My Assessment of the Weather Industry

My overall assessment: once funding tightened, capital in the weather industry moved to upstream observation hardware, value accrued to the downstream users that convert weather into profit, and the software and data services in the middle are squeezed from both sides.

Three segments of the value chain

This post splits the weather and geospatial value chain into three segments:

1. Upstream observation: upper-air sounding, radar, satellites;
2. Data and forecast services in the middle: general-purpose forecasts, data processing, analysis software;
3. Downstream applications: energy trading, insurance, grid dispatch, and other organizations that earn operating income directly from weather information.

When funding was easy, investors accepted the long-term narrative that “the value of data will keep rising”; once funding tightened, investors required existing paying customers. The three segments diverged from there. Upstream observation companies usually hold government contracts and continue to raise money; downstream users internalize weather capability, building their own teams and buying raw data; the middle layer has neither scarce assets nor direct operating income, and is compressed at the same time by open data from public agencies and by open-weight AI models, so its funding rounds were the first to shrink in scale. Software engineers are concentrated in this segment.

China

From one angle, China’s weather industry is closer to a traditional industry that happens to ship internet-style products. The core reason is that the China Meteorological Administration (CMA) and the meteorological departments at each level control the national observation network: by figures from the State Council Information Office in October 2025, the network consists of 9 Fengyun satellites, 842 weather radars, and more than 90,000 surface weather stations, of which about 11,000 are national-level surface stations. The Meteorological Law requires other observing entities to submit their data to the meteorological authorities. Commercial companies can obtain basic data in the open catalogue free of charge; everything else depends on agreements with the meteorological departments or their affiliated units. The meteorological departments also derive part of their income from this data through specialized meteorological services and affiliated enterprises, and compete with commercial companies in the same market (in the 2023 final accounts of the meteorological departments, operating income was about 4.2 billion yuan, roughly one tenth of the year’s revenue; a 2026 industry analysis by Huxiu describes data acquisition costs and competition from affiliated enterprises in more detail).

The funding difficulties I observed among Chinese weather companies began in the summer of 2019. Fundraising by foreign-currency funds contracted sharply in the same period: by Zero2IPO statistics, foreign-currency fundraising in the first 11 months of 2019 fell 51.5% year on year, and RMB funds accounted for 88.8% of total fundraising. Before that, startups with an internet background relied on USD funds to support high valuations and long-horizon growth expectations, and weather companies raised money on the same “internet product” narrative; the RMB funds that took over preferred hard technology, domestic substitution, and policy-backed directions, with clearer requirements on payback period and exit path. The weather business has stable cash flow, slow growth, and sits outside the policy mainline, so neither type of fund was interested. The most public data point from that period: Moji Weather’s ChiNext IPO application was rejected by the Issuance Examination Committee on 11 October 2019. The committee’s concerns were operating-license compliance, user data collection compliance, advertising revenue exceeding 90% of the total, and related-party transactions. After signing a listing tutoring agreement in September 2020, the company did not refile.

In the last two years a number of commercial companies have added observation dimensions through satellite remote sensing and similar technologies: on 30 December 2024, GNSS radio occultation data from the 23 Tianmu-1 satellites of Aerospace Tianmu and the 12 Yunyao-1 satellites of Yunyao Aerospace entered CMA’s operational observing system, the first time commercial small-satellite data has entered domestic meteorological operations. If the environment holds, the commercial weather market in China still has room to grow. Satellite remote sensing is a capital-intensive direction. Tianmu-1 is operated by Aerospace Tianmu, which has a CASIC background; Yunyao Aerospace is privately owned and had launched 47 satellites as of June 2026, 31 of which are in CMA operations. Companies with positive cash flow that do not depend on fundraising are relatively safe in this environment, with a limited valuation ceiling. I am not very familiar with the funding situation of Chinese remote sensing companies, and am only describing the current state.

The mechanism behind tighter funding differs between the West and China: in the West it was rising interest rates; in China it was the withdrawal of USD funds. The outcome is the same: the middle layer that traded funding for growth exits first, and hardware and infrastructure with government or large customers remain. China’s turning point came three to four years earlier than the West’s.

