The freight train outside Gothenburg wasn't supposed to derail.
It was June 29, 2022—a Wednesday afternoon, 34 degrees Celsius, the kind of heat that makes Swedish summers feel borrowed from somewhere further south. The rails had expanded beyond their tolerances, warping into S-curves that no trackside inspector could have spotted in time. Metal does that when it gets hot enough. By the time the distortion became visible, it was catastrophic.
Similar failures play out every summer across aging infrastructure networks, though most never make headlines. Power substations overheat during demand surges. District heating pipes leak beneath city streets, wasting energy and destabilizing the pavement above. Airport runways soften in ways that pavement inspectors miss until cracks appear. The common thread? Heat stress that legacy monitoring systems simply can't see.
Now a handful of startups think they've found the answer—and it involves launching satellites.
Constellr, Hydrosat, OroraTech, and Satellite Vu are racing to blanket Earth with thermal imaging satellites capable of detecting temperature anomalies at resolutions fine enough to flag a single transformer, a hundred-meter rail segment, or a leaking pipeline junction. It's infrastructure triage from orbit. And the first operational data is already streaming down, revealing blind spots that ground-based monitoring has lived with for decades.
Sharper Than Before
Constellr got its first satellite up on January 14, 2025. The SkyBee spacecraft achieves 30-meter native thermal resolution, with products sharpened algorithmically to 10 meters. By mid-March, the German company—spun out of Fraunhofer EMI, the defense research institute—published its first thermal map of Tokyo. The image showed building-level temperature gradients across the metro area with a clarity that wasn't commercially available before.
A second satellite followed in June, doubling collection capacity.
The pitch is deliberate: "sovereign-grade thermal intelligence" for European governments wary of depending on foreign imagery to monitor critical infrastructure. It's a nationalism-tinged sales argument, but one that's landed contracts.
Hydrosat took a different path. Its VanZyl-1 satellite, launched in August 2024, captures longwave infrared at roughly 70-meter resolution paired with 30-meter visible and near-infrared bands. A second satellite—VanZyl-2—went up in June 2025, quadrupling capture capacity and enabling daily revisits over priority areas.
CEO Pieter Fossel frames the mission around agriculture: field-scale water stress monitoring to optimize irrigation, the kind of precision farming that venture capital loves to fund. But the same thermal data reveals cooling tower inefficiencies at power plants and thermal plumes downstream of dams. Turns out, heat is heat.
Satellite Vu pursued what may be the most audacious specification: 3.5-meter thermal imaging, day and night. Its HOTSAT-1 pathfinder launched in June 2023. Then, in December that year, the company disclosed an anomaly. Details were sparse, the kind of vague engineering setback that makes investors nervous.
Despite that, CEO Anthony Baker argues the use case justifies the technical risk. Identifying heat-inefficient buildings across entire cities, then verifying post-retrofit performance—it's a compelling narrative, especially for decarbonization-focused municipal governments. Investors including Lockheed Martin and In-Q-Tel have backed the pursuit of HOTSAT-2 and HOTSAT-3, though neither has launched yet.
OroraTech narrowed its focus to wildfires. Eight satellites launched in March 2025 aboard a Rocket Lab Electron, filling afternoon and overnight coverage gaps that leave forests vulnerable for hours at a time. In February, the Canadian Space Agency awarded the company and Spire a C$72 million contract to develop payloads for WildFireSat, a national wildfire monitoring system.
It's a rare case of a commercial constellation winning public procurement on speed and specialization—governments, it seems, don't always build faster.
Why This, Why Now
The commercial Earth observation market is projected to grow from approximately $5 billion in 2024 to more than $8 billion by 2033, according to Novaspace. Defense contracts and higher-end products drive much of that expansion. North America held 44 percent market share in 2023; Europe, 22 percent.
Thermal imaging sits at the premium end. Higher cost per pixel, certainly, but revealing phenomena invisible to optical or radar sensors.
Climate adaptation finance creates another tailwind, though the gap between need and reality remains staggering. The UN Environment Programme's 2024 Adaptation Gap Report estimates developing countries need between $187 billion and $359 billion annually for adaptation measures. International public flows reached only $28 billion in 2022. Infrastructure owners in wealthy nations face similar pressure—quantify climate risk, harden assets against temperature extremes—even as federal mandates wobble.
Consider the U.S. Securities and Exchange Commission's climate disclosure rule. Adopted in March 2024, stayed in April amid litigation, then withdrawn entirely from defense in March 2025. Legal limbo. California's SB 253 survived—requiring large companies to report Scope 1 and 2 emissions by August 10, 2026, with Scope 3 disclosures to follow in 2027. But SB 261, the state's climate risk reporting law, was enjoined by the Ninth Circuit in November 2025.
