After publishing on a wave energy proposal for offshore data centers, I received a useful challenge. A reader pointed to CorPower Ocean as a counterexample. That was worth taking seriously. CorPower is not a render-first startup selling a fantasy of floating artificial intelligence infrastructure in the deep Pacific. It has been around since 2012. It has built a full-scale device. It has deployed it offshore in Portugal. It has exported electricity to the grid. It has survived large Atlantic storms. It has attracted credible public and private funding. It has a more sophisticated engineering story than most wave-energy firms ever manage. That makes CorPower more interesting, not less. The right question is not whether the concept is physically plausible. It is. The right question is whether serious engineering is enough to overcome the reference class for wave energy, marine machinery, offshore maintenance, seals, corrosion, biofouling, reciprocating rods, gearboxes, and small-unit fleet economics. That is a harder question, and it is the one that matters. CorPower’s founding story isn’t the usual clean-tech origin myth. The company traces its core inspiration to Swedish cardiologist and inventor Stig Lundbäck, whose work on the pumping dynamics of the human heart informed the idea of a compact wave-energy converter that could tune and detune its motion, absorbing energy in ordinary seas and protecting itself in storms. That is a good story, and in this case it appears to have led to real engineering rather than just a metaphor stretched into a pitch deck. But it also raises one of my standard red flags. Energy hardware startups founded around insights from outside the target industry can sometimes produce useful lateral thinking, but they more often underestimate the accumulated brutality of the domain they are entering. The ocean is not a patient circulatory system. It is salt, grit, fouling, corrosion, storms, vessel schedules, insurance, and maintenance invoices. A heart-inspired machine can be clever. It still has to survive like offshore industrial equipment. CorPower is a point-absorber wave-energy converter. In simple terms, it is a floating buoy that moves relative to a structure connected to the seabed. That vertical motion is converted into useful electricity through a power take-off system inside the buoy. CorPower’s distinctive claim is that its WaveSpring system allows the machine to tune itself to ordinary waves, amplifying motion when energy capture is useful, and detune itself in storms, reducing motion when survival matters. Two of CorPower’s public claims sound odd at first. One is that 1 m waves can produce about 3 m of machinery motion. The other is that in large waves, the device can be detuned so that it becomes partly “transparent” to wave energy. Neither claim breaks physics. A resonant system can have motion larger than the forcing motion. A playground swing is the easy analogy, because small pushes at the right time can build much larger movement. A detuned system can also reduce its response when conditions are dangerous. Motion is not energy. A 1 m wave producing 3 m of machinery motion does not mean free energy has appeared. It means the system is accumulating and exchanging energy across cycles. The right test is available wave power, capture width, power take-off efficiency, capacity factor, downtime, maintenance, and annual MWh. A 300 kW device at a claimed 40–60% capacity factor would produce roughly 1,050 to 1,580 MWh per year before downtime. At 50%, the midpoint, it would produce about 1,314 MWh per year. That is not trivial, but it is a small offshore machine. Small matters offshore. A 10 MW CorPower-style array requires about 34 devices. Each device carries a floating body, a seabed-connected structure, rods, seals, scrapers, coatings, moorings, power cables, sensors, controls, a power take-off system, and a maintenance story. Every one of those elements has to survive saltwater, storms, marine growth, cyclic loads, and long gaps between convenient vessel windows. Offshore renewables are not judged by the heroism of the first machine. They are judged by how boring the fleet becomes. The “transparent to waves” claim also needs to be read carefully. It does not mean invisible to ships, zero load, or zero risk. It means lower hydrodynamic response in survival mode. That may be useful for storm survival, but low-freeboard or partly submerged wave devices in heavy seas are still marine obstacles. A field of them has to be charted, marked, lit, and managed through aids to navigation, notices to mariners, and practical exclusion or avoidance zones. A wave farm becomes industrial sea space, not a harmless addition to open water. Storm survival is a qualification test. It is not a business model. The siting comparison is harsh. CorPower’s target depth range overlaps with good fixed-bottom offshore wind territory. Forty meters of water is not exotic for offshore wind. It is close to the middle of the modern fixed-bottom opportunity space. Offshore wind turbines are already bankable, industrialized, supported by mature supply chains, and scaling into the 20 MW class. One 20 MW offshore wind turbine has the nameplate capacity of roughly 67 CorPower 300 kW devices. It also has far more capacity per major serviced machine and per offshore maintenance campaign. That is a brutal comparison for any small moving machine in the ocean. That does not mean wave energy has no possible niche. It might have value co-located with offshore wind if it can