NanoTherma

**Product and the concrete problem it solves.** NanoTherma attacks one of the least-glamorous but most economically consequential inefficiencies in the AI build-out: nearly all of the electricity a data center consumes ends up as low-grade heat that operators spend still more energy to remove and then throw away. As one of the company's founders frames it, "In the age of artificial intelligence, energy is becoming the limiting factor. If we want [AI] to keep growing, we must stop wasting the energy we've already paid for." NanoTherma's answer is a solid-state **thermoelectric generator (TEG)** delivered as thin, lightweight, modular panels that install plug-and-play into existing server racks and exhaust paths, sitting between the hot and cold zones of a facility. Because the panels contain no liquids, pumps, or moving parts, the pitch is that they can be retrofit into operating sites with minimal disruption and no re-plumbing. The panels convert the temperature difference across them directly into usable electricity that can be fed back into the facility, nudging down its Power Usage Effectiveness (PUE), operating cost, and carbon footprint. The company's own stated figures — which should be read as vendor claims pending independent validation — are roughly 3-10% heat-to-power recovery, up to ~10,000 kWh generated per rack per year, and on the order of ~$2,000 in annual savings per rack. Its tagline compresses the thesis: "Harvest the Heat. Power the Future."

**Core technology and how it actually works.** Thermoelectric conversion via the Seebeck effect is old physics; the hard part is doing it usefully at the *small* temperature differentials that real infrastructure produces. Server-exhaust heat is "low-grade" — often only tens of degrees above ambient — and conventional bulk thermoelectric modules convert such gradients at efficiencies too low to justify their cost. NanoTherma's differentiation, per company and press descriptions, is a proprietary panel engineered to operate efficiently across a low differential of roughly **10-15°C**, built on advanced thermoelectric **nanomaterials and two-dimensional materials** produced with **semiconductor-compatible manufacturing** so that the panels can, in principle, be made at scale and low cost rather than as exotic lab devices. The self-optimizing, no-moving-parts design is meant to adapt to varying thermal conditions and to be maintenance-light — attractive properties for uptime-obsessed data centers. The credibility hinges on two intertwined questions a diligence process must answer with data: (1) how much net electricity the panels actually deliver at data-center-realistic gradients after parasitics, and (2) whether the manufacturing route yields a cost-per-watt that produces a defensible payback. The founders' materials-science and semiconductor-process backgrounds are directly on point for both, but the public record does not yet contain third-party efficiency or cost validation.

**Market, customers, and go-to-market.** The go-to-market wedge is deliberately narrow and timely: AI and high-performance-computing data centers, whose power draw and heat rejection are exploding and whose operators are under acute pressure on energy cost, grid interconnection, and sustainability reporting. A per-rack, modular, retrofit form factor is a sensible land-and-expand motion — a customer can trial a few racks and scale by rack count rather than committing to a facility redesign. Beyond data centers, NanoTherma names a broad secondary pipeline wherever large low-temperature heat streams are routinely discarded: **desalination plants, heavy industry, factories, ships, and large vehicles**. That list matters strategically because it maps the same core panel onto water-security infrastructure (desalination) and mobile/marine platforms. The most concrete near-term go-to-market asset is not a signed customer but a channel: by winning the Access innovation competition, NanoTherma secured a structured pathway to pilot with the engineering teams and operating sites of the **Shapir** and **Baran** groups — large Israeli infrastructure, engineering, and construction players — which is exactly the kind of real-world deployment access an unproven hardware startup most needs.

**Traction, funding, and third-party validation.** NanoTherma is extremely young and the honest characterization is pre-commercial. It was **founded in February 2026** and, as of mid-2026, has not disclosed a priced venture round; public sources do not report a funding total, revenue, or customer contracts, and those should be treated as unconfirmed rather than assumed. The single most substantive external validation to date is winning the **third annual Access innovation competition** run by the innovation division of the **Shapir Group**, where the judges reportedly selected NanoTherma unanimously on the basis of near-term implementation feasibility, clear market demand, team expertise, and environmental potential — and awarded a support program pairing the company with Shapir and Baran engineering teams and access to live sites. This is meaningful for an early hardware company: it is a route to design-partner deployments rather than a cash headline. The realistic read is that NanoTherma sits at the idea-validated, seeking-first-deployments stage; the near-term milestones that would convert interest into evidence are an independently measured pilot, a disclosed funding round, and a named data-center or infrastructure customer.

