Robotonomics

Markets again struggled to find direction this week. The S&P finished roughly flat (+0.1%) and the Nasdaq +0.4%, but, as has been the case in many weeks recently, the muted weekly numbers mask significant intra-week volatility. Stocks fell early in the week as President Trump teased renewed strikes on Iran, then rallied hard on Thursday (S&P +2%, Nasdaq +3%) as he floated a peace deal that would reopen the Strait of Hormuz. Crude swung in tandem, with Brent sliding from ~$95/bbl toward a roughly two-month low near $89. In equity markets, tech stocks stayed volatile on continued AI-related news. Chips rallied Thursday (Micron +11%, AMD +8%, Intel +10% on a BofA upgrade) while Oracle fell ~9% on a cloud revenue miss and rising AI infrastructure costs. YTD, the S&P/Nasdaq are up roughly 8%/11%. Our NPM Price data, which shows the average price performance of the 50 largest names in our internal data, is up 38% YTD. (Bloomberg; NPM)

Humanoid Robotics: A Framework for Industry Size, Earnings Power, and Ultimate Market Capitalization

Humanoid robotics has become one of the most talked about industries in private markets, one that is often described as a “multi trillion dollar” opportunity. However, many industry participants in our view, rely more on general sentiment vs. actually quantifying the industry in a way that connects robot adoption, revenue, earnings power, and ultimately market capitalization. Here we will attempt to provide a loose framework for that.

Our central view is that humanoid robots should not be valued simply as a “robotics” category. They should be valued, rather, as a new form of labor. The global labor market represents roughly $66tn of annual compensation (Source: OECD). If embodied AI can perform a meaningful percentage of economically useful work, the resulting industry could ultimately become one of the largest sectors in the global economy.

The key question is therefore not simply “how many robots will be sold?” but “what percentage of global labor spending will eventually migrate to humanoid robots?.”

Commercialization Timeline and Pace of Growth

Humanoid robotics has begun to cross from pilot projects into early commercial deployment (e.g. in 2Q BMW announced it would move towards deployment in factories). Multiple platforms are now in paid pilots or limited production, such as Figure AI’s 03, Tesla’s Optimus, Agility’s Digit, Apptronik’s Apollo (with Mercedes-Benz, GXO, and Google DeepMind), Unitree’s G1, and a deep bench of additional Chinese competitors. The commentary from our industry management we have met points to meaningful production scaling in 2027–28 and mass market availability, comparable to today’s industrial robots, by 2029–30. To date deployments have concentrated in a limited number of structured environments—primarily automotive and warehouse logistics—where tasks and processes are defined and repetitive. We continue to believe at home deployments of robots are at least 5 years out, and potentially far longer.

Global humanoid shipments totaled only roughly 13k-18k units in 2025 (Omdia; IDC). From there, Goldman Sachs projects approximately 1.4mm to be shipped annually by 2035 (~70% CAGR). IDC projects global shipments exceeding 510k units by 2030, a CAGR near 95%. Production capacity is being built to match these curves: Tesla is converting its Fremont plant toward a stated capacity of up to 1mm/yr, Figure’s BotQ facility is targeting >100k/yr, and Agility’s RoboFab is ramping toward >10k units.

We have attempted to provide a straw man for thinking about the ultimate TAM for investors in the sector. We start by assuming a 7 year replacement cycle for robots and various levels of adoption. Our replacement assumption is very much an average given our view that industrial robot life will be shorter (4-5 years) and home robot life will be longer.

We next try to estimate industry potential revenues on both hardware sales and services (i.e. fleet service, maintenance, upgrades, subscriptions etc.). We then assume potential margin levels and potential P/E ratios based on other hardware/industrial companies to back into potential industry market cap. We note that inputs like margin structure remain huge unknowns, but we point out to investors that ultimately margins for these companies are likely to be a blend of hardware and software economics. Over time we would expect monthly fees (note we are ignoring a potential “robot as a service” model in our analysis, which may be end up being the main model for industrial deployments) to dwarf hardware sales.

While a cranky investor could easily poke holes in our analysis, suffice to say that, even under our “downside” adoption scenario the numbers are quite large. Under our framework, the global industry could plausibly support anywhere from $500bn to $30tn of aggregate equity value.

We would note that our figures are largely corroborated by external estimates. Morgan Stanley projects a $5tn revenue market by 2050 and Citigroup models as much as $7tn by 2050. If the sector ultimately supports technology-like valuation multiples, the long-run humanoid ecosystem could plausibly command $5–20+ tn in aggregate market capitalization, exceeding today’s global automotive sector at ~$4tn. (Bloomberg)

The wide dispersion in outcomes is also part of our point: the ultimate market capitalization of the sector is a function primarily of a single variable, namely how much of the global labor market migrates to autonomous robots. Even small changes in that penetration rate swing aggregate equity value by trillions.

Where does China fit in?

Note that our analysis is for the global humanoid robotics space vs. simply the US industry, and China has already proven to be a stalwart competitor in robotics. Unitree, AgiBot, and other Chinese manufacturers already appear to be shipping substantially more units than US peers and benefit from world class manufacturing ecosystems and lower cost supply chains. The risk for US companies is that humanoids eventually resemble EVs, where manufacturing scale and cost advantages matter more than early software leadership.

Chinese manufacturers accounted for almost 90% (Omdia) of global humanoid sales in 2025, and six of the largest global producers (Agibot, Unitree, UBTech, Leju, Engine AI, and Fourier Intelligence) are Chinese. This leadership is not accidental, as the Chinese government designated humanoid robotics a national priority in 2015. China’s advantage also stems from a hardware supply chain built up through various industries, including EVs and drones. China is a leader in LiDAR, rare earth mining and a variety of controllers and components used in robots.

Are Chinese robots a threat to U.S. companies? Yes, but, in our view, a qualified one. The US appears to still lead on an important axis: AI, foundation models for robotics, autonomy software, and vertically integrated platforms.

The bear case for US robotics companies is the EV analogy: humanoids commoditize, manufacturing scale/cost win and US companies find it difficult to compete. US EVs currently have only about a 10% global market share (S&P Global). That risk, to us, is most prevalent if humanoids are ultimately a hardware story.

The bull case for U.S. companies rests on the fact that the “body” is currently ahead of the “brain” and the latter is where the US excels. Reliable autonomy, dexterity and proven safety remain under development. In addition, policy may protect US manufacturers, such as the proposed Humanoid ROBOT Act and the American Security Robotics Act. A potential outcome, in our view, is a bifurcated market, with China dominating cost and volume and the U.S. leading in software and autonomy, which we believe to ultimately be the larger prize.

What are the risks?

Our bottom up framework above and the Street estimates converge on the same conclusion: if humanoids remain simply a factory automation niche, the sector is likely worth on the order of $500+bn. However, if they become a genuine substitute for human labor, it becomes one of the largest equity complexes in the world.

In addition, nosebleed growth trajectories assume continued technology improvement, cost declines and reliability gains, none of which are guaranteed.