Brett Adcock’s AI hardware company Hark closed a $700 million Series A on May 21 at a $6 billion mark with Nvidia, AMD Ventures, Intel Capital, and Qualcomm Ventures all on the cap table, per TechCrunch. Mind Robotics, RJ Scaringe’s industrial robotics spinout from Rivian, has raised roughly $1 billion in six months across seed, Series A, and Series B, with Kleiner Perkins leading the $400 million B in May and Rivian serving as both first customer and shareholder, per The Robot Report. Sereact raised a $110 million Series B led by Headline on April 27, a four-times step-up from its Series A in fifteen months, for its Cortex 2 vision-language-action model, per Bloomberg. Physical Intelligence is sitting at $1.1 billion of total capital at a $5.6 billion mark and is reported to be in market for an additional $1 billion follow-on, per PYMNTS. Four data points, one finished-business question. The capital that is being priced for physical AI and industrial robotics is paying for the foundation-model and orchestration layer above the hardware, not the hardware itself. Founders who finish in this category decide which layer they are building before the engineering plan locks in, not after the prototype is in the field and the comparable is already pointing at the layer above them.
If You Are Building a Company in This Environment
The default first-time physical AI or industrial robotics founder treats the hardware as the primary product and the intelligence layer as a feature that ships inside it. The build-phase logic is to clear the mechanical design, prove the form factor, sign a lighthouse customer, and let the software roadmap fill in over the next two years against revenue. The engineering plan is sized to that sequence. The hiring plan is sized to that sequence. The fundraise pitch positions the company as a robotics company first and a software company second. The trouble with the sequence in 2026 is that the buy-side is no longer pricing the company that way. The strategic chip investors who showed up on the Hark Series A want exposure to the silicon-demand engine the intelligence layer will run on. The growth investors who priced Mind Robotics at platform-level marks want the foundation model that compounds across the install base. The Series B that traded at a four-times step-up on Sereact priced a robot brain, not a robot. The founder who closes the next round with a hardware-first pitch deck and a software-second roadmap is going to get priced as a hardware vendor inside someone else’s platform, at a multiple that does not justify the depth of the engineering organization the company already has.
The version of the company that finishes through this environment is the one whose founder makes the platform-versus-hardware decision early enough that the engineering organization, the customer architecture, and the capital plan all reflect a single answer. If the company is going to be the foundation-model layer, the first customer is structured as a training-data partner, the supply chain is designed to ship across third-party hardware OEMs, and the round is raised against a software-comparable. If the company is going to be the hardware layer, the engineering plan goes deep on the form factor and the platform layer is sourced from someone like Sereact or Physical Intelligence on day one, freeing the company from a multi-year software roadmap promise that the institutional buyer has stopped paying for. The trap a founder cannot afford to keep open is the third option, which is to leave the question unresolved and let the engineering organization spend years optimizing both layers at the same time without the cap table reflecting either pricing model. That is the company that does not finish.
The same dynamic is now operating across MedTech, diagnostics, defense hardware, climate hardware, and any regulated or capital-intensive category where institutional buyers are converging on a posture of paying above the asset for the cadence and orchestration layer the asset runs inside. The CMS-FDA RAPID coverage pathway is doing the same work in MedTech that DAWG is doing in defense and that Kleiner Perkins is doing in physical AI, which is signaling that the durable procurement appetite is one layer up from the asset itself. The book’s framework exists to make the architectural decision early enough that the company is positioned for that buyer rather than for the buyer it would have been priced against three years ago.
What the Four Physical AI Rounds Show as a Case Study
The Hark, Mind Robotics, Sereact, and Physical Intelligence rounds are useful as a case study because each fixes a different architectural decision a finished physical AI company has to get right. Hark fixes the strategic capital question. The Nvidia, AMD, Intel, and Qualcomm Ventures syndicate is the strategic supply chain pre-loading the demand for what their silicon will run, which means the window to bring chip strategics in at a defensible mark is open and the founders who read the signal early are getting cap-table positions that the next cohort will not be able to replicate. Mind Robotics fixes the customer architecture. The Rivian customer-and-shareholder structure is a single-environment training partnership that lets the foundation model compound across every future deployment, which is a fundamentally different first-customer shape than a traditional pilot. Sereact fixes the OEM-or-platform question. A third-party robot-brain layer that ships inside the hardware OEM’s form factor either competes with the platform layer the founder thought they were building or partners with it, and the founder has to decide which. Physical Intelligence fixes the funding sequence. A pre-revenue $5.6 billion mark with a $1 billion follow-on in market sets the comparable that the next round of foundation-model companies will be priced against, and the founder who is not credibly competing for that comparable will be priced for a different category entirely.
The Five Questions for the Platform-Versus-Hardware Decision
The five-question framework in Founders Who Finish rewrites the operating plan when the capital environment is rewriting itself faster than the engineering plan can keep up with. Each question maps to a decision that has to be settled before the engineering organization commits to a multi-year build sequence that the buy-side may not pay for.
Question 1
What are you actually finishing?
If the answer is the robot, the form factor, or the end-of-arm tool, the company is finishing a product. The finished business is the layer the institutional buyer will pay a software comparable for, which in 2026 means the foundation model, the orchestration layer, or the platform that compounds across hardware instances. Founders who finish in physical AI and industrial robotics identify which side of that line they are on and write the engineering plan to support that finish from the first capital decision forward.
Question 2
Who decides you are done?
The first manufacturing customer reads the operational capability. The strategic chip investor reads the silicon demand the intelligence layer will create. The institutional growth investor reads the comparable Mind Robotics and Physical Intelligence have just set. The strategic acquirer reads which layer the company owns and which layer they would have to buy from a competitor. A finished company in 2026 has to read convincingly to all four, and the operating plan is what aligns the answer across them.
Question 3
What does your evidence actually prove?
Sereact has 200 deployed systems and more than 1 billion production picks, with roughly one in every 53,000 requests requiring remote human intervention. That is evidence the buy-side reads as a robot-brain that works at scale. Mind Robotics uses Rivian’s factory as a training environment and is priced on the model that compounds across future customers. The evidence plan a founder writes has to produce the data the buy-side is paying for at the layer the company is building, not the layer the engineering organization finds it easiest to instrument.
Question 4
What does your path to contract, scale, and capital actually look like?
The path to contract has to fit the customer architecture the company has chosen, which for a platform-layer company means structuring the first customer as a training and integration partner rather than as a fixed-quantity buyer. The path to scale has to be financed against the layer the buy-side is pricing, which in physical AI today is the foundation-model layer for growth investors and the orchestration layer for strategic acquirers. The path to capital has to be priced against the Mind Robotics, Hark, and Physical Intelligence comparable, not against the airframe or end-of-arm-tool multiple that used to apply. Founders who finish put all three work streams against the same operating plan.
Question 5
What does the finish line look like to your buyer, your investor, and the counterparty pricing the contract?
The strategic acquirer is pricing the foundation-model and orchestration layer above the hardware. The growth investor is pricing the model that compounds across the install base. The chip strategic is pricing the demand engine for the silicon. The buy-side is converging on a posture where the layer above the asset commands the multiple. The slot the founder picks at that layer is what determines whether the finished business reads as a scarce platform or as a hardware vendor with a roadmap promise. Founders who finish make the slot decision early enough that the comparable lands cleanly when the round is priced and the strategic conversation begins.
Founders Who Finish
The guide for founders building in regulated and capital-intensive markets
The five-question framework for building medical device, diagnostics, defense, climate, industrial robotics, and physical AI companies that finish what they start, in the procurement, capital, and platform-layer environment as it actually exists.
Get the Book