The longest walk in bomb disposal is the one you take alone, toward something that might kill you. Minerva Humanoids just raised $10 million to make sure the next person who takes that walk is a robot named Roger.
What Actually Happened
On October 6, 2026, Minerva Humanoids emerged from stealth with a $10 million pre-seed round led by General Catalyst, with Long Journey Ventures and Credo Ventures as co-leads. Additional participants include MVP Ventures, Kaya, RunwayVC, Compound, Founders Embassy, and a collection of institutional and angel investors including Hugging Face and Codecademy. The company has been operating in stealth, and the funding announcement coincides with the launch of its first paid pilot programs in fall 2026, according to the GlobeNewswire announcement. General Catalyst's involvement signals institutional conviction in the hazardous-operations segment before the company has public revenue, which is a strong endorsement given how selective the firm has become in deep-tech hardware bets.
Minerva was founded by CEO Sandor Felber and CTO Maurice Rahme to build humanoid robots for hazardous energy operations and public safety applications. Its first robot, Roger, is a semi-autonomous humanoid designed to operate in two primary environments: offshore oil and gas facilities, where explosive atmospheres, fall risks, and chemical exposure represent severe ongoing hazard, and explosive ordnance disposal contexts, where a single deployment failure is fatal. Dominic Asquith, a British Special Forces veteran now serving as Minerva's Head of Hazardous Operations, provided the operational framing in the announcement: "Every EOD operator knows the long walk," referring to the ritual approach to a suspected device that defines the profession's most dangerous moment. Roger is designed to take that walk instead, allowing specialists to apply their expertise remotely while keeping their physical bodies away from the hazard. The company also notes that NATO consistently cites EOD and hazardous materials response teams as the highest-casualty-risk segment in military and civilian emergency operations, a claim corroborated by independent coverage from The Manila Times.
The company's website positions Roger as a "specialist telepresence platform" rather than a fully autonomous system. Specialists retain control and apply their judgment remotely while the robot executes physical tasks in the hazardous environment. This semi-autonomous framing is a deliberate product architecture choice that distinguishes Minerva from the factory automation plays that have dominated the humanoid robot funding narrative. Figure AI, Agility Robotics, and Unitree are building toward autonomous repetitive-task robots. Minerva is building a platform for high-stakes, low-frequency, expert-judgment scenarios where the human must stay in the loop because the cost of an autonomous error is unacceptable by any engineering or regulatory standard currently in existence.
Why This Matters More Than People Think
Most of the humanoid robot investment narrative of 2025 and 2026 has been captured by the factory automation angle: robots doing repetitive assembly work at BMW, Toyota, and Foxconn facilities. That is a real market, but it is also a market with established competition, long procurement cycles, manufacturing unions, and labor-relations politics that slow adoption. Hazardous operations represent a categorically different entry point. The buyer is motivated by liability reduction and insurance cost savings rather than labor arbitrage. Procurement timelines are shorter because the alternative is leaving a dangerous job permanently understaffed or unfilled. There is no political opposition to deploying a robot in place of a human operator who faces daily risk of death. Minerva is entering from the side of the market where willingness to pay is highest, regulatory opposition is lowest, and the value proposition is unambiguous to every decision-maker in the purchase process.
The global EOD and hazardous materials response market represents a large, underserved market that has historically been served only by wheeled platforms with severe functional limitations. NATO's own assessments identify EOD teams as the highest-casualty-rate segment of military and civilian emergency operations across its member states. The US military has spent billions on remote-operated ground vehicles since 2003, but existing solutions are wheeled platforms with limited dexterity, unable to climb stairs, open doors, manipulate complex objects, or perform the fine-motor tasks that many hazardous operations still require human presence to execute. The humanoid form factor in this context is not a design preference, it is a functional requirement, because hazardous environments are built for humans to navigate, and the robots that operate in them must match that geometry.
