Why Creandum and USV co-led a $40m seed into Machenta, a robotics lab with no product

We have spent the last three years hearing that the era of cheap money is dead. We were told that founders need to focus on unit economics, clear paths to profitability, and shipping a product in weeks, not years. Then a company like Machenta comes along and blows that entire narrative out of the water.

Machenta, a London-based robotics lab, recently closed a $40 million seed round. For those keeping track, that is a massive figure for an initial check. What makes it even more striking is that they do not have a product. They are not out there demoing a prototype to customers or showing off a beta. They are building in what many would call deep stealth, focusing on the fundamental intersection of physical robotics and large-scale AI models.

The round was co-led by Creandum and Union Square Ventures. These aren't the kind of firms that throw money at every passing trend. They are looking for generational shifts. When firms of this caliber put $40 million into a lab with no commercial output, they aren't buying a business; they are buying an R&D timeline.

The Return of the Talent Bet

For founders, this is a signal that the pendulum is swinging. For a while, the market favored the 'hustler'—the person who could glue together three APIs and find a way to charge $20 a month for it. But in the world of AI and robotics, the complexity is getting so high that the 'builder' is being redefined as the deep technical specialist.

Machenta is being led by people who have spent their lives at places like DeepMind and Waymo. The investors are betting that the bottleneck in robotics isn't the hardware anymore—it is the brain. We have had robotic arms for decades, but they are stupid. They do one thing repeatedly. If you move the object an inch to the left, the robot fails. The goal now is to create general-purpose intelligence that can navigate the messy, unpredictable physical world.

If you are a builder looking at this, you have to realize that 'no product' doesn't mean 'no work.' It means the work is so hard that it requires tens of millions of dollars just to reach the starting line. This is a return to the moonshot mentality.

Why Robotics, Why Now?

The timing here isn't accidental. We are seeing a convergence. On one side, we have Large Language Models that have proven we can scale intelligence through data. On the other side, we have a desperate need to automate physical labor as demographics shift and manufacturing costs rise. The bridge between those two is the kind of work Machenta is doing.

Investors are looking at the current AI landscape and realizing that the software layer is getting crowded. There are a thousand wrappers for GPT-4. But there are very few teams capable of building the fundamental software layer that allows a machine to perceive, reason, and act in a warehouse or a kitchen. That moat is much deeper, and that is why the price tag for entry is $40 million.

The market is starting to distinguish between companies that use AI to optimize a process and companies that use AI to solve a physical reality. The latter is where the real value will be captured in the next decade.

The Founder’s Dilemma

If you’re a founder without a $40 million war chest, this might feel discouraging. But there is a lesson here in how you frame your vision. Machenta didn't get this money by promising a better vacuum cleaner. They got it by pitching a fundamental shift in how machines learn to move.

When you are in the early stages, you have to decide if you are building a feature, a product, or a platform. Machenta is clearly aiming for the platform level. They are building the 'operating system' for movement. For builders, this means if you want to attract high-tier VC interest right now, you need to be working on a problem that is technically defensible. If a weekend coder can replicate your 'moat' using a few Python scripts and an OpenAI key, you aren't going to see these kinds of rounds.

The Risk Profile

Let’s be honest: this is high-risk. Many labs like this have failed in the past. Remember Rethink Robotics? They had the pedigree and the funding, but they couldn't find the market fit fast enough. The danger for a company like Machenta is that they spend three years in the lab, build something incredible, and realize the hardware costs make it unscalable for the average business.

However, the involvement of USV and Creandum suggests they are looking at a longer horizon. They aren't expecting a return in 24 months. They are looking at the 2030s. This is 'patient capital,' something we haven't seen much of lately.

What Builders Should Take Away

If you are building in the AI space, stop looking at the shiny UI and start looking at the plumbing. The most valuable companies of the next era won't just be the ones that talk to us; they will be the ones that do things for us.

• Talent is the ultimate currency: In deep tech, your team's pedigree acts as the initial collateral for funding.
• Fundamental problems attract fundamental capital: If you are solving something that hasn't been solved in 50 years, the usual rules of 'traction' are often waived.
• The UK is becoming a hub for this: London’s ecosystem, fueled by the legacy of DeepMind, is proving to be a serious competitor to Silicon Valley for AI-heavy robotics.

We are moving past the 'AI hype' and into the 'AI application' phase. Machenta represents the most extreme version of this—building the foundation before they even know exactly what the house will look like. It’s a bold bet, and for the rest of us, it’s a reminder that the biggest opportunities usually lie where the technical difficulty is highest.

Read the original at Sifted →