The EU's Corporate Sustainability Reporting Directive continues, albeit with a two-year delay for some reporting waves under an April 2025 "Stop-the-Clock" directive.
Regulatory uncertainty hasn't stopped procurement, though. Asset owners still face investor scrutiny, insurance premium escalations, and operational failures that thermal monitoring promises to anticipate. A district heating network in Gothenburg doesn't wait for SEC clarity before hunting for leaks that waste energy and erode margins.
Money finds problems.
What These Satellites Actually See
Constellr's technology demonstrates the shift from raw pixels to actionable anomaly detection. The company's Thermal Intelligence Atlas processes 30-meter land surface temperature data through machine learning pipelines trained to flag deviations. A rail embankment running 5 degrees Celsius hotter than adjacent sections during a June afternoon. A substation transformer bank glowing warmer than historical baselines under load.
Radiometric accuracy hovers around 1.0 to 1.5 Kelvin, with noise-equivalent delta temperature between 0.05 and 0.15 Kelvin. That precision matters when distinguishing normal diurnal warming from stress-induced heat buildup—the difference between "everything's fine" and "something's about to fail."
Hydrosat pairs its thermal imagery with visible bands and external data: weather models, IoT sensor feeds, operational logs. The fusion approach infers water stress in crops or cooling system inefficiencies at industrial facilities. The company markets an IrriWatch analytics layer for agriculture, but the same logic applies to refineries, where abnormal process heat or idled units show up as thermal anomalies correlated with maintenance schedules or market disruptions.
OroraTech's wildfire satellites detect sub-pixel hotspots in near-real time, triggering alerts when fires ignite during afternoon heat or overnight—the hours when geostationary weather satellites' coarse resolution misses the earliest stages. A fire detected three hours earlier translates to smaller burn scars and lower suppression costs, particularly in remote forests where ground patrols are sparse.
Satellite Vu's proposed 3.5-meter resolution—if HOTSAT-2 and HOTSAT-3 succeed—would enable building-level thermal audits at city scale. UK local authorities piloted airborne thermal mapping programs with providers like Bluesky, identifying properties with severe heat loss for fuel poverty interventions and retrofit targeting. Satellite Vu aims to make such surveys routine and scalable, covering entire metro regions on repeating schedules rather than one-off airborne campaigns that cost tens of thousands per city.
If it works.
The Infrastructure Blind Spots
Rail operators have used handheld thermal cameras for years to spot overheated wheel bearings and electrical faults. Satellite thermal extends that logic to corridor-scale triage.
Studies demonstrate that infrared imaging can identify rail sections prone to buckling under heat stress: embankments with poor ballast drainage, curves with constrained expansion joints, sun-exposed stretches that reach critical temperatures hours before shaded segments. A thermal satellite flags macro zones; track inspectors follow up with ground surveys and targeted interventions. It's not perfect, but it's faster than waiting for something to buckle.
Pavement thermography follows similar principles. The U.S. Federal Highway Administration notes that infrared surveys detect subsurface defects—delamination, voids, moisture intrusion—that appear as temperature differentials on the surface. Airports use thermal imaging during paving operations to ensure uniform compaction and catch cooling anomalies that predict premature failure.
Satellite land surface temperature trends, aggregated over time, can prioritize runway segments under persistent heat extremes for preventive maintenance. Maybe not glamorous work, but essential.
Power grid operators face compounding stress. Transformer hotspots detected via infrared indicate bushings or connections approaching failure. During heatwaves, when air conditioning loads spike, cooling systems at substations may struggle—detectable as elevated surface temperatures around equipment enclosures.
Multiple studies validate automated hotspot detection using convolutional neural networks trained on infrared substation imagery, achieving sensitivity around 65 percent and specificity above 97 percent for anomaly flagging. Translation: the models catch roughly two-thirds of genuine problems while rarely crying wolf.
District heating networks—common in Scandinavia and Central Europe—lose efficiency through underground leaks that waste energy and destabilize urban streets. A research team demonstrated a machine learning method (SHEDAD) that identifies anomalous substations in heating networks with roughly 65 percent sensitivity and 97 percent specificity, using operational data.
High-resolution satellite thermal offers a complementary layer. Thermal line anomalies along buried pipes can triage inspection crews to suspect corridors before they excavate. Less digging, lower costs.
Dams and levees present geotechnical monitoring challenges at scale. Infrared thermography detects seepage and erosion zones: cool patches on dam faces where groundwater infiltrates, warm zones on levee toes where internal erosion channels heat. Lab and field studies validate the approach for large civil structures, though satellite resolution historically fell short.
Now, 30-meter to 70-meter thermal satellites can screen hundreds of kilometers of levees, prioritizing sections for UAV or ground follow-up.
