The companies building the physical execution layer for AI are worth tens of billions. Figure AI: $39 billion. Apptronik: $5.3 billion. Physical Intelligence: $11 billion. Every one of them is private. The rounds are institutional size. You’re not getting an allocation.

AI has solved the intelligence layer. What it hasn’t solved, and what it’s actually making more urgent by the day, is the physical execution layer.

Think about what’s being planned right now. Hundreds of gigafactories. A $1.7 trillion reshoring wave pulling manufacturing back to the US. An energy transition at a scale the world has never attempted. And that’s before you look at what already exists. The American Society of Civil Engineers puts the US infrastructure investment gap at $3.7 trillion. Look at Europe. Every major city is sitting on decades of deferred physical work. Housing supply hasn’t kept pace with demand since your parents were young. Streets designed for a different century. Logistics infrastructure from the 1970s.

The physical backlog is staggering. And the workforce to address it doesn’t exist.

By 2030, 2.1 million manufacturing jobs will go unfilled (Deloitte/NAM). The construction workforce is set to lose 53% to retirement this decade. The Semiconductor Industry Association projects 58% of needed semiconductor technicians won’t exist when the fabs come online. Skilled trades are running a 4:1 demand-to-supply ratio today: 600,000 job postings and 150,000 new workers entering the workforce per year.

AI doesn’t just create new projects that need physical execution. It’s accelerating the reckoning with a backlog of physical work that has been accumulating for decades. And here’s the part most people miss: the same AI making the bottleneck worse is also speeding up the solution. Better models mean faster robot training, faster iteration on hardware, faster path to deployment. The demand curve and the supply curve are both moving, but demand is moving faster. That gap is the trade.

The capital knows. $55.8 billion flowed into robotics in 2026 alone. Goldman revised their humanoid market projection from $6 billion to $38 billion by 2035. That’s a 6x revision forecast change in under two years.

The Pentagon’s FY2026 budget included a $13.4 billion dedicated autonomy and robotics line, the first time it has existed as a standalone category. FY2027 projections show $54.6 billion for the Defense Autonomy Working Group, a 180x expansion in 18 months.

Amazon has deployed over 1 million robots across its fulfillment network. Productivity per worker went from 175 packages per year to 3,870: a 22x improvement. Morgan Stanley estimates Amazon’s robots save the company $4 billion annually, rising to $10 billion by 2030.

Apple moved its secret robotics unit into the hardware division in April 2025. Caterpillar announced a partnership with Nvidia in 2026 to bring autonomous systems to construction and mining equipment. China set a national target of 100,000 humanoid robots deployed by 2027.

These are not speculators. These are the entities that move the world. And they’re all moving in the same direction.

Nvidia did 20x from the ChatGPT launch through. Microsoft, the hyperscalers, all of them captured real upside on AI. Those were all good trades. But the biggest returns in any technology wave don’t come from the giants. They come from getting into the right companies before they’re publicly accessible.

SpaceX just had the biggest IPO in history. The numbers were absurd. But the people who made generational money on SpaceX weren’t buying the IPO. They were in years earlier, when it was still a speculative rocket company. The majority of the move happened in the private rounds, before anyone could touch it.

That same story is loading again right now. OpenAI is heading to a trillion-dollar IPO. Anthropic is right behind it. You didn’t miss these because you didn’t see them coming. You missed them because the private rounds weren’t open to you. Sovereign wealth funds and top-tier VCs locked it up. By the time it trades on an exchange, the structure of who wins is already decided.

The BOTT humanoid robotics ETF is up 116% in the last 12 months. A broader robotics basket returned 167.6% while the S&P returned 16%. The repricing in robotics has started. Figure AI, Apptronik, and Physical Intelligence are the names that will define physical AI the way OpenAI defined software AI. They’re raising at billions. They’ll IPO at multiples of that. Boston Dynamics has targeted a Nasdaq listing for 2027 and that IPO will not be cheap. The leading humanoid companies are staying private by design: private capital is abundant and public market scrutiny would constrain the hardware pivots they’re still making. These rounds go to institutional allocators and established VC funds. There is no clean public instrument for this thesis. At least, there wasn’t.

You're reading this before it's obvious. That matters.

In the late 1990s, Masayoshi Son watched the internet coming before most institutional money had even processed it. His move: build a listed vehicle, SoftBank, to hold concentrated stakes in private internet companies that the average investor couldn’t access. The logic was simple. Own the infrastructure before the crowd. Let the listed vehicle do the access work for you. SoftBank’s early bets on Alibaba, Yahoo Japan, and Vodafone turned it into one of the most consequential investment vehicles of the internet era.

