Humanoids Are the Greed Trade. The Money Is in the Boring Middle.

Last week, Barclays analyst Zornitza Todorova told CNBC the humanoid robotics market is "around $2 to $3 billion" today and will "surge to $200 billion by 2035." Same week, SoftBank’s Masayoshi Son told an audience the next "trillion-dollar enterprise" will come out of physical AI. The retail brain, conditioned by a decade of Tesla autonomy demos and Figure factory clips, hears one thing: the next NVDA is bipedal and it is coming. Buy the pure-plays. Buy the picks-and-shovels. Buy the ETFs that have not been built yet.

This is the part of the cycle where allocators get separated from speculators. Not because the Barclays number is wrong — it might be conservative — but because the path from $2B to $200B does not run through the company on the demo reel. It runs through a far less photogenic place: a warehouse outside Memphis where an autonomous mobile robot is moving a pallet right now. The deployment data published last week says so explicitly, and almost nobody on financial media bothered to read it.

The two phases Barclays actually described

Buried under the $200B headline was the part of Todorova’s framing that matters for capital allocation. Barclays calls this cycle "automation 3.0," and they split it into two phases that almost no retail commentary picked up. Through 2030: manufacturing, logistics, agriculture, construction. Industrial floors. Loading docks. Greenhouses. Job sites. Boring, dirty, measurable. After 2030: healthcare, elder care, education, hospitality. The consumer-facing humanoids. The ones retail is being sold today.

Read those two phases like an operator. The first phase is a buyer’s market — the customer has a CFO, a CapEx committee, a per-pallet cost spreadsheet, and a willingness to sign a five-year lease on a fleet that returns 18%. The second phase is a consumer adoption problem with regulatory risk, liability risk, and a hardware bill of materials that has to compete with human wages in the most labor-cost-sensitive industries in the economy. Phase one is shipping. Phase two is a pitch deck.

If you are buying the public humanoid story in 2026, you are buying phase two. The companies in phase one already have customers, contracts, and revenue. Most of them are private, late-stage, or buried inside industrials.

The deployment data nobody is quoting

ABI Research published a survey on June 4 covering 490 supply chain operators. The headline numbers are blunt enough to print on a t-shirt. 77% of supply chains are considering, piloting, or beginning implementation of mobile automation. 65% say AI and generative AI capability is important or very important to their next technology purchase decision. 55% plan to spend more than $100,000 on machine vision in the next two years. ABI’s qualitative read from MODEX 2026 was even sharper: the industry has moved from "experimentation to active deployment" and operators are now demanding "clear operational value — not vague promises."

Translate that out of survey-ese. The buyers stopped being curious and started being procurement. They want references. They want unit economics. They want a vendor that can quote a 36-month payback with a maintenance SLA. The hype cycle for "robots will do everything" is over inside the customer base. What replaced it is a buying cycle for things that already work.

The market that is shipping into that buying cycle is the logistics robot market. Fact.MR’s latest sizing puts it at $14.4 billion in 2025, $16.1 billion in 2026, and a projected $48.2 billion by 2036 — an 11.6% CAGR with autonomous mobile robots taking 29.4% of the product mix and warehouse automation at 33.7% of the application mix. That is not an AI fantasy curve. That is a CapEx procurement curve. And it is happening this fiscal year, not in 2035.

The Fear and Greed inversion

Here is the part allocators need to internalize. The Fear and Greed dial inside robotics is not pointed at the same companies the public market thinks it is. Retail greed is highest where the demo videos are loudest — humanoid bipeds folding shirts, sorting parts, walking up stairs. Operator greed is highest where the procurement cycle is closing — pallet movers, autonomous forklifts, vision-guided pickers, sortation systems with AI dispatch. The first crowd is bidding on a 2031 narrative. The second crowd is bidding on a 2026 invoice.

When the retail dial spikes greed, the operator dial spikes inventory. That is not a coincidence. That is how every previous automation cycle has cleared. Industrial robotics in the 1980s, semiconductor capex in the 1990s, cloud infrastructure in the 2010s — in every case, the headline trade got hyped two years after the boring middle of the supply chain had already locked in the volume. The investors who outperformed bought the boring middle. The investors who underperformed bought the avatar.

What we are watching at AdValorem

This is exactly the kind of split where edge actually lives, and it is the reason our AI and Robotics research vertical does not start with humanoid pure-plays. It starts with the question every allocator should be asking right now: which companies inside the phase-one buyer’s market are still private, still pre-IPO, and still allocating cap-table room to non-institutional capital? The answer is not zero. The answer is also not retail-accessible through a public ticker. It sits in the late-stage secondaries market, in accelerator alumni who shipped real industrial product before the 2024 robotics money got loud, and in a small set of warrant positions on companies that did the unglamorous integration work before the demo reels existed.

None of that is a deal. It is a research thesis. The point is not to predict the winner — the point is to be reading the deployment data while everyone else is watching the demo. Position before you predict. The Barclays $200B headline will still be there in five years. The procurement window for the boring middle is open right now.