Closing AEInnova: What Eleven Years Taught Me

I left AEInnova on 22 March 2022. This month, it's closing for good.

I want to write down what I actually think happened before the story hardens into something simpler than it was, before it becomes "another startup that failed" and nothing else. The truth is more specific than that, and more useful, I think, to anyone building something similar right now.

Where it started

AEInnova was a spin-off from the Universitat Autònoma de Barcelona, founded on 17 October 2014. I was there the day it was founded (the photograph accompanying this post was taken then). The founding team was fairly insular, a group of researchers who already knew each other well. A few months earlier, they had been my professors. Now they were my cofounders and coworkers, and it was strange and exciting in equal measure, sitting across the table, as an equal, from people I'd looked up to in the classroom. That's a good way to start a company. It's also a way to inherit blind spots that nobody inside the group notices, because everybody came up through the same lab, the same funding logic, the same habits of thought.

One of the earliest signs that this might actually go somewhere was getting selected by Fundación Repsol, early enough that we still had no real budget for anything, travel included. So we took a night bus from Barcelona to Madrid to be there. It's a small thing, but it's stayed with me as the shape of those first years: a real institution taking us seriously enough to want us in the room, and us showing up for it having spent the night upright on a bus because that was what we could afford. It felt harder in hindsight than it did at the time. At the time it just felt like the company was real.

The first product we tried to build recovered waste heat and injected the energy back into the grid. It was a good idea on paper and a hard one commercially, mostly because the amount of energy you could actually recover this way was really low, nowhere near enough to justify the cost of the equipment needed to capture and feed it back in.

 

I remember exactly where that became undeniable. We installed the first prototype of the waste heat recovery unit at SEAT Martorell, during a summer maintenance shutdown. We had to cut into a chimney to put our module in, with a huge crane holding the rest of the chimney section suspended in the air while we worked, and once the module was in, we put the chimney back on top of it. It's the kind of image that sounds almost absurd written down, a crane holding a piece of industrial chimney over your head while you install a small module meant to harvest a sliver of the heat passing through it, but it's a fair picture of the gap between the idea and the return. All that effort, that access, that risk, for a genuinely small amount of recoverable energy. The margins didn't work at our scale, and the sales cycle for anything touching grid infrastructure was long enough to bankrupt a fifteen-person company twice over before the first contract closed. I was the one who pushed hardest to kill that direction, arguing we should stop trying to inject energy back into the grid and instead use the same heat-harvesting core technology to power something cheaper to deploy and easier to sell: wireless sensor networks, powered by waste heat instead of batteries. That pivot happened, and it became the actual company, InduEye, ATEX certification, the whole later chapter. I still think it was the right call.

The pivot cost us people, though. A few of the founders had joined for the original vision, grid-connected energy recovery, and once the product stopped being that original product, they left. They were never really there for the company as such, they were there for that specific idea. We brought in new hires to fill the gaps, and that shuffle changed more than the org chart. New responsibilities landed on whoever was still there, myself included, and I ended up with team leadership on top of everything else I was already doing.

There was another kind of friction underneath all of this, one about how we actually worked. By this point we had people putting in real full-time hours, eight hours a day, five days a week, building the company as their main job. But some of the founders were still only giving it spare hours on the side, fit in around whatever their main job actually was. And yet, because they were founders, their decisions carried weight over people who were putting in far more time and who, frankly, could have stepped into those same roles without much trouble. It was a strange setup to keep running, and the only reason it kept running as long as it did was the company's origins, the fact that being there first came with authority that had nothing to do with how much you were actually there.

What we built

Industrial machinery runs hot. Pumps, motors, compressors, turbines, all of it radiates heat that normally just disappears. We built devices that captured a sliver of that heat and used it to run wireless monitoring hardware indefinitely, with no batteries and no wires. The timing helped: NB-IoT, Sigfox and LoRaWAN were all emerging around then, low-power, long-range connectivity standards that made battery-free, self-powered sensors a genuinely practical idea rather than a lab curiosity.

Most battery-powered industrial sensors run on non-rechargeable lithium batteries that eventually need replacing, and in an ATEX environment that's not a small thing. It means access, labour, maintenance planning, and an intervention in a space where every intervention comes with extra procedure and risk attached. Ours used no batteries at all, so there was nothing to replace and no cable to run either. That solved a maintenance headache plant operators genuinely felt.

It worked, and we got there technically. We picked up ATEX certification, which put us among a small number of European startups manufacturing heat-powered sensors safe for use in explosive environments, and along the way we collected national and international awards, EU funding, and enough validation pilots with large industrial clients to believe we had something. Good engineers, a product that did what we said it would do. None of that was the problem. The problem showed up after "it works," in what we decided to do with it.

If I'm honest about why the company lasted as long as it did, a big part of the answer is that we were genuinely good at winning European funding. We got consistently good at writing proposals, hitting the right calls, and bringing in grant money year after year, and that funding did as much to keep the lights on as anything we sold. It's not a comfortable thing to admit in an article about product decisions, but it would be dishonest to leave out: without that money, we probably wouldn't have had eleven years to make the mistakes I'm describing here. The funding bought us time. It didn't buy us a business.