share grid connections, substations, cables, ports, vessels, and consenting envelopes. It might have value where wind is constrained by visual, military, radar, shipping, fishing, or permitting barriers. It might have value for islands with expensive diesel generation and strong local wave resources. It might have value near harbours, aquaculture, desalination, or remote industrial loads where local resilience has more value than wholesale electricity. But those are niches. They are not yet a broad energy-transition market. The mechanical concerns are where the public story starts to look difficult. Externally, CorPower’s device presents two parallel, equal-sized rods descending from the buoy to the seabed attachment, looking a lot like a motorcycle’s front fork shock absorbers. The buoy moves relative to those rods and the lower structure. That means rods, seals, scrapers, coatings, grease systems, cathodic protection, alignment, and marine growth management are not peripheral details. They are central to the economics. Exposed reciprocating marine interfaces are not impossible. Offshore and subsea systems use rods, seals, wipers, coatings, and hydraulic components all the time. The question is whether this can be done cheaply, predictably, and with long service intervals in a small 300 kW wave machine. The risk pathways are mundane and severe. Biofouling can build up on rods and nearby surfaces. Scrapers can wear or clog. Seals can abrade. Coatings can be damaged. Corrosion can appear around seal gland assemblies. Salt, grit, biological material, shell fragments, and corrosion products can enter the working environment. Grease systems have to keep working. Alignment has to stay within tolerance under cyclic loads. CorPower’s own post-deployment inspection reinforces the concern. After its first ocean campaign, the company reported lessons in biofouling, corrosion and robustness. It upgraded the tidal regulator under the device with a new grease system and improved seal and scraper solution. It reported good performance from some rod coatings, but also corrosion around parts of the seal gland assemblies where cathodic protection connections had been inadequate. That is not a scandal. It is what real engineering development looks like. It is also not a proof point for commercial bankability. It shows that the exact areas one would worry about from the outside are the areas that required post-deployment improvement. Wave energy’s enemy is not the first storm. It is the thousandth ordinary operating day. The second mechanical concern is the internal power take-off system. Wave devices face a hard conversion problem. The ocean gives slow, high-force, reversing vertical motion. The grid wants controlled electricity. CorPower’s Cascade Gearbox is a clever answer, distributing load across multiple small pinions and converting linear motion into rotation. CorPower has also done more serious testing than many marine-energy companies. It has invested in dry hardware-in-the-loop testing, high-load cycling, and staged validation. It has reported testing loads up to about 4 tons, and the first C4 deployment did not come with public evidence of a failed gearbox. That matters. But the gearbox remains a high-cycle fatigue and maintenance risk until there are fleet data. CorPower states that the Cascade Gearbox is designed for over 100 million load cycles. That sounds reassuring until the arithmetic is done. A device operating in waves with periods around 5 to 10 seconds sees roughly 3.2 million to 6.3 million cycles per year if active continuously. A 100 million-cycle target corresponds to roughly 16 to 32 years of continuous cycling. That is a design target across a long operating life, not public proof that a fleet has achieved it. Gear teeth, bearings, racks, lubrication, alignment, load-sharing, torque reversals, control transients, and generator coupling all have to stay in acceptable ranges across years of real sea states. This is where Flyvbjerg’s reference class forecasting becomes useful, especially when paired with a simple Monte Carlo simulation. For technologies with limited fleet data, the right question is not what the company hopes, or whether the first machine survived storms. The right question is what similar machines tend to do under similar conditions, then what happens when those outside-view failure rates are repeatedly sampled and scaled across a 34-device, 10 MW array. Reference class forecasting does not tell us CorPower will fail, and the Monte Carlo simulation does not predict its actual future. Together, they create a stress-tested outside view of what CorPower has to beat. For CorPower, the two most relevant reference classes are the two mechanical risk areas already described: external reciprocating marine interfaces and internal high-cycle drivetrains. Public wave-energy reliability studies are sparse, but they are not silent. Generic 300 kW point-absorber models show actuator leakage, seal failure, rod corrosion, mechanical failure, and bearing failure as meaningful contributors to maintenance risk. Project summaries on wave-energy power take-off systems also identify leakage, fatigue, and foreign material as recurring issues. Using a public-data outside view, then adjusting for CorPower’s better-than-average engineering and testing, the exposed rod, seal, scraper, grease, and fouling system forecasts as the larger near-term risk. For the external-interface bucket, I would use a median outside-view rate of about 0.30 significant events per device-year. The