US upstream observation: funding continues

Among the US weather companies I follow closely, those with their own observation hardware have continued to raise money even through these years of high interest rates. Several rounds between 2024 and 2026:

What these companies have in common is existing paying government contracts. WindBorne’s main customers are government agencies; NWS buys its sounding data and assimilates it into GFS, and the Series B funds go to expanding the commercial team serving private customers such as investment funds. Tomorrow.io was founded in 2016 as ClimaCell doing software and APIs. The SPAC listing announced at the end of 2021 was terminated in March 2022. It launched its first own radar satellite in April 2023, then won a US Air Force contract and a NOAA microwave sounder data purchase ($7.3M in June 2026), and in September 2026 selected York Space Systems to build the first 12 satellites of DeepSky. A middle-layer company preserved its ability to raise money by moving upstream into hardware. Rainmaker’s paying customers are state governments: the 2025–2026 season pilot in the Bear River basin was funded by the Utah Division of Water Resources and the Idaho Water Resource Board, capped at $3.496M and $0.95M respectively. Climavision’s funding came in June 2021 (a $100M strategic investment from TPG Rise Fund), outside the table’s time range, but its revenue sources since then follow the same pattern: its radar data has been integrated into the NWS AWIPS II warning system, NOAA renewed its Mesonet contract in September 2024, and it has public-private partnerships with county governments: the county provides a site such as a water tower or rooftop, and Climavision covers construction and maintenance and provides data access. As of September 2026 it operates 31 radars in 15 states; of the 14 radar blind spots in Texas, 6 had been filled as of March 2026, and the 7th came online in August of the same year.

In terms of funding direction, rounds in the hundred-million-dollar range are concentrated in observation hardware and the AI models that go with it. The pure-software side of weather and geospatial data processing and analysis, which is what I am better at and more familiar with, has already lost many positions to AI in recent years. That is my personal observation.

Upstream raising money does not mean upstream revenue is secure. Hardware build-out is measured in years and constrained by supply chains: Tomorrow.io launched its first satellite in April 2023 and did not have a NOAA data purchase contract until June 2026; a radar network has to negotiate sites and permits station by station. The demand-side cycle is set by government budgets, capital markets, and geopolitics, each independent of the build-out cycle. When the two fall out of step, a hardware company’s revenue fails to materialize. Two kinds of mismatch have already occurred:

1. Government budget cycle: GeoXO’s instrument contracts were signed in 2023–2024; after the change of administration in 2025, three instrument contracts were cancelled and the contractors’ signed orders were voided. Details in the next section;
2. Capital market cycle: Spire Global holds NOAA radio occultation data contracts, yet its 2024 annual report still disclosed a going-concern doubt. It only paid off about $102M of debt after selling its maritime business in April 2025, and its 2026 guidance is still unprofitable; in September of the same year it won a new $33.16M NOAA order. Tomorrow.io’s SPAC, announced at the end of 2021, was terminated after the market turned in March 2022.

A third kind that has not occurred yet but holds just as well is geopolitics: Japan’s SAR satellite orders are driven by defense demand, and the persistence of that demand depends on the defense budget. Details in the Japan section.

Deploying an own constellation or radar network is expensive and slow; hardware delays, launch failures, and slower-than-expected customer adoption all consume cash. Tighter funding means harsher terms on later rounds, so these companies need to generate commercial revenue within two to three years, and any of the mismatches above can land in those two to three years. So “funding moves upstream” in this post describes only the direction of funding; upstream revenue is not determined by funding direction alone.

US public observation infrastructure: from building to buying

Satellites

Since 1975 NOAA has deployed a generation of geostationary satellites roughly every twenty years. The current GOES-R series is expected to retire in the early 2030s; its successor GeoXO was originally planned for a first launch in 2032 with service to 2055. The original plan was six satellites and five instruments, with a life-cycle cost estimate of $19.6B for 2020–2055. NASA, as the procurement agent, completed all contract awards between March 2023 and September 2024. In 2025 the White House determined the plan’s cost was unsustainable and directed NESDIS to cancel all major instrument and spacecraft contracts and recompete them as fixed-price contracts, reducing the number of satellites to four and keeping only the infrared and visible imager and the hyperspectral infrared sounder (CRS R49026). As of 24 July 2026, NOAA had terminated the atmospheric composition, lightning mapper, and ocean color instrument contracts (CRS IF12898); the contract values CRS lists for atmospheric composition and ocean color are $262M and $371M respectively. Under the new plan no satellite carries a lightning mapper. The FY2026 appropriations passed by Congress in January 2026 require all GeoXO satellites to retain the imaging and sounding instruments and keep the 2032 first launch, but do not restore the three cancelled instruments; the Department of Commerce approved the $11.9B four-satellite plan in April 2026. Senate bill S. 3923, introduced in February 2026, requires imaging and sounding as core instruments on all satellites. It passed the Senate Commerce Committee in March and had not become law as of this writing (S. Rept. 119-154).