The pivot I pushed for, and the fight I lost

Winning the argument about the grid-injection pivot early on didn't mean I kept winning arguments later. When I pushed for the wireless sensor network direction, I had a specific vision for it: not just sensors that report data, but a system that could plug directly into a plant's existing SCADA infrastructure where one existed, and stand on its own as a monitoring application where it didn't. Two paths into the same customer, and both had to work without leaning on us being alive and online forever.

At some point I was stretched across too many things at once and delegated the application side of the product to someone else on the team. That was a reasonable thing to do, I couldn't hold every piece myself, but it's also roughly where the direction started slipping out of my hands. The person who took it over convinced the CEO that a cloud-first approach was the way to go, and over time the device became architecturally dependent on our own cloud service just to be useful at all.

I want to be precise about what actually bothered me here, because it's easy to flatten this into "the cloud was the mistake," and that's not quite it. Plenty of good industrial products lean on the cloud. What bothered me was that our hardware stopped being able to deliver its core value on its own, without our specific service running on our specific servers. I warned, more than once, that we were tying a piece of industrial hardware, sold to a client who might use it for a decade, to something that only existed as long as a fifteen-person startup kept its lights on. No one listened, or at least nothing changed.

There was a second, more practical problem with that setup. Heavy industry treats operational technology and IT as separate worlds for good reason, and security teams tend to be wary of anything that opens a path between the plant floor and the outside internet. A device that needed our cloud to function wasn't only a bet on our own longevity, it was also one more thing for a customer's IT and security people to push back on once they got involved in a purchase, which in this industry they always eventually do.

I don't think this was a fight about who was smarter, it was a fight about what kind of company we thought we were. The cloud-first push wasn't an accident either, it followed from a real strategic choice: in leadership meetings, the CEO said plainly, more than once, that he wanted AEInnova generating recurring revenue rather than depending mostly on hardware sales. I understood the appeal, investors tend to value predictable subscription revenue and software-style margins more than the messier economics of selling a physical device once. But in my view the business model started dictating the product, rather than the other way around. We were shaping the architecture around the revenue model we wanted instead of the one the deployment reality actually called for. I thought we hadn't earned the right to make that bet yet, that we needed to be a hardware company that worked first, standalone, SCADA-compatible where it made sense, and worry about recurring revenue once we had a base of customers who trusted us enough to wait around for it.

Why I left

I left because I could see where the cloud-dependency decision led, having said so out loud more than once with no result, and staying to watch it happen felt worse than leaving. It wasn't a dramatic exit, no blow-up, no single bad day, just a growing certainty that the company was optimizing for a version of itself that didn't match its size, and that I'd said what I had to say about it already. I was also, honestly, ready for something more. Several years inside one company, however formative, don't hand you everything you need to keep growing as an engineer, and I could feel I'd reached the edge of what staying could still teach me. That said, I don't want to undersell what those years gave me. I grew enormously there, and a real part of who I am professionally now was built inside AEInnova, whatever else is true about how it ended.

Leaving didn't mean disappearing. I stayed an owner and stayed in touch with the founders, which gave me some visibility into how the company performed in the years that followed. That's part of why I feel able to write this at all, I wasn't purely guessing from the outside. But I also want to be honest about the limits of that view. I wasn't in the daily decisions, every customer conversation, or the full financial picture, so what follows is my interpretation of what I saw, not a definitive account of AEInnova's final years.

What that vantage point showed me was this: for most of the years I was there, I was pushing a lot of the new ideas, and just as often pushing back on the ideas other people brought in. After I left, that back-and-forth seemed to fade. Fewer new ideas got seriously challenged, at least from what I could tell. The company kept running largely on decisions I'd already set in motion before I left, plus the one decision I fought hardest against and lost, the cloud dependency. It's a strange thing to watch from a distance, a company running for years on momentum you helped build, while the one call you couldn't win quietly shapes how the story ends.

Four years is a long time to believe you were right about something you had no power to change. I don't say that to be smug about it, plenty of founders push for the wrong things and get lucky, and plenty push for the right things and still fail for other reasons anyway. I don't think my read on the cloud dependency question is the whole explanation for why AEInnova is closing. Startups die from a hundred contributing causes, and revenue never showing up in enough volume, over enough years, is the blunt instrument that actually kills a company, whatever strategic bet did or didn't pay off along the way. But I think this is part of the story, and it's the part I actually have something to say about.

What I'd tell someone at year three

If you're building hardware and you're small, be honest with yourself about what "recurring revenue" actually costs you. It's not free, it's a promise, and promises have a price, paid in trust, that you need to be big enough to keep. Using the cloud isn't the risk in itself. Making your product's core function depend on your own company's continued existence is. A product that works without you is a worse pitch deck and a better company, and I'd take the second one again.

I don't know if AEInnova's underlying technology was ever the real problem. I think the thermoelectric energy harvesting idea was sound, is probably still sound, and someone will eventually make real money on some version of it. What I know is that eleven years with no durable revenue stream is its own answer, whatever the full cause turns out to be. I'm writing this down mostly for myself, while it's still fresh, but also because I'd rather someone who was closer to the later years read this and push back on the specifics than have it settle into a vague story about the market not being ready. I don't think the market was the problem. I think we were.

---

If you were closer to this than I was in the years after I left, or if you think I've got the feasibility argument wrong, I'd genuinely like to hear it. This is my read from the outside, not the last word.