The oil and gas opportunity is similarly underframed in coverage of the robotics sector. Offshore platforms operate in explosive atmospheres where a single static discharge can trigger a catastrophe with consequences measured in billions of dollars and multiple fatalities. Current practice involves human workers trained in specialized procedures performing inspections, valve operations, and equipment maintenance that could theoretically be handled remotely with sufficient robot dexterity. The insurance and liability costs associated with offshore fatalities run into the tens of millions of dollars per incident, a figure that makes a $10 million robot pilot program appear trivially cheap against a single avoided casualty. Minerva's paid pilots starting in fall 2026 will stress-test whether the unit economics actually close at the field level, which is the most important unanswered question in the business model.
The Competitive Landscape
The hazardous operations humanoid space is genuinely nascent, which creates both an opportunity and a warning. The incumbents in bomb disposal robotics are wheeled platforms: iRobot's defense division, Endeavor Robotics, and Northrop Grumman's Remotec subsidiary. These companies have decades of field experience, established procurement relationships with government agencies and defense ministries, and deep MIL-SPEC certification expertise accumulated over dozens of conflict deployments. Minerva's humanoid platform must eventually pass qualification gauntlets that these incumbents have already navigated, and that process takes years rather than months. A company with $10 million in pre-seed capital is starting a race where the finish line requires regulatory approval infrastructure that typically costs $50 million or more to build.
However, the critics who argue that Minerva faces an insurmountable incumbency advantage miss what is actually different about the current moment. The wheeled platform incumbents built their products for an era when the only viable control paradigm was direct teleoperation with joysticks and camera feeds. The current generation of foundation models and embodied AI has created a new control paradigm: a specialist operator can now describe intended actions in natural language, and the robot can translate those intentions into physical movements with a level of contextual adaptation that previous systems could not provide. This is not an incremental improvement on existing EOD robot designs, it is a different product category that the incumbents are not positioned to build because their architecture is designed around joystick teleoperation, not embodied AI control.
The broader analog that illuminates Minerva's strategic position is Palantir's early positioning in government data analysis: a company that identified a buyer with extreme willingness to pay, high mission criticality, and tolerance for nonstandard procurement processes, and built its early revenue on government contracts that civilian competitors either could not or would not pursue. If Minerva executes correctly, its first defense and energy contracts could generate revenue disproportionate to its current team size, providing the runway to build toward broader civilian safety applications. The risk is that the certification and procurement timelines extend beyond what a $10 million pre-seed can sustain without a bridge round, which is the central near-term financial risk for the company.
Hidden Insight: Semi-Autonomy Is the Smarter Architecture
The most important design choice in Minerva's product is the word "semi-autonomous," and it is almost entirely absent from coverage of the launch. The dominant narrative in humanoid robotics frames autonomy as the destination: robots eventually operate without human supervision, and every current deployment is a stepping stone toward that endpoint. Minerva is explicitly building in the opposite direction. Roger is a remote-operation platform that amplifies human expertise rather than replacing human judgment. That is not a limitation imposed by current AI capability, it is a deliberate product architecture that reflects a sophisticated understanding of the regulatory and operational reality of the markets they are entering.
In high-consequence, low-frequency scenarios, a bomb that might not be a bomb, a gas leak in an explosive atmosphere with ambiguous pressure readings, the value of human involvement is not in the physical presence. It is in the expert's ability to read ambiguous signals, apply probabilistic risk assessment, and take contextual actions that no training dataset can fully anticipate. The current generation of AI systems, including the frontier models from OpenAI, Anthropic, and now Reflection AI, is not reliable enough for fully autonomous operation in adversarial or genuinely novel environments where a single error is catastrophic and irreversible. Minerva's semi-autonomous model sidesteps this problem entirely: the human stays in the decision loop, the robot is the actuator, and the entire system can be deployed today in environments where fully autonomous certification is not available and will not be available for years.
The revenue model this unlocks is underappreciated by most observers. Robot-as-a-service for hazardous operations, where the robot is operated by a trained specialist at a remote console, creates a recurring revenue stream tied to operational hours rather than unit sales. A single offshore platform might require continuous robot monitoring during active drilling operations, generating thousands of operating hours per year. At a premium pricing tier appropriate for safety-critical applications, the economics per robot could be far better than the $25 per operating hour that Figure AI charges BMW for general manufacturing automation. The customer's willingness to pay is driven by insurance savings and liability reduction rather than labor cost, which is a structurally higher ceiling for pricing.