Radar

A recent example: from 2 to 5 October 2026, data from five NEXRAD sites in the southeastern US stopped being distributed to the public because of a transmission line failure. The radars themselves were operating normally, and NWS forecasters could still see the data. Jim Cantore used the occasion to call for privatizing the radar network, and the quotes and replies contain considerable argument over whether nationwide weather radar observation should be handed to private companies. This failure was in the data distribution link and unrelated to radar age, but the discussion itself shows that privatization is already a public question.

NEXRAD technology was developed in the 1980s and built out in the 1990s, entering operations in 1992, with 159 WSR-88D units in total (122 NWS, 37 FAA and Department of Defense) and a design life of 20 years. A $150M, nine-year Service Life Extension Program ran from 2015 to 2024, replacing signal processors, transmitters, pedestals, shelters, and emergency generators. The last unit was completed in August 2024, and NWS describes it as extending operations beyond 2035 (NWS announcement). A July 2026 CRS report lists the concerns as coverage gaps, signal interference, and the end of design life in the early 2030s; the NWS Radar Next request for information lists aging infrastructure, discontinued parts, and supply chain disruption as the three factors affecting reliability. The replacement program Radar Next is at the option-comparison stage, with candidates including another round of life extension, an upgraded WSR-88D, and phased array. The timeline CRS gives is planning complete by 2028, design complete by 2032, operations and maintenance from 2035, and installation complete by 2040.

Until Radar Next lands there is a window on the order of a decade. The demand for private gap-filling radar and this window are two different things: X-band gap-filling radars fill the low-level blind spots that NEXRAD’s station spacing was never designed to cover, and that demand does not disappear when Radar Next lands; the window affects procurement of the replacement itself, including phased array suppliers and data processing software. Both have already been accepted as procurement targets by NWS operational systems. For comparison with China: the 112 radars of the CINRAD backbone were mainly built between 2004 and 2012, an average age more than ten years younger than NEXRAD. The two countries’ radar networks are at different stages of their life cycles.

The only official reason for the GeoXO cuts is “cost unsustainable”; the CRS reports do not mention interest rates or debt. What can be traced is a separate line: NOAA’s commercial data program grew from $8M in FY2020 to $27.5M in FY2024 (CRS IF12671), awarded $67.05M in radio occultation data contracts in a single round in September 2026, and began purchasing Tomorrow.io’s microwave sounder data in June 2026. The three cut GeoXO instruments have no corresponding commercial data purchase; lightning data is only at the “evaluation” stage. My inference: the shrinking budget for self-built systems and the growing budget for data purchases are happening at the same time, and private observation companies gain anchor customers as a result. No official document supports a causal link between the two. The effect on forecast quality itself is unverified.

Japan

Beyond the US, orders and contracts in Japan’s remote sensing satellite sector are also growing fast, but mostly in the sensitive area of synthetic aperture radar satellites, tied to defense, military industry, and intelligence, where my knowledge is limited. For scale: in March 2024 the Ministry of Defense signed a 5.649 billion yen contract with iQPS for a technology demonstration satellite, and the Cabinet Office’s SAR demonstration project the same year was 1.538 billion yen; the Ministry of Defense’s satellite constellation development and operation project in February 2026 totaled 283.1 billion yen (tax included), of which Synspective’s SAR imagery acquisition subcontract was 105.6 billion yen. On the civil weather side there is no free, open, high-resolution operational model at the level of HRRR; JMA’s MSM and LFM are distributed for a fee through the Japan Meteorological Business Support Center, and providing forecasts to third parties is subject to licensing under the Meteorological Service Act. This means the open-data pressure on Japan’s middle layer is weaker than in the West, and this post’s assessment of the middle layer holds to a lesser degree in Japan.

Europe

I have no direct contact with European companies; what follows comes only from funding announcements and agency annual reports. Europe gets its own section because it went through the same rate rises and public data opening as the US, with a different outcome: commercial upstream funding comes from defense demand, the public upstream is sustained by multi-year civil contributions, and there is almost no layer in between where meteorological agencies buy commercial data.

