In 2015, IBM paid $2 billion for The Weather Company’s digital assets, believing weather data was the next frontier of enterprise AI. In February 2024, IBM sold those assets to private equity firm Francisco Partners for an undisclosed sum — almost certainly at a significant loss. The divestiture was not a one-off misstep. It was a structural signal. In the twelve months that followed, Google embedded AI weather forecasts into Search for 5 billion users — for free. ECMWF open-sourced its AIFS model weights. Nvidia open-sourced Earth-2. Meanwhile, Tomorrow.io raised $500 million, hit $1 billion valuation, and is building its own satellite constellation to bypass traditional data providers entirely. The commercial weather data middleman — the business that aggregates government weather data, adds a layer of processing, and sells API access — is being squeezed from above by free AI and from below by vertically integrated startups. The Weather Company problem is not about one company. It is about an entire business model.
The commercial weather data business has operated on a simple model for decades: ingest free government data (from NOAA, ECMWF, and national meteorological services), run proprietary processing and presentation layers on top, and sell the result via enterprise APIs, consumer apps, and advertising. The Weather Company, AccuWeather, DTN, and others built substantial businesses on this model. The Weather Company alone reached 415 million monthly users and served over 2,000 enterprise clients.[1]
That model is now being dismantled from two directions simultaneously.
From above: Google now provides AI weather forecasts to 5 billion users through Search, Gemini, Pixel Weather, and Maps — all for free, powered by WeatherNext 2. For enterprises and developers, Google offers the same data through Earth Engine, BigQuery, and Vertex AI. ECMWF released its AIFS model weights under a permissive licence, allowing anyone to run a world-class AI weather model. Nvidia open-sourced Earth-2 for governments and businesses. When the world’s best forecasts are free, the value of repackaging government data collapses.[4][5][6]
From below: Tomorrow.io raised $175 million in February 2026 at a valuation exceeding $1 billion, bringing total funding to approximately $500 million. The company is building DeepSky — the world’s first AI-native weather satellite constellation — to generate its own proprietary atmospheric data rather than depending on government sources. With ~$100 million ARR, a Palantir partnership, US Air Force contracts, and 250+ enterprise customers including Uber, Delta, and Ford, Tomorrow.io represents the new model: own the observation infrastructure, own the AI, own the customer relationship. Traditional data middlemen are cut out entirely.[3]
In the middle: IBM’s divestiture is the diagnostic event. IBM purchased The Weather Company’s digital assets for $2 billion in 2015, believing weather data would power its Watson AI platform. By 2023, IBM was shopping the assets. Francisco Partners, a PE firm specialising in technology turnarounds, completed the acquisition in February 2024. The Weather Company is now a standalone company, led by CEO Sheri Bachstein, with a mandate to expand beyond forecasting into health and well-being. But the structural question remains: when Google gives weather away for free and Tomorrow.io owns its own satellites, what does a weather data middleman sell?[1][2]
| Dimension | Evidence |
|---|---|
| Revenue / Financial (D3)Origin · 72 | IBM $2B acquisition → PE divestiture (2024). Weather Company now standalone under Francisco Partners. AccuWeather and DTN face commoditisation. The revenue cascade is structural. IBM’s divestiture signals that the weather data aggregation model no longer generates the returns that justified a $2 billion acquisition. Meanwhile, Tomorrow.io’s $1B+ valuation and ~$100M ARR show that the market is not shrinking — it is migrating. Value is moving from data aggregation to data ownership (Tomorrow.io’s satellites) and platform integration (Google’s Search/Maps). The AI weather modelling market ($1.1B → $7.2B by 2033) is growing. The legacy segment of that market is not.[1][3] |
| Quality / Product (D5)L1 · 65 | The product differentiation of legacy providers is collapsing. When ECMWF open-sources AIFS and Google delivers hourly resolution ensemble forecasts to 5 billion users for free, the traditional value proposition — repackaging NWS/ECMWF data with a proprietary processing layer — becomes a commodity. The new quality bar is set by AI-native products: Tomorrow.io’s impact-based forecasting, Google’s FGN ensemble approach, and ECMWF’s 51-member AI ensemble. Legacy providers must either build their own AI capabilities or accept commoditisation.[4][5] |
| Operational (D6)L1 · 68 | The infrastructure paradigm is shifting from data aggregation to data ownership + AI inference. Tomorrow.io’s DeepSky satellite constellation represents the new model: proprietary observation infrastructure feeding AI-native models. Google’s single-TPU inference makes enterprise-grade forecasts nearly free to produce. Legacy providers operate on an infrastructure model built for ingesting, processing, and redistributing government data. That model requires constant updating, licensing agreements, and compute infrastructure that AI approaches bypass entirely. The operational transformation is not incremental — it is architectural.[3] |
