: if compute can be sited where power is cheapest and most reliable, the economics of conventional data center footprints in congested U.S. hubs weaken over a multi-year horizon. That said, the near-term risk is execution, not concept—geothermal resource quality, drilling capex, permitting, and satellite connectivity reliability can each kill IRR before the AI thesis matters. The market will likely overvalue the narrative in the next few weeks if it extrapolates from incorporation to commercial scale; the real re-rating window is 12-24 months, after the company proves it can secure land, wells, and anchor customers.

The contrarian angle is that this may be a financing story disguised as an infrastructure innovation. A 51/49 governance structure with the IP assignment still pending leaves meaningful uncertainty around control, dilution, and monetization, and the economics of satellite-backed compute are unproven versus fiber-backed sites. If hyperscalers conclude the network layer is too latency-sensitive or too expensive at scale, the addressable market shrinks to edge/AI inference rather than training workloads, which would materially cap upside.

For public-market positioning, the cleaner expression is to own the pick-and-shovel winners rather than the small-cap vehicle itself. The asymmetry is strongest in names that benefit from rising power demand and constrained supply even if one novel project fails, because the broader thesis is intact across the sector. Any enthusiasm here should be paired with discipline: treat this as a catalyst for volatility, not a durable fundamental rerating yet.

Key Decisions for Investors

• Long IREN or CIFR on any 3-5% pullback over the next 1-2 weeks; these names have cleaner exposure to power-constrained AI infrastructure and should outperform if investors rotate toward 'energy-first' compute narratives.
• Long RGTI / short a high-multiple colocation proxy for 1-3 months as a thematic pair: geothermal-adjacent power innovation versus conventional data-center capex, with the trade monetizing if the market starts pricing differentiated access to cheap baseload energy.
• Avoid chasing the venture vehicle on the headline; if you must express it, use a very small, tactical position only after confirmation of site control and customer LOIs, because the first commercial milestone is likely 12-24 months away.
• Consider a call spread on ORA or other geothermal-exposed infrastructure names for 6-12 months; upside comes from the market re-rating geothermal as an AI-enabling baseload asset, while downside is limited if project execution stalls.
• If power-price-sensitive data center beneficiaries sell off on the news, buy the dip in the strongest balance sheets; the broader AI power shortage thesis is still secular, and one prototype does not resolve the supply-demand mismatch.