Commercial funding

Funding rounds for European weather and Earth-observation companies in 2024–2026, sorted by amount:

The amounts fall into three tiers. The hundreds-of-millions tier has one company, ICEYE: FY2025 revenue above €250M, EBITDA above €100M, contracted backlog above €1.5B (company announcement); seven European governments have bought its satellite systems, the German Ministry of Defence awarded it and Rheinmetall a €1.7B SAR constellation contract in December 2025, and the two formed the joint venture Rheinmetall ICEYE Space Solutions in June 2026. Its one-year revenue exceeds the combined funding of every other company in the table, and the demand has the same source as the SAR orders in the Japan section: defense budgets. The tens-of-millions tier is the thermal-infrared satellite companies and Meteomatics. constellr’s Series A announcement describes the business as “defence-grade thermal intelligence”; SatVu’s 2026 round included the NATO Innovation Fund; OroraTech’s paying customers are the Greek government (four satellites procured through ESA, €20M) and the Canadian Space Agency (the WildFireSat contract, Can$72M, awarded to Spire Canada with OroraTech supplying the payloads). OroraTech and constellr are both initial partners of the Rheinmetall ICEYE joint venture above; in September 2026 OroraTech also signed with Eutelsat to host 48 thermal-infrared sensors on its future LEO satellites, with an option for 48 more, targeting public authorities and security services. Meteomatics is the only company in this tier with weather as its main business. It was founded as a software company and later built its own Meteodrone sounding drones, the same structure as WindBorne. Its public customers are still at the trial stage: DWD began testing Meteodrones at the Lindenberg observatory in June 2025, and it holds a CRADA with NOAA. Below ten million are the pure-software companies, at the same scale as Brightband and Silurian in the previous section.

The investors differ from the US. Every ICEYE round includes Solidium, the Finnish state holding company; OroraTech and constellr have Bayern Kapital from the state of Bavaria; constellr and Skyfora have the EU’s EIC Fund; SatVu has the British Business Bank. Pure venture rounds led by General Catalyst and General Atlantic appear only in ICEYE’s 2025–2026 rounds. Government money enters these companies as equity and whole-satellite procurement. I found only one data-purchase contract: EUMETSAT signed an operational radio occultation data contract with Spire in December 2023, renewed it for two years at €3M in October 2025, and extended it for one more year at about €4M in August 2026. EUMETSAT Council policy is that commercial data procurement “should remain marginal compared to EUMETSAT own-programmes” and that the technology should be European (briefing to WMO, June 2024); NOAA’s commercial data budget for the same fiscal year was $27.5M. EUMETSAT’s 2024 annual report also notes that the radio occultation products distributed to member states include US commercial data obtained through NOAA’s purchases, and that it discussed with the China Meteorological Administration and the China National Space Administration the possibility of China sharing its commercial data globally. The report does not name the providers, but the passage sits in the section on commercial radio occultation data, and the preceding sentence describes the discussion as one on radio occultation data procurement. The commercial data may be the Tianmu and Yunyao data from the China section. The discussions in the report took place over the course of 2024, and the two constellations’ data only entered CMA operations at the end of December that year, so I have not been able to confirm that the timelines match.

Acquisitions are small as well. Before buying Atmo, Vaisala made four acquisitions in a row: Speedwell Climate in September 2024, Nevis Technology in October, Maxar’s WeatherDesk business in December ($70M, 2023 sales of $13M), and Quanterra in September 2025. Windyty, the operator of Windy, took a majority stake in meteoblue in June 2024, and the agricultural weather station company Sencrop was acquired by ISAGRI in January 2025; neither amount was disclosed. I found no public record of a European weather company shutting down in 2024–2026.

