Inside the $35 Billion race to interpret what infrastructure inspections actually reveal

Michael Petrosjan spent part of his early career commanding soldiers who launched reconnaissance aircraft into terrain nobody wanted to walk into on foot. The aircraft existed to answer one narrow question at speed, which was what sat out there and what condition it was in. His service as an officer in the German airborne forces, where he commanded a unit operating Aladin reconnaissance drones, gave him an early view of what unmanned systems could do outside defence. What he found on moving into business was that the industries with the sharpest version of that same question, the utilities and network operators responsible for the physical systems entire economies run on, were still answering it by sending people up ladders.

Gathering asset information the traditional way is expensive, slow and frequently dangerous, and the European Institute of Innovation and Technology, which later accelerated the company, described the pattern precisely when it noted that workers scaling wind turbines often return with data of poor quality that teams then struggle to share. In September 2015, Petrosjan joined Andreas Dunsch, Christian Caballero and Holger Dirksen to build software that would remove the trade-off. The company opened in Hamburg, moved its headquarters to Leipzig after joining the SpinLab accelerator, and released Map2Fly, a flight-planning map that has since handled more than 12 million location queries from over 100,000 users.

Eleven years on, the description Petrosjan gives of the business has moved a long way from aircraft. FlyNex now calls itself an infrastructure intelligence company, and the problem it describes is one of interpretation rather than capture.

“Critical infrastructure operators generate enormous amounts of visual data through drones, cameras and inspections, yet most maintenance and investment decisions are still made with fragmented information,” Petrosjan, Chief Financial Officer and Co-Founder of FlyNex, said.

The inspection problem has grown faster than the workforce sent to solve it

The market context behind that observation has shifted sharply in the company’s favour. The International Energy Agency puts current global spending on electricity grids at around $400 billion a year against roughly $1 trillion on generation, and calculates that annual grid investment must climb by approximately 50% by 2030. The agency also counts more than 2,500 GW of renewable, large-load and storage projects sitting in connection queues worldwide, and estimates that meeting stated national targets requires adding or refurbishing over 80 million kilometres of grid by 2040, a figure equivalent to the entire existing global network.

Every kilometre of that network becomes an asset somebody has to inspect, document and defend to a regulator. The ageing profile compounds it. Eurelectric has warned that many of Europe’s distribution grids will be more than 40 years old by 2030, approaching the end of their design lives, while demand growth and climate exposure push them harder. In the United States, the American Society of Civil Engineers graded national infrastructure at C- in its 2025 report card, with more than 56,000 bridges classified as structurally deficient. The volume of assets requiring condition assessment is rising at a rate no realistic hiring plan absorbs.

Trust, rather than model accuracy, decided how quickly utilities adopted

The technical work was never the constraint. Petrosjan is direct about where the difficulty actually sat, and his answer is one that most enterprise AI vendors selling into regulated industries will recognise.

“The biggest challenge was never building the AI models. Convincing traditionally conservative infrastructure industries to trust AI-driven decision-making was the harder problem,” he said. Operators, he added, “do not simply need AI models. They need reliable workflows, regulatory compliance, scalability, and measurable business outcomes.”

The approach FlyNex took was to sell the outcome rather than the algorithm. “We overcame this by working closely with utilities and infrastructure operators and focusing on measurable business outcomes rather than technology alone,” Petrosjan said. Company figures compiled by Dealroom put inspection cost reductions at up to 60% and time savings at up to 75%, claims that matter to a utility only insofar as they survive contact with a rate case or an internal capital committee.

Regulation has moved in parallel. Under the European Union Aviation Safety Agency rules, beyond visual line of sight operations sit in the Specific category, where infrastructure inspection is among the primary commercial use cases, and operators can obtain a Light UAS Operator Certificate to self-authorise repeat missions inside an approved risk envelope. Authorisation timelines still run to three to six months across much of continental Europe and six to 12 months in the United Kingdom, which keeps compliance tooling inside the product rather than adjacent to it. FlyNex demonstrated the operational end of this early, completing a beyond visual line of sight inspection flight over roughly 40 kilometres of overhead lines in Saxony.

Owning the full workflow separates data vendors from decision systems

The competitive claim FlyNex makes rests on scope. “While many solutions focus on collecting data or analysing data, FlyNex connects the entire workflow from data acquisition and asset intelligence to maintenance, risk, and capital investment decisions,” Petrosjan said.

That distinction carries weight in a category where most vendors occupy one segment of the chain. Intel Market Research values the drone inspection software market at roughly $1.97 Billion in 2025, rising to $3.63 Billion by 2034, and counts DroneDeploy, Flyability, Pix4D, vHive and Skydio among the principal players. The same analysis records consolidation running through the sector, with Trimble’s acquisition of Pix4D as one example of photogrammetry specialists being absorbed into larger engineering stacks. Independence on features alone is getting harder to sustain.

Commercially, the model is conventional enterprise software. “FlyNex operates a SaaS-based business model, generating recurring software revenues through subscriptions, platform licences, and enterprise agreements,” Petrosjan said, with further revenue from implementation, integrations and analytics modules. The customer list skews toward the asset-heavy operators the thesis depends on, and includes the transmission operator 50Hertz, the distribution utilities Netze BW, MITNETZ Strom, SachsenEnergie and enviaM, alongside Deutsche Bahn, ÖBB-Infrastruktur, Deutsche Telekom, Vodafone, Linde, TÜV Rheinland and the construction groups Goldbeck and Implenia.

“Utility companies use FlyNex to manage and analyse powerline and substation inspections, while infrastructure operators use the platform to document assets, detect defects, and monitor asset conditions at scale,” Petrosjan said.

Capital is following the asset base rather than the aircraft

Investor behaviour has tracked the same logic. A multi-million euro round in 2021 brought in STIHL Digital as a strategic investor alongside High-Tech Gründerfonds, TGFS, GPS Ventures and Snowflake Ventures, and took the team to 40 people. A further round completed in July 2025 saw TGFS follow on while SBG invested for the first time through Saxony’s Innovation Capital programme, which is co-financed by the German government’s Future Fund. PitchBook records around $7 million raised across 12 investors, a modest sum against the venture money flowing into AI generally, and characteristic of European deep-tech companies selling into procurement cycles measured in quarters.

The wider category is expanding faster than the software layer within it. Fortune Business Insights values the global drone inspection and maintenance market at $8.43 billion in 2025, forecasting $35.35 billion by 2034 at a compound annual growth rate of 16.9%, with North America holding 34.52% of the market in 2025 and the software segment growing at 18% a year. Marketintelo puts Asia Pacific ahead in autonomous inspection software on the strength of build-out across China, India and Southeast Asia, which places the growth centre of gravity some distance from Leipzig and raises the question of how a European company of this size reaches it.

Deloitte’s 2026 power and utilities outlook makes the demand case plainly, noting that operators are under pressure to deliver more reliability from the same resource base, which requires analytics and automation to drive capital and operational efficiency, with edge AI running from drones through to substation sensors. The commercial prize sits in capital allocation. An operator that can rank 400 substations by condition with evidence behind the ranking spends its capital budget differently from one working through a maintenance schedule by calendar.

“Our mission is to make critical infrastructure safer, more efficient, and more resilient by turning visual asset data into actionable intelligence,” Petrosjan said.