**Founders and team background.** The company was co-founded by **Dr. Nimrod Gazit** (CEO) and **Dr. Roi Levi** (CTO), both materials scientists whose backgrounds span **Intel, NVIDIA, and deep-tech startups**. That pedigree is unusually well matched to the specific problem: the technical crux is nanomaterial/2D-material thermoelectric engineering plus semiconductor-compatible fabrication, and both large-scale semiconductor manufacturing (Intel) and high-performance-compute thermal reality (NVIDIA) are directly relevant reference points. The obvious open questions are the ones facing any two-person deep-tech founding team: depth of the broader engineering and commercial bench, whether the founders have prior company-building (versus research/industry) experience, and the ability to hire the manufacturing, reliability-engineering, and enterprise-sales talent that a hardware-in-the-datacenter business demands. Headcount and the full team are not publicly documented and should be verified directly.

**Competitive dynamics.** NanoTherma competes on two fronts. (1) Against the *incumbent approach of doing nothing* — data centers today simply reject waste heat — its hurdle is proving a payback attractive enough to overcome procurement inertia and any perceived airflow/thermal risk to precious compute. (2) Against other **data-center energy-and-heat plays**: liquid- and two-phase-cooling vendors (e.g. Israel's ZutaCore, and global players) *remove* heat more efficiently but do not convert it to power; waste-heat *reuse* schemes typically pipe hot water to district heating rather than generating on-site electricity; and rival thermoelectric-materials efforts (academic spinouts and specialty firms globally) chase the same Seebeck economics. NanoTherma's plausible edges are: (i) a low-differential-optimized panel aimed squarely at server-grade temperatures where generic TEGs fail; (ii) a plug-and-play, no-fluids retrofit form factor that lowers adoption friction; (iii) semiconductor-compatible manufacturing as a cost-scaling story; and (iv) founder domain fit. The countervailing reality is that thermoelectrics have a long history of promising more than the thermodynamics and unit economics ultimately deliver, so the moat is unproven until measured in the field.

**Defense, security, and resilience dual-use relevance.** The dual-use case is genuine but adjacency-grade and pre-fielded, and should be stated with calibration. The core capability — compact, solid-state, no-moving-parts conversion of low-grade waste heat into electricity — maps onto well-documented defense and resilience priorities: reducing fuel-logistics burden by recovering energy from generators, vehicles, and ships is a long-standing military interest (fuel convoys are a casualty and cost driver), and silent, maintenance-light power generation is attractive for forward or expeditionary use. NanoTherma explicitly lists **ships and large vehicles** among target platforms, and its **desalination** application touches water-security infrastructure, while the data-center use case is itself part of AI-compute and critical-infrastructure energy resilience. The honest calibration: these are enabling adjacencies served through the same commercial panel, not a demonstrated, qualified, or fielded defense capability, and there is no disclosed military customer, ruggedization, or certification in the public record. Dual-use here is credible as *energy-resilience and logistics-efficiency infrastructure*, and its strategic weight will rise only if the technology is validated and then hardened for those environments.

**Growth stage, trajectory, and key diligence risks.** NanoTherma reads as a genuinely **early-stage** company — founded February 2026, pre-priced-round, pre-fielded-product — whose appeal is a timely thesis (AI's energy bottleneck), a credible founder-problem fit, and a concrete channel into infrastructure pilots via the Access/Shapir win. The trajectory to watch is the classic deep-tech hardware gauntlet from lab claim to validated field pilot to manufacturable, bankable product. The key diligence risks are, first and foremost, **physics-and-economics risk**: thermoelectric conversion at 10-15°C differentials is intrinsically low-efficiency, and the stated 3-10% recovery and per-rack yields require independent measurement before they can be relied on; second, **cost and manufacturing risk** — the whole thesis depends on a semiconductor-compatible route to an attractive cost-per-watt at volume; third, **adoption-friction risk** — data-center operators are conservative about anything touching airflow and uptime around expensive GPUs; fourth, **funding and team-scaling risk** for a two-founder company with no disclosed round; and fifth, **disclosure risk** — headquarters, headcount, funding, and customers are not publicly confirmable. Progression from here would be evidenced by an independently verified pilot, a disclosed financing, named design-partner deployments, and quantified payback data.