| Customer / Market (D1)L1 · 62 | Enterprise weather data buyers are shifting. Tomorrow.io serves Uber, Delta, Ford, National Grid, and the US Air Force. Google’s Vertex AI and Earth Engine provide enterprise-grade access to WeatherNext 2 data. The customer migration follows the value: from generic repackaged data to AI-native, impact-based, industry-specific intelligence. Legacy providers retain customer inertia — existing integrations, familiar interfaces, contractual lock-in — but the switching cost decreases as AI-native alternatives mature and as Google’s free tier captures the commodity layer.[3][4] |
| Employee / Talent (D2)L2 · 55 | The talent drain follows the capital. Tomorrow.io has 223 employees and is hiring aggressively across engineering, sales, and C-suite. Google DeepMind’s weather team attracts top atmospheric science and ML talent. ECMWF’s Anemoi open-source framework creates a community of contributors. Legacy weather data companies face a talent squeeze: the ML engineers they need to compete are drawn to better-funded, higher-growth AI-native companies. IBM’s decision to divest rather than invest was itself a talent signal — the company concluded weather was not worth the R&D commitment.[3] |
| Regulatory / Governance (D4)L2 · 38 | The regulatory dimension is minimal but emerging. There is no regulatory barrier protecting legacy weather data providers from free AI competition. In the US, government weather data is a public good — NOAA data is freely available by statute. This means legacy providers have no regulatory moat. The potential regulatory tailwind for legacy is in accountability: if Google’s free AI forecasts prove unreliable for critical applications (aviation, emergency management), enterprise buyers may value the liability and SLA commitments that established providers offer. But this has not materialised yet. |
-- The Weather Company Problem: 6D Diagnostic Cascade
FORAGE weather_data_middleman_disruption
WHERE incumbent_divestiture = true
AND acquisition_price > 1_000_000_000
AND free_ai_competitors >= 3
AND ai_native_challenger_valuation > 1_000_000_000
AND open_source_models_available = true
AND proprietary_observation_infrastructure = true
ACROSS D3, D5, D6, D1, D2, D4
DEPTH 3
SURFACE weather_company_cascade
DIVE INTO middleman_squeeze
WHEN free_from_above AND vertical_from_below AND incumbent_divests
TRACE commoditisation_cascade
EMIT diagnostic_signal
DRIFT weather_company_cascade
METHODOLOGY 85 -- IBM $2B investment, Weather Company 415M users, 2000+ enterprise clients, established API business
PERFORMANCE 35 -- Divested to PE, commoditised by Google free, bypassed by Tomorrow.io satellites, no AI model moat
FETCH weather_company_cascade
THRESHOLD 1000
ON EXECUTE CHIRP diagnostic "IBM paid $2B for The Weather Company in 2015 and divested to PE in 2024. Google gives AI weather to 5B users for free. ECMWF and Nvidia open-sourced their models. Tomorrow.io raised $500M and is building its own satellite constellation. The commercial weather data middleman model is broken. The value has migrated from data aggregation to data ownership and platform integration."
SURFACE analysis AS jsonRuntime: @stratiqx/cal-runtime · Spec: cal.cormorantforaging.dev · DOI: 10.5281/zenodo.18905193
When upstream data becomes free and downstream distribution gets captured by platforms, the middleman gets compressed. Stock photography agencies faced this with smartphones and Instagram. News aggregators faced this with Google and Facebook. Travel agents faced this with Kayak and Google Flights. Weather data providers are now facing the same structural compression. IBM’s divestiture is not a company-specific failure — it is the diagnostic event that confirms the pattern has arrived in meteorology.
Tomorrow.io’s $1B+ valuation is not built on repackaging government data. It is built on proprietary satellite infrastructure (DeepSky), proprietary AI models, and direct enterprise relationships. The lesson is the same one that Google learned with Maps and Street View: when you own the observation layer, you control the value chain. Traditional weather providers license government data. AI-native providers generate their own. The defensibility gap is unbridgeable.
Google’s WeatherNext 2 is free for consumers and available at cloud commodity pricing for enterprises. ECMWF’s AIFS weights are free to download. Nvidia’s Earth-2 is open-source. When three of the world’s most capable weather AI providers choose to give the technology away, it is not philanthropy — it is platform strategy. Google monetises through Search advertising and cloud services. ECMWF is funded by 35 member states. Nvidia sells the GPUs. The forecast itself has zero marginal cost. Legacy providers cannot compete on price with free.
The commercial weather providers that survive will be those that move up the value chain from data to intelligence. Tomorrow.io’s impact-based forecasting — not just what the weather will be, but what it means for your specific operations — is the template. AccuWeather’s health and well-being pivot is another. DTN’s industry-specific solutions for agriculture, energy, and maritime represent a third. The common thread: domain-specific, decision-support intelligence that raw forecasts cannot provide. The data layer has been commoditised. The intelligence layer has not — yet.
One conversation. We’ll tell you if the six-dimensional view adds something new — or confirm your current tools have it covered.