The Research Sandbox is building a hybrid AI + physics platform to predict weather at 10-100 meter resolution, starting with hyper-local wind forecasting for wind-energy operators penalized for over/under-prediction on node-based grids. The differentiator is less accuracy than computational efficiency — simulations that run faster than the event — powered by rare VFX-world HPC talent (Internal).
The founders are genuinely elite: Millicent Maier (Pomona '05, Oxford Astrophysics PhD) led the research lab behind the Avatar simulations at Wētā FX, with GTM co-founder Inge Rademeyer. But this is a science project today: no product, ~9-12 months to a competitive model, no round, and a crowded, well-funded field (Tomorrow.io >$1B, WindBorne, Meteomatics) already shipping sub-kilometer AI wind forecasts (Internal/Web). StoryHouse's read is honest — not one to pursue and not to lead, but credible scientists worth staying in touch with or referring to a 1517 program.
Weather intelligence is a real, high-single-digit-growth market ($7.8B in 2026 → $13.5B by 2033) with the fastest growth in the exact hyper-local micro-weather niche Sandbox targets (~12% CAGR to $2.5B by 2030) (Web). But it is now a well-capitalized, crowded field: Tomorrow.io (>$1B, ~$100M ARR), WindBorne (Khosla-backed, out-forecasting government agencies), and Meteomatics are already delivering sub-kilometer AI wind forecasts to energy operators today, and Vaisala's $70M June-2026 acquisition of Atmo shows incumbents are buying AI-forecasting talent rather than waiting (Web). Against that backdrop a pre-product, pre-round NZ team faces steep catch-up and capital-intensity risk, even with genuinely elite simulation credentials (Web + Internal).
The addressable weather-intelligence market is real and growing high-single-digits, and the sharpest growth is in the smart micro-weather / hyper-local segment (~12% CAGR) the company is targeting. But the specific SOM (wind-energy operators penalized on node-based grids in NZ/AU) is a narrow beachhead relative to the headline TAM (Internal).
| Player | Positioning | Funding / Stage | Edge vs. them |
|---|---|---|---|
| The Research Sandbox | Physics+AI, GPU-native, 10-100m resolution; VFX-HPC talent from Wētā FX | Pre-round | — |
| Tomorrow.io | Weather-intelligence unicorn; proprietary satellites + AI, enterprise/gov | $508M raised, >$1B val, ~$100M ARR | Scale, satellite data moat, enterprise GTM |
| WindBorne Systems | AI WeatherMesh model + balloon constellation; 1km local forecasts in seconds | $15M Series A (Khosla), ~$85M val | Proprietary in-situ data + fast AI model, out-forecasting agencies |
| Meteomatics | EURO1k 1km physics NWP + ML; +50% wind-forecast accuracy for operators | Growth-stage (Europe) | 1km operational model live, existing energy-trader customers |
| Amperon | Asset-specific wind-gen forecasts, physics+ML, hourly retrain | VC-backed | Direct energy-market GTM, live utility customers |
Moat: Proposed moat is computational efficiency (simulate faster than the event) plus rare VFX-world HPC/simulation talent, but every named competitor is already shipping sub-km AI forecasts while Sandbox has no product.
Most probable exit is acquisition by a weather-intelligence major or industrial/instrument player. Live comps: Vaisala acquired AI-weather co Atmo for $70M (June 2026); Tomorrow.io consolidating the space at >$1B; DTN, The Weather Company, StormGeo, AccuWeather are all active acquirers; energy-trading firms and utilities are plausible strategic buyers. Base case is a modest ~$50-100M talent/tech M&A (Atmo-style), which for a pre-seed check could return well but is unlikely to be fund-returning unless the platform generalizes beyond wind into ag/aerospace/wildfire (Web + Internal).
Strong intro call with Millicent and Inge on a hyper-local (50-100m) wind-forecasting platform for energy operators. Credible scientists; very early (no product, no round).