Andrew Kang looked at robotics in 2023 and saw the same setup. Kang is an entrepreneur and investor who built his reputation running Mechanism Capital, one of the better-known early-stage crypto funds. Not the typical robotics investor. Most VCs he consulted at the time told him NOT to invest in robotics. He went the other direction. He put $190 million into Figure AI’s Series B at a $2 billion valuation. Figure is now valued at $39 billion. That’s nearly a 20x on paper before a single share has traded publicly.

In May 2026, Kang launched RoboStrategy (Nasdaq: BOT), the first public closed-end fund dedicated exclusively to private robotics and physical AI companies. His own framing: “SoftBank was the original pioneer of this.” The structure is deliberate. A listed fund that holds private equity stakes, so that anyone with a brokerage account can get exposure to companies they would otherwise have no path into.

The thesis is straightforward. If humanoid robotics is a multi-trillion-dollar market, and the leading companies are going to stay private for several more years before they IPO, you need a vehicle that does the access work. BOT is that vehicle.

BOT is a closed-end fund registered under the Investment Company Act of 1940. 15 companies across the physical AI stack. The three largest positions are equal-weighted at roughly 19% each.

Figure AI ($37.25 million via Series B SPV): Brett Adcock’s humanoid company, currently valued at $39 billion. Backed by Microsoft, OpenAI, Nvidia, and Jeff Bezos. Has BMW as a commercial deployment partner. One of the most-watched humanoid companies in the world.

Apptronik ($37.25 million across two tranches): Building Apollo, a humanoid designed for industrial deployment. Google has invested $870 million in total. NASA is a development partner. Targeted at logistics, manufacturing, and hazardous environments where labor is hardest to place.

Dyna Robotics ($37.25 million, Series A direct): Early-stage, focused on dexterous manipulation, the hardest unsolved problem in physical AI. Getting a robot to handle objects with the precision of a human hand is the bottleneck that separates demo robots from deployable ones.

Then Dexmate, Standard Bots, Path Robotics, and smaller positions covering medical robotics, defense autonomy, defense, and logistics infrastructure.

The mechanism they call R.I.S.E. works like this: BOT issues new shares at a premium to NAV, uses the proceeds to acquire more private robotics equity, which pushes NAV per share higher for existing holders, which defends the premium, which lets them issue more shares. It compounds.

The closest precedent is MicroStrategy. MSTR ran the same playbook with Bitcoin: issue shares at a premium, buy more BTC, NAV rises, repeat. BOT cites that 97% of MSTR’s 55x Bitcoin-per-share gain since 2020 came from that premium harvesting, not from BTC price appreciation alone.

BOT has $2 billion in committed financing from Roth Capital to run this at scale. Between June 19-25, they drew $36.5 million at $29.65 per share. NAV per share jumped 23% in three weeks, from $7.24 to $8.92. The flywheel is already running.

BOT trades at roughly 3x its stated NAV. On the surface that sounds expensive. But NAV here is based on the last VC round each portfolio company raised, not any live market price. These companies don’t trade. There’s no ticker to check. The $39 billion Figure AI valuation in the NAV is what someone paid in a private round months ago, not what Figure is worth today or what it will IPO at.

So the 3x premium may not be a premium at all. If Figure’s real value today is $70 to $80 billion, the NAV is understated and you’re closer to fair value than the headline suggests. The metric isn’t wrong, it’s just measuring the inaccurate thing.

What you’re actually buying is IPO optionality. The bet is simple: these companies go public at multiples above their current private marks, NAV resets overnight, and the flywheel compounds from there. That’s the structure working before a single portfolio company has gone public.

Not saying buy it. Not saying don’t.

But if you believe robotics is the next infrastructure wave, if you believe the labor shortage is real, the reshoring is real, the Pentagon budget is real, and that Figure and Apptronik and the companies being built right now are going to define the next decade of physical work, then there’s finally a way to trade that conviction without being a VC or a sovereign wealth fund.

BOT is the only listed vehicle that gives retail access to the private robotics companies that are actually leading this. Not the ETFs holding the public names. Not Nvidia at $3 trillion. The private companies themselves, before they IPO, at the stage where the real upside has historically lived.

That’s the point. If the thesis is right, the access was always the hard part. Now it isn’t.

Personally trading this on Hyperliquid via the BOT perp trade.xyz just launched. 10x leverage, 24/7, USDC-settled. If you want to follow the trade, I post my positions on X; link in bio.

The chain keeps the score.