Kairos Power gets up to $100M from Samsung group to build nuclear reactor for Google

Kairos Power selected Samsung C&T to support construction of its 50MW Hermes 2 demonstration reactor for Google, with Samsung providing up to $100 million of equity and in-kind engineering services, including a $70 million equity investment. Kairos targets Hermes 2 operations by 2030, delivering the first 50MW under Google’s broader agreement for roughly 500MW of nuclear capacity by 2035. Samsung’s nuclear construction experience could help Kairos accelerate follow-on reactor deployment to meet AI data-center power demand.

Analysis

The investable read-through is not near-term nuclear generation economics; it is Google’s willingness to underwrite firm-power capacity before commercial proof, reducing the financing hurdle for advanced-reactor vendors. For GOOG, a 50 MW initial tranche is immaterial to consolidated earnings, but it creates an option on reducing data-center power-delivery bottlenecks in constrained markets. The more relevant 6-18 month valuation implication is whether this supports an AI-capex narrative in which power access—not chips or buildings—is the binding constraint; that would favor hyperscalers with diversified long-duration power procurement over smaller cloud competitors reliant on merchant-grid availability.

Samsung C&T’s equity-plus-services structure is a useful financing signal but not validation of reactor economics. It aligns a proven EPC counterparty with schedule execution, yet transfers little of the key risks: first-of-a-kind construction, TRISO fuel availability, licensing amendments, and the ability to replicate units quickly enough to meet the larger procurement ambition. A single schedule slip would have limited direct financial impact on GOOG but would weaken the market’s assumption that advanced nuclear can solve AI power shortages before the mid-2030s, benefiting gas-turbine, grid-equipment, and conventional generation alternatives.

Consensus may over-credit nuclear announcements as immediate relief for data-center load growth. Over the next 1-3 months, power scarcity should remain a regional transmission, interconnection, and dispatchable-generation trade rather than a public-equity advanced-nuclear trade. The contrarian positive is that customer-backed offtake plus EPC capital can establish a repeatable project-finance template; confirmation via a second hyperscaler contract or fixed-price EPC terms would materially improve odds of a scalable fleet model.

Key Decisions for Investors

• No directional GOOG trade on this development alone: the capacity is too small and too distant to alter FY2026-27 earnings. Reassess if management quantifies power-constrained AI capacity, commits incremental nuclear offtake, or a reactor delay forces higher-cost replacement power.
• Maintain a 6-18 month long ETN / short hyperscaler-relative basket focused on power-constrained cloud exposure only after verifying regional data-center interconnection exposure; grid build-out captures nearer-term spend while advanced-reactor output remains unproven.
• Use GE Vernova (GEV) as the liquid watch-list beneficiary of a nuclear-timeline disappointment: a 12-24 month delay in advanced-reactor commissioning would reinforce demand for gas generation and grid equipment. Enter only on evidence of capacity-contract repricing or raised turbine backlog guidance; key risk is faster renewable-plus-storage deployment.
• Treat Samsung C&T’s $70M equity commitment as a private-market diligence datapoint, not a sector signal. Watch for fixed-price versus cost-plus EPC allocation, fuel-supply contracts, and construction milestones; failure to disclose these should cap confidence in replication economics.