gearbox and power take-off bucket forecasts lower because CorPower has done serious dry testing and because there is no public evidence of first-device gearbox failure. For that bucket, I would use about 0.12 significant events per device-year. Combined, that gives about 0.46 significant mechanical events per device-year. Plain English matters here. A combined rate of 0.46 events per device-year means about one significant mechanical intervention every 2.2 years per device. That is not a claim that every device fails catastrophically every two years. It is a reference-class stress test suggesting that, across a fleet, mechanical corrective work involving rods, seals, scrapers, gearbox, bearings, power take-off, or related systems could arise often enough to dominate the maintenance model. Now scale that to a 10 MW array. At 300 kW per device, the array needs about 34 devices. At a 50% capacity factor, it produces about 44.7 GWh per year. If the base reference-class forecast is right, that fleet faces roughly 10 to 20 major mechanical interventions per year. The optimistic case might be 4 to 7 per year. The pessimistic case might be 25 to 40 or more. The difference between those outcomes is the difference between an interesting marine-energy project and a maintenance treadmill. Infographic of likely CorPower maintenance burden by author with ChatGPT The economics are not subtle. As a round-number stress test, assume €150,000 per major mechanical intervention. That is not a wild number once vessel mobilization, weather windows, tow-back or retrieval, port handling, inspection, parts, labour, recommissioning, and lost output are included. At 44.7 GWh per year, 4 to 7 interventions add roughly €13 to €24/MWh. That might be survivable if every other cost bucket is controlled and the energy has a high-value niche. A base case of 10 to 20 interventions adds roughly €34 to €67/MWh from this mechanical intervention bucket alone. A pessimistic case of 25 to 40 or more adds roughly €84 to €134/MWh or worse. That is before routine monitoring, planned maintenance, insurance, staff, spares inventory, cable faults, mooring inspections, export equipment, project management, financing, and ordinary operational overhead. The mechanical maintenance bucket does not have to be the whole levelized cost of energy to be fatal. It only has to be too large before the rest of the system is counted. If the bankability target for this one mechanical bucket is around €25/MWh, a 10 MW array at €150,000 per intervention needs fewer than about 8 major events per year. The base case is above that. The mid-case looks uneconomic unless CorPower beats the reference class by a material margin. That is the key finding. The public outside view does not say CorPower’s physics are wrong. It does not say CorPower’s engineers are unserious. It says the company has to prove that its rods, seals, scrapers, grease systems, coatings, gearbox, bearings, and power take-off can achieve much lower intervention rates than similar marine systems would lead us to expect. It has to beat the reference class, not by a rounding error, but by enough to move the combined mechanical event rate below roughly 0.1 to 0.2 events per device-year. The evidence that would change the conclusion is straightforward, and it is the kind of evidence investors, insurers, utilities, and project finance teams will care about. Several years of ocean operation are needed. Multiple devices are needed, not one. Measured availability through winter seasons is needed. Actual MWh between interventions are needed. Unplanned retrievals per device-year are needed. Inspection data are needed after millions of cycles for seals, scrapers, rod coatings, corrosion, gearbox oil, vibration, bearings, racks, and pinions. Mean time to repair, including weather delays, is needed. Actual cost per retrieval is needed. Array-level performance matters more than single-device performance. One device for two years does not prove much. Thirty device-years with few or no unplanned mechanical retrievals would start to shift the prior. For a 34-device array, bankability is not proven by a heroic machine. It is proven by boring records. Low retrieval rates. Predictable service intervals. Clean inspection reports. Dry invoices. Crews that do not need to improvise. Parts that do not surprise anyone. Ports that are not clogged with returning machines. Availability that remains high when the ocean is inconvenient. That puts CorPower in a specific category. It is not the same thing as the latest wave-powered data center concept. It is not a cartoon. It has real engineering, real testing, real investors, and a real path through staged projects. It deserves to be treated as one of the credible wave-energy companies. But credible technical demonstrator is not the same as bankable infrastructure. CorPower is still trying to prove that its clever machine can escape the marine-energy reference class that has defeated many clever machines before it. The wider lesson is useful. The ocean is a poor place for small, complex, high-cycle mechanical equipment unless the value per machine is high, the maintenance interval is long, and the service model is boring. CorPower may yet prove that it has solved that combination. The physics look plausible. The engineering looks serious. But until the retrieval rate, service interval, and mechanical cost per MWh are boring, the economics remain unproven. It’s just another example of why wave energy is dead tech floating in my professional opinion.
