Robotics just became the second largest category in private markets. Here's everything for the week!

a16z published a piece this week that deserves more attention than it got. Buried inside a broader analysis of market cycles and capital rotation is a data point that reframes how seriously the investment world is now taking physical AI and robotics. Measured by the aggregate value of the top 100 private companies by category, robotics and physical AI did not register as a category a decade ago. As of 2026, it has overtaken fintech and payments to become the second largest segment, trailing only AI and data infrastructure.

That is not a gradual climb. That is a category that went from invisible to dominant in under ten years.

The deal flow data from PitchBook backs it up. Q1 2026 set a record for robotics and physical AI on both deal value and deal count, with approximately $16 billion invested across just under 500 deals. To put that in context:

• Deal count is roughly 2x the average from the 2021 to 2025 stretch

• Deal value is approximately 4.5x larger than the same period

This is not a blip driven by one or two mega-rounds. Nearly 500 separate transactions closed in a single quarter, which means capital is spreading across the category at every stage, from seed through late-stage growth.

a16z frames this inside a broader thesis they have been building for a while: the market is rotating from bits to atoms. The post-pandemic investment cycle has flipped the script on what worked after the global financial crisis. Back then, capital-light, consumer-oriented sectors like healthcare, media, and consumer products delivered double-digit returns. Those same categories are now doing low single digits. Meanwhile, energy, raw materials, construction, and financials, all of which lagged in the 2010s, have moved into the mid-to-high double digits. Asset-heavy companies have overtaken asset-light companies after trailing for an entire decade.

Tech remains the exception as a repeat winner across both cycles, but there is a split happening within tech itself. Hardware is the real standout of the current cycle. Software, which dominated the last one, has actually followed the broader pattern of inversion and cooled off. That distinction matters because the robotics surge is not a software story. It is a hardware story, enabled by software, but grounded in physical machines doing physical work.

The a16z piece draws a comparison to electrification that is worth sitting with. When factories first got access to electric power, most manufacturers simply swapped out the steam engine for a large electric motor and kept everything else the same. The shafts, the belts, the floor layout, all unchanged. Productivity barely moved. The real gains only arrived when manufacturers realized they could put smaller motors at each individual machine and redesign the entire factory around distributed power. That was not an incremental improvement. It was a complete architectural rethink that unlocked one of the largest productivity leaps in industrial history.

The parallel to robotics and AI is direct. Most companies today are still in the “swap the motor” phase, applying AI to existing processes without rethinking the process itself. a16z cites a study of 515 high-growth startups that found firms who redesigned their workflows around AI, rather than just inserting AI into existing steps, discovered 44 percent more use cases, generated roughly double the revenue at the top end, and consumed 40 percent less capital. The productivity gains are real, but they live downstream of a discovery problem that most organizations have not yet solved.

For robotics specifically, the implication is that the current wave of investment is not just about building better robots. It is about building the infrastructure layer for a physical economy that gets redesigned around what AI-enabled machines can do, rather than slotting robots into workflows that were built for human hands and human judgment. Defense is the highest-profile frontier right now, and expanding global defense budgets are pulling capital into autonomous systems at an accelerating rate. But the a16z thesis suggests the rotation to atoms may be deeper, wider, and longer than any previous modern tech cycle, extending well beyond defense into manufacturing, logistics, energy, agriculture, construction, and eventually domestic environments.

The bottom line from this data is straightforward. Robotics is no longer a niche category sitting at the margins of venture capital. It is now the second largest concentration of private market value on the planet after AI infrastructure, and the rate at which capital is entering the space has no historical precedent. Whether this cycle produces lasting companies or an overheated correction depends on whether the firms absorbing this capital can convert it into deployed systems that generate revenue, not just prototypes that generate excitement. But the capital allocation signal is no longer ambiguous. The money has decided that the next decade belongs to machines that operate in the physical world.

This one caught me off guard. General Intuition is training physical AI models not on robot demonstrations, not on simulation data, but on billions of gaming clips that people uploaded to Medal, a platform where gamers share their gameplay moments. The company’s CEO, Pim de Witte, also co-founded Medal, which is how this whole thing started.

It sounds odd until you think about what a gaming clip actually contains. A human is looking at a 3D environment, reading what is happening around them, deciding where to move, and executing that decision in real time. Every clip comes with embedded action labels recording exactly which button was pressed and when, so you get a clean mapping between what the player saw, what they chose to do, and what happened next. That is closer to how physical AI needs to think than anything you can get from text descriptions or scripted robot demos. And the scale is hard to compete with. Nobody else in the space has access to billions of action-labeled clips across thousands of different environments, and the dataset keeps growing on its own every time someone posts to Medal.

General Catalyst led the $320 million round. Jeff Bezos and former Google CEO Eric Schmidt also participated. Total funding is now $454 million, with the company valued at $2.3 billion. What surprised me is that General Intuition has actually been around since 2015, quietly building two model types: action models that decide what to do, and world models that predict what happens after. The funding goes toward scaling compute and pretraining the next model version, with a public API expected this summer.

The open question is obvious. Gaming environments are rich and diverse, but they are still not the real world. Friction, weight, material compliance, lighting conditions, none of that translates perfectly from a rendered game engine to a physical robot handling objects on a factory floor. General Intuition’s bet is that the perception and decision-making patterns transfer even if the physics do not match exactly, and that the sheer volume and diversity of the data compensates for the domain gap. Whether that holds up when the models start driving real hardware is something the company has not yet demonstrated publicly. But the approach is genuinely novel, the data moat is real, and the investor confidence at this valuation suggests people closer to the technology believe the thesis has legs.