There is also a data acquisition dimension that is not discussed in the announcement but is inherent in the business model. Every hour Roger operates in a real hazardous environment generates sensor data, telemetry, and human operator decision records that no synthetic training dataset can replicate. That operational data becomes a proprietary training asset for the autonomous systems that Minerva will eventually build as AI reliability improves. Companies that control the data pipeline from real-world hazardous operations will have a durable advantage in certifying higher-autonomy systems, because regulators will require real-world performance evidence that only field-deployed systems can provide. Minerva is building that evidence base from the first day of operations, even if that framing does not appear in the press release. The semi-autonomous phase is not just a compromise with current AI limits, it is a data strategy for future autonomy.
What to Watch Next
The 30-day signal is the first named customer announcement. Minerva has committed to paid pilots in fall 2026, which means an announcement of a named early customer is likely before year end. The identity of that first customer will define the company's strategic trajectory: a US military or NATO defense agency as the first customer signals a government-first strategy with long but transformative contract timelines; a named offshore energy operator signals a commercial-first strategy with faster initial revenue but narrower initial scale. Watch for joint press releases from energy majors including Shell, BP, or ExxonMobil, which have been the most systematic adopters of new robotic technology in the upstream energy sector over the past three years.
The 90-day signal is General Catalyst's follow-on behavior. At $10 million, Minerva has approximately 18-24 months of runway at typical pre-seed burn rates for a small team. The critical question is whether the paid pilots generate a revenue signal compelling enough to support a Series A conversation by Q2 2027, or whether the regulatory and certification timeline requires a bridge round before that milestone is achievable. General Catalyst's portfolio includes multiple deep-tech companies with extended development cycles, so they understand hardware timelines. But a stealthy robotics company with no public revenue needs a clear customer validation story to raise at favorable terms in a market that has become more selective about capital-intensive hardware bets since the 2024 robotics funding peak.
At 180 days, the substantive question is whether Minerva can demonstrate a cost-per-incident reduction for its first operator customers that passes procurement committees. The unit economics need to show that a robot deployment pays for itself within a defined time horizon compared to the cost of human exposure, in insurance premium reductions, avoided workers' compensation claims, and regulatory compliance savings. If Minerva can produce a validated case study by Q2 2027 showing a clear return on investment for a named customer, it has a repeatable sales motion that the next round of funding can scale. If the economics remain ambiguous after two quarters of paid pilots, the company faces a harder conversation about whether its current product architecture is addressing the right initial application at the right price point.
The bomb disposal robot that gets certified first will define every one that follows.
Key Takeaways
- $10M pre-seed led by General Catalyst, Minerva Humanoids exits stealth targeting two sectors with extreme willingness to pay: EOD and offshore oil and gas hazardous operations
- Semi-autonomous design with specialist in the loop, Roger is a remote-operation platform that keeps expert judgment in the decision chain, enabling deployment in environments where fully autonomous certification is unavailable
- First paid pilots in fall 2026, Minerva moves immediately from stealth to commercial deployment, compressing the research-to-revenue timeline relative to most humanoid robot competitors
- Humanoid form factor as a functional requirement, EOD and offshore environments built for human operators require human-shaped robots that wheeled platforms with limited dexterity fundamentally cannot replace
- Data flywheel from real hazardous deployments, operational sensor data and human decision records from EOD and offshore work become a proprietary training asset for future higher-autonomy certification
Questions Worth Asking
- Semi-autonomy sidesteps today's AI reliability limits and enables immediate deployment, but at what point does human-in-the-loop operation become a competitive liability rather than a regulatory advantage, and how does Minerva plan to navigate that architectural transition as AI reliability improves?
- The defense and energy procurement process can run two to five years from first pilot to full contract, does a $10 million pre-seed provide enough runway to survive that timeline, and what does a bridge-round scenario look like if major contract announcements slip past the 18-month mark?
- If Minerva succeeds in demonstrating ROI in EOD and offshore energy, what prevents Figure AI, Sanctuary AI, or a well-capitalized defense contractor from entering the same market with deeper procurement relationships and existing MIL-SPEC certification infrastructure?