Industrial sites reveal operational patterns through heat. Refineries show abnormal process temperatures when units shut for maintenance or experience upsets. Gas flaring intensity—a proxy for production levels or waste—appears as bright thermal signatures. Kayrros, a geospatial analytics firm, demonstrated predictive refinery observation by fusing thermal, radar, and operational data, allowing market analysts to infer throughput and maintenance schedules independent of company disclosures.
Transparency by satellite. Refineries hate it; analysts love it.
The Dual-Use Dilemma
Thermal satellites carry a defense pedigree, and nobody in this industry pretends otherwise.
Night-time operations. Activity monitoring in denied environments. Infrastructure surveillance. All of it attracts military and intelligence customers willing to pay premium rates for tasked collections. Constellr explicitly markets "sovereign-grade" capabilities, emphasizing reliability and data hosting within European jurisdictions for NATO allies. OroraTech's rapid-tasking wildfire alerts double as a demonstration of responsive satellite operations—a capability with obvious defense applications.
The U.S. streamlined its NOAA Commercial Remote Sensing Regulatory Affairs licensing in 2020, lifting many restrictive conditions by 2023 to maintain competitiveness against foreign operators. Export controls still apply to certain thermal cameras and focal plane arrays under the Export Administration Regulations, but operational satellites face fewer constraints than a decade ago.
Hydrosat holds a NOAA remote sensing license. European operators navigate ESA and national frameworks that balance openness with sovereignty concerns.
This dual-use dynamic subsidizes civil applications, frankly. Defense budgets fund constellation buildout and sensor R&D that spill over to infrastructure monitoring. It also shapes market access—countries wary of foreign surveillance may favor domestic or allied providers for critical infrastructure imagery, fragmenting the market by jurisdiction.
What Comes Next
Public missions will materially expand medium-resolution thermal supply over the next three years.
TRISHNA, a joint ISRO-CNES mission, targets a 2026 launch with approximately 57-to-60-meter thermal resolution and three-day revisit. The European Space Agency's Copernicus Land Surface Temperature Monitoring (LSTM) mission, scheduled for 2028, will field two satellites achieving 50-meter resolution and two-to-four-day revisits. NASA's Landsat Next, also targeting the late 2020s, will add five thermal infrared bands at 60-meter resolution with six-day global coverage.
These public missions establish a baseline: free, global, medium-resolution thermal data for science and policy. Commercial constellations compete on revisit frequency, tasking flexibility, night imaging, and analytics. Constellr and Hydrosat can task their satellites for priority areas within hours, delivering same-day thermal intelligence that public missions with fixed orbits cannot match.
Satellite Vu's proposed 3.5-meter resolution, if realized, would occupy a niche public missions don't address. But that's a big "if."
The analytics layer may matter more than raw pixels in the long run. Infrastructure operators don't need global thermal coverage; they need anomaly alerts tied to asset registries, correlated with maintenance logs, and scored by failure probability. Constellr's Thermal Intelligence Atlas, Hydrosat's IrriWatch, and similar platforms aim to become decision-support layers—ingesting thermal data, fusing it with other sensors and external data, then pushing alerts to facility managers and grid operators.
Cross-sensor fusion is already standard practice. Thermal satellites pair with synthetic aperture radar for all-weather monitoring and optical imagery for context. Ground-based IoT sensors provide real-time validation. The winning models will likely combine high-quality land surface temperature with tight radiometric accuracy, domain-specific training for rail, pavement, district heating, or substations, and workflows that integrate into asset management systems and ESG reporting platforms.
Standards and disclosure frameworks keep evolving, though inconsistently. California's SB 253 creates a compliance wedge—companies reporting emissions must quantify operational risks, and thermal monitoring offers comparable, facility-scale heat metrics. The EU's sustainability taxonomy and CSRD, despite delays, maintain pressure on infrastructure owners to demonstrate climate adaptation and resilience planning.
Even in jurisdictions where mandates stall, insurance underwriters and bondholders increasingly demand quantified climate risk assessments. Money, again, finds problems.
Supply-side expansion from 2026 to 2028—TRISHNA, LSTM, Landsat Next—means more frequent, freely available thermal data at 50-to-60-meter resolution. Commercial operators must differentiate on speed, resolution, or analytics to justify premium pricing. Some may pivot toward defense and intelligence customers willing to pay for exclusive tasking and rapid turnaround. Others will chase volume in agriculture, forestry, and urban heat mapping, where business models depend on subscriptions and data-as-a-service.
The infrastructure monitoring use case remains underserved, perhaps surprisingly so. Rail operators, grid utilities, and district heating networks lack continuous, asset-level thermal surveillance. Ground sensors cover critical points but not corridors or entire networks. Aerial surveys cost too much to run frequently.
Satellites offer the only scalable path to routine thermal anomaly detection across thousands of kilometers of rail, pipeline, and transmission line—assets that will buckle, leak, or fail in ways invisible until they don't.
Like that freight train outside Gothenburg.