Public observation

Both of Europe’s new satellite generations are at the start of deployment. Meteosat Third Generation (MTG) in geostationary orbit has six satellites, four imagers and two sounders, designed for 20 years of operation; the budget approved by EUMETSAT in February 2011 was about €2.37B at 2008 economic conditions. MTG-I1 launched in December 2022 and entered service on 4 December 2024; MTG-S1 launched on 1 July 2025 carrying an infrared sounder with 1,953 spectral channels and the Sentinel-4 atmospheric composition instrument, and the sounder’s pre-operational data was released to all users on 30 July 2026; MTG-I2 launched on 27 August 2026, was handed over to EUMETSAT on 9 September, and is expected to enter service in the second quarter of 2027. The two instruments cancelled on GeoXO, lightning imaging and atmospheric composition, have their MTG counterparts, the Lightning Imager and Sentinel-4, already in orbit. Metop Second Generation in polar orbit also has six satellites. Metop-SGA1 launched on 13 August 2025; its radio occultation sounder entered service on 5 May 2026 and its microwave sounder on 27 August, the latter with five water-vapour channels degraded after an anomaly in March (EUMETSAT). Metop-SGB1’s planned window on EUMETSAT’s launch timeline is June–August 2026; it has not launched as of this writing, and the remaining four are scheduled between 2032 and 2040.

The difference between the two sides is the budget mechanism. EUMETSAT has 30 member states, and mandatory programmes are funded on a GNI scale; the 2024 annual report shows expenditure of €763.2M, of which MTG €219.1M and EPS-SG €254.3M, and member contributions of €506.0M, with Germany at €100.7M, France €68.9M, and the UK €67.7M. The programme approved in 2011 has run on its original schedule to this day; the GeoXO instrument contracts were signed in 2023–2024 and cancelled after the 2025 change of administration. ESA’s November 2025 ministerial council subscribed €22.3B, of which Earth observation €3.4B, up 27% from 2022 (ESA briefing). The same council created the European Resilience from Space Earth-observation programme, with a €100M envelope for its first phase and €166.6M subscribed by member states, €55M of it from Poland; defense demand is starting to enter ESA’s civil Earth-observation budget, at an amount still small against the €3.4B.

There is no Europe-wide radar network; each country renews its own. The UK completed the dual-polarization upgrade of its 15 radars in January 2018 for about £10M; Sweden completed its 12 between 2014 and 2021; Spain’s roughly €80M renewal is expected to complete in August 2026. EUMETNET’s OPERA programme combines national radars into a 1 km, 5-minute European composite. These networks were built about twenty years after NEXRAD and are at a different point in their life cycle from the US network.

My assessment: funding in Europe also flows upstream, but the upstream splits into two layers with no transactions between them. The commercial layer is driven by defense demand; every round and order in the hundreds of millions comes from defense ministries and governments buying whole satellites, while observation hardware companies with weather as their main business all raised under $22M. The public layer is sustained by EUMETSAT’s multi-year civil contributions, builds to the plan set in 2011, and buys €3–4M of commercial data per year. The shift “from building to buying” in the US section has not happened in Europe, and agency data purchases have not become a revenue source for commercial companies. The middle layer is squeezed by ECMWF open data just as in the US; see the next section.

The middle layer: open data and open models

Another reason the software layer struggles to raise money is that the software side of weather and geospatial data processing has been significantly reshaped worldwide in recent years.

Traditionally NOAA in the US was the main institution releasing data free to the world over the long term; ECMWF in Europe has reached the same point in recent years. In December 2019 the ECMWF Council approved a phased transition from a restrictive data policy to fully open data, originally to be completed in 2026 and later brought forward to 1 October 2025. The open data portal went live in January 2022, with 0.4° data released under CC BY 4.0 and third-party redistribution permitted; in March 2024 the free subset was raised to 0.25°; from July 2024 all real-time data at 0.4° and coarser was open under CC BY 4.0. On 1 October 2025, the entire real-time catalogue moved to CC BY 4.0, the information fee was abolished, and a service fee is charged only for high-volume or customized distribution; the previous licence terms were no redistribution, no brokering, no resale. The free open subset was still 0.25° as of early October 2026. A note from the ECMWF Data Support team on 30 September 2026 expects 0.1° (about 9 km) to go live before the end of October, with a 2-hour delay as in the plan published in 2025. Since 1 October 2025 Open-Meteo has redistributed IFS HRES through its API at the full 9 km resolution with no added delay. On usage, ECMWF’s March 2025 announcement gave a 150% increase in downloads since the expansion of open data and about 680 TB per month distributed through the data portal; the ECMWF forum gives 1292 TB distributed through the portal in November 2025, excluding cloud mirrors.

The effect of this process on the middle layer: businesses that resell or package ECMWF data lose their pricing basis. ECMWF’s own fee structure is free data with charges for distribution and customization. As free open data keeps improving in quality, the margin for commercial data is compressed.

