Pick Your Layer Before the Market Picks It for You

Brett Adcock’s AI hardware company Hark raised $700 million at a $6 billion valuation on May 21 with Nvidia, AMD Ventures, Intel Capital, and Qualcomm Ventures all on the cap table, per TechCrunch. RJ Scaringe’s Mind Robotics has now raised roughly $1 billion in six months, with Rivian as both first customer and shareholder, per The Robot Report. Sereact raised a $110 million Series B at a four-times step-up for its vision-language-action robot brain, per Bloomberg. Physical Intelligence is reportedly in market for a $1 billion follow-on at the top of $1.1 billion already raised, per PYMNTS. Four data points, one read every founder should be making. The market just announced which layer of the physical AI stack it is willing to pay a platform multiple for, and any founder building hardware in any category needs to be honest about whether the company they are running matches the layer the market is paying for or whether they have an asynchronous business that is being financed against the wrong comparable. Understanding the market is only part of the job. Making money, building systems, and getting to where you want to go — those are the problems The Build exists to help you think through, and the platform-versus-asset question that the physical AI rounds just sharpened is exactly the kind of structural decision most founders defer until the next round forces it on them.

The Market Will Pick Your Layer For You If You Wait

The default first-time founder building hardware writes the operating plan as if the company will graduate from asset to platform over time. The build-phase logic is to ship the prototype, sign a customer, prove the form factor, and use the revenue from the hardware to fund the platform-layer software roadmap that will become the moat in years three and four. That sequence worked when the institutional capital was patient about the platform layer and willing to pay a hardware multiple until the platform showed up. The sequence does not work in a market that has just paid Mind Robotics a foundation-model multiple six months after the company was formed, that has put four chip strategics on Brett Adcock’s Hark cap table at a $6 billion mark before a single product has shipped, and that is willing to fund Physical Intelligence at $5.6 billion with no end customer disclosed. The market is no longer waiting for the platform layer to show up on a hardware company’s revenue. The market is funding the platform layer directly, on a different cap table, with a different comparable. The hardware company that was planning to evolve into the platform now has the platform layer being capitalized around it by competitors who are not slowed down by a form-factor build.

The version of the business that compounds through that environment is the one whose founder is honest about which layer they are actually building and writes the company accordingly from the first capital decision forward. Honesty here has a specific operational meaning. It means looking at the engineering organization and asking whether the people, the roadmap, and the technical debt are pointing at a hardware moat, a foundation-model moat, an orchestration moat, or a data-flywheel moat. It means looking at the customer architecture and asking whether the first deal is structured as a training-data partnership, a fixed-quantity order, an integration into a third-party platform, or a flat-rate service contract. It means looking at the cap table and asking whether the existing investor base is priced against the layer the company is building or against the layer it was sold as building. The discipline is unglamorous. It does not change the demo. It changes the structural shape of the company in ways the comparable will see when the next round is priced, when the strategic conversation begins, and when the acquirer decides whether the company is the platform they want to buy or the vendor they want to compete with.

The same dynamic is running underneath every regulated and capital-intensive hardware category right now, not only physical AI. The CMS-FDA RAPID coverage pathway in MedTech is pulling commercial timing forward in a way that rewards the platform-layer features above the device. The DAWG procurement shift in defense hardware is paying for orchestration above the airframe. The pharma-partnership pattern in diagnostics is pricing the pipeline-level collaboration depth above the panel. The grid-integration software contract in climate hardware is paying more than the asset it runs on top of. Founders who read the layer the buyer is paying for and write the operating plan to match are building businesses that will compound through the next cycle. Founders who hope the market will pay a platform multiple for the asset are going to discover that the market has already moved its checkbook to a different layer.

Three System Decisions Worth Studying

The first decision is the customer architecture decision. Mind Robotics has Rivian as both first customer and shareholder, structured so the factory floor is the training environment that compounds across every future deployment, per The Robot Report. That is a fundamentally different first-customer shape than a pilot. A pilot teaches the company about the customer. A training-data partnership teaches the model about the world. The founder who structures the first customer relationship as a training partnership ends up with a model that is already worth something on customer two. The founder who structures the first customer as a pilot ends up with a reference and a roadmap.

The second decision is the strategic cap table decision. The Hark Series A had Nvidia, AMD Ventures, Intel Capital, and Qualcomm Ventures participating, per TechCrunch. Three of the four major AI-chip strategics, plus a major mobile-compute strategic, all writing the same round at the same time. That syndicate is the strategic supply chain pre-loading the demand for what their silicon will run. For founders building any hardware that will eventually run a meaningful compute load, the window to bring chip strategics in early at a defensible mark is open right now. The founders who recognize the signal and structure the round to capture it get a strategic syndicate that the next cohort of founders will not be able to replicate without paying a higher mark.

The third decision is the comparable decision. Sereact raised a $110 million Series B at a four-times step-up from its Series A in fifteen months for a robot brain that ships inside any third-party hardware OEM’s form factor, per Bloomberg. Physical Intelligence has $1.1 billion raised at a $5.6 billion mark with no end customer disclosed, per The Robot Report and PYMNTS. Those are the comparables the next physical-AI and industrial-robotics rounds are going to be measured against. A founder who is writing the next round’s deck against the OEM-vendor multiple from 2024 is going to find the math very unfriendly. A founder who is writing the company against the layer those comparables price is going to find the round much easier to land at a mark the next strategic conversation can pay against.