Bernstein: Humanoid Robot Integrators Are a Better Bet Than Pure Component Suppliers
Alina Collins
Bernstein's new humanoid-robot industry report argues that OEMs deserve a higher structural bet than pure component suppliers — because OEMs can stack multiple differentiation roles, while component makers are often squeezed into a single lane of engineering excellence.
Why is this industry "easy to enter, but with a very high ceiling"?
Roughly 150 companies have piled into the humanoid-robot space, which itself proves the barrier to entry is low.
But Bernstein argues the real determinant of returns is not entry — it is how many high-value roles a single company can play at once.
The report defines four roles: Scientist (cracking the robot brain), Inventor (defining the whole-machine form), Engineer (refining components), and Gardener (running the data and skills ecosystem).
This means → the crowd is large, but most players fill only one or two roles; companies that wear multiple hats hold the structural edge.
How far along is the "robot brain"?
The technical paradigm is iterating fast: from large language models (LLMs) to vision-language-action models (VLAs — models that see, understand, then act), and now toward various "world action models" (WAMs).
Bernstein's read: "Light at the end of the tunnel, but the tunnel is still long."
In plain terms = brain tech is advancing, but not mature enough for a robust valuation on pure "brain" concept plays — which is why the report steers away from betting on brain alone right now.
Why bet on OEMs over component makers?
OEMs can freely combine invention, research, engineering, and ecosystem roles to build a moat; their degrees of differentiation freedom are far greater.
Pure component makers are often forced onto a single narrow lane — engineering excellence — and once that edge is matched, the moat disappears.
Many OEMs have already integrated brain-model development upward, turning their product into a "cerebellum + brain" hardware-software package with stronger pricing power and customer lock-in.
This means → Bernstein's screening rule is explicit: the more differentiation roles a company stacks, and the closer it sits to the ecosystem center, the greater its structural return potential.
The best humanoid plays aren't listed — so what now?
The most-watched humanoid-robot companies — China's Unitree, and the US-based Physical Intelligence and Figure AI — are all still private.
Bernstein therefore projects the same framework onto industrial automation and other "adjacent rooms," hunting for proxy picks with proven business models.
FANUC is positioned as the "Inventor" of industrial robotics — and is already "tending the garden" alongside Nvidia and Google, combining OEM and ecosystem identities.
Keyence is a "pinnacle-level Inventor," with over 6,000 products and roughly 20% of revenue each year from new launches.
Six picks — how does Bernstein value them?
The valuation anchor across all picks is EV/EBITDA — a ratio of enterprise value to pre-tax, pre-depreciation profit; a higher multiple means the market assigns a larger premium. DCF serves as a long-term cross-check; target multiples "reference historical cycles, adjusted for secular or competitive trends."
Three Japanese names, all rated Outperform: FANUC at JPY 7,000 (~22.5× EV/EBITDA); Harmonic Drive at JPY 7,800 (~45.5×, earning the highest premium on ~50% global share in strain-wave reducers); Keyence at JPY 86,000 (~21.5×).
US-listed Cognex is rated Outperform at USD 75.00 (~36.5×); the key variable is Moritex integration and new-customer pipeline progress.
Two Chinese names form a contrast: Inovance (汇川技术) is rated Outperform at CNY 82.00 (~24.0×); Estun (埃斯顿) is the only Market-Perform in the group, target CNY 26.00 (below current price), ~35.9×, with the Cloos acquisition synergy as the make-or-break variable.
What is the biggest risk across these picks?
Bernstein flags clearly: FANUC and Keyence carry significant cyclical exposure — the report even notes that in a severe downturn similar to 2009, Keyence revenue could decline.
Harmonic Drive's risk is more specific: a ~50% global share is both a moat and a vulnerability — any shift in the competitive landscape hits harder than it would for peers.
Estun's issue is post-merger integration: whether Cloos synergies materialize directly determines if the current lofty valuation (~116.5× forward P/E) can hold.
Content is for reference only, not financial advice.