Other factors in the same direction:

1. Open-weight weather models (GraphCast, AIFS, Aurora, and others) let individual developers and small teams run medium-range forecasts on a single GPU, matching the capability of commercial weather model data from a few years ago. Among them, AIFS Single’s weights are CC BY 4.0, while the GraphCast and Aurora weights are restricted to non-commercial use; ECMWF’s AIFS output is also in the open catalogue;
2. The Overture Maps map data project launched by the Linux Foundation in December 2022 has, in my experience, data quality in Western countries already sufficient to support map services without tight freshness requirements; Microsoft replaced Bing Maps’ own buildings layer with Overture’s from July 2024;
3. The open PMTiles data format has brought access to map basemaps, traditionally expensive, down to the level where an individual can self-host.

Pure-software AI weather model companies could still raise money in 2024–2026; the difference is in scale. Jua raised a €10M Series A in June 2025, Atmo about $17.6M in December 2024, Silurian $6M in April 2026, and Brightband a $10M Series A in September 2024; Series B and later rounds on the hardware side in the same period were in the hundred-million-dollar range. Exits: Atmo was acquired by Vaisala in June 2026 for a fixed consideration of $70M plus up to $60M in earn-outs, with 2025 revenue of about $2M; ClimateAi shut down in August 2026 after raising $38M; Salient and Zeus AI have had only grants and SBIR awards in the last two years, with no new venture rounds. Brightband is a separate case: Google’s WeatherNext 3 documentation lists Brightband’s Operational WeatherBench as an independent evaluation; as of September 2026 no paying customers had been disclosed, and its operating situation is unclear.

In the years of easy funding, software service companies of this kind could still trade funding for growth; with funding tighter and free alternatives improving, this path gets harder.

A summary I made during my earlier work: “good weather forecasting has two prerequisites, first natural disasters, second money.” It now needs a third: “sufficiently open data.” Countries and regions that open high-quality data the way the US does will develop faster, and with a more mature investment environment have the chance to produce leading weather and geospatial technology companies. Tighter funding determines that the “money” part concentrates in hardware and users.

Downstream: value shifts to users

The economic value of weather information is concentrated in the organizations that use it to make decisions. This came from a book I read years ago about Koch Industries, Kochland, which mentions that its energy trading desk in Houston hired forecasters away from The Weather Channel and similar outfits in order to get weather forecasts better than public channels, and used them to judge demand and price natural gas; Koch itself bought the 9,271-mile United Gas Pipe Line in 1992 and held a substantial natural gas pipeline business. Energy businesses such as wind and solar generation that rely on weather forecasts to optimize operations and trading follow the same logic: the cost of buying weather forecasts and analysis data is far below the operating gains that result. These organizations do not depend on venture capital, and their motive for internalizing weather capability is independent of the funding environment: forecast costs fall, decision gains stay the same, building a team and buying raw data costs less than subscriptions, and the strategy itself stays in-house. WindBorne using its Series B funds to sell raw observation data to investment funds is one example of this direction. So my assessment is that in the future, weather capability is more likely to be internalized by this kind of user, and the market for independent weather vendors will shrink.

The other way out for a middle-layer company is to become a user itself, and China already has a precedent. Seniverse (心知科技) was founded in 2016 doing weather big data, and from 2019 used weather technology to enter the energy and power industry; in April 2021 the Shanxi electricity spot market began settlement trial operation (April to June was the first quarterly settlement, continuous operation from July), with prices formed every 15 minutes and a cap of 1.5 yuan/kWh. Seniverse focused on power trading services from 2021, and that business grew year by year into the core (36Kr report). Its current form is a full-process managed service for generators, retailers, virtual power plants, and storage customers, covering strategy, trade execution, and settlement analysis. The company describes tens of billions of kWh under annual management and offers a price guarantee service with shared returns and shared risk. A price guarantee means the company bears trading risk on its own balance sheet, and the business has changed in nature from software to trading. Externally it only talks about how many fen per kWh the return improves; the weather forecast has receded to the input layer.

Japan’s market is at a similar stage. The FIP scheme makes renewable generators bear imbalance risk, aggregators need power forecasts to reduce imbalance charges, and JEPX spot volatility has increased. Japan’s current stage is close to China’s 2021 turning point, and companies of the same kind are more likely to emerge among aggregators.