Capital is flooding into physical AI with a sharp thesis: generalist robots and AI-driven engineering tools will capture more value than narrow automation. The premise holds that manufacturers facing labor shortages cannot afford to wait for humanoid robots to mature, and that task-specific machines no longer justify their cost when production lines constantly shift. Barcelona's Theker just raised $85 million in what the company calls Europe's largest robotics Series A, led by CRV with backing from Samsung and Bernard Arnault's Aglaé Ventures. The startup builds reconfigurable factory robots whose hands, arms, and form factor can be swapped depending on whether they're sorting packages, packing clothing, or handling bottles. Inditex, Zara's parent, signed on early.
Co-founder Carla Gómez Cano framed the opportunity plainly: most processes do not involve putting the same cookie in the same box, and existing automation breaks down when variability rises.
The ambition is sweeping: replace large swaths of engineering work with AI that automates design and manufacturing of complex physical systems.
The same logic is drawing strategic capital. Samsung invested in Theker and is in advanced discussions to become a customer and supplier simultaneously, a trifecta that would give the startup revenue, credibility, and supply-chain integration in one move. Theker's near-term focus is retail and logistics, but the company's broader ambition is heavier industrial manufacturing, where manual task complexity is higher and the cost of downtime steeper. The funding wave follows a record seed round less than a year prior, underscoring how quickly the generalist factory-robot thesis has gained traction among investors betting that configurability beats specialization when labor is scarce and production demands are heterogeneous.
At the far end of ambition sits Prometheus, the physical AI startup co-founded by Jeff Bezos and former Verily co-founder Vik Bajaj. The company announced a $12 billion raise at a $41 billion valuation, bringing total funding to $18.2 billion after an initial $6.2 billion round late last year. Backers include Bezos himself, JPMorgan Chase, Goldman Sachs, and BlackRock. Prometheus is building what it calls an artificial general engineer: software capable of automating the design and manufacturing of complex physical systems, from jet engines to drug compounds. The 150-person company, with offices in San Francisco, London, and Zurich, has not disclosed what it has built, but Bezos told CNBC that a large portion of capital will fund compute infrastructure to support the engineering models.
Bezos is also staking out a contrarian labor thesis. While many AI leaders predict widespread job losses, he argues that productivity gains will create labor scarcity, his term for a world where demand for human workers outpaces supply. He suggested that two-earner households may become one-earner households and that overtime work will decline as living standards rise. The view is shaped by experience: Amazon, where Bezos serves as executive chairman and largest individual shareholder, employs over 1.5 million people and has laid off tens of thousands over the past year while accelerating automation under CEO Andy Jassy. At $41 billion, Prometheus is among the most valuable private companies building physical AI, a category that barely existed as an investor focus two years ago.
The infrastructure layer is also attracting capital. Physical Intelligence, co-founded by Karol Hausman alongside researchers Sergey Levine and Chelsea Finn, has raised more than $1 billion to develop foundation models that serve as a general-purpose AI brain for robots. The company's thesis mirrors the scaling dynamics that transformed language models: instead of programming robots for a single task, build models that generalize across machines, environments, and use cases. Hausman, who worked at Google Brain and Stanford before founding Physical Intelligence, has argued that the central challenge in robotics is not mechanical design but intelligence—how robots learn, generalize, and interact with a physical world far harder to simulate than describe. The bet is that resisting commercial pressure toward specialization will unlock the same exponential gains seen in language AI.
The May unicorn crop reflects how quickly capital is rotating from pure software into embodied intelligence. AI services and robotics led the 29 companies that joined the Crunchbase Unicorn Board last month, a shift toward businesses helping enterprises deploy AI in the real world. OpenAI and Anthropic each launched multibillion-dollar deployment ventures staffed with forward-deployed engineers, while a long list of startups building AI infrastructure, autonomous software, and robotics reached unicorn status. China dominated the robotics cohort; Canada led in quantum. The Unicorn Board reached $9.9 trillion in value during May, though that figure is set to drop sharply as SpaceX, Anthropic, and OpenAI move toward public listings. Anthropic filed confidentially for an IPO after surpassing OpenAI to become the second most valuable private company behind SpaceX.
The investment wave is driven by two converging realities. First, manufacturers are showing growing interest in startups that promise faster automation without the usual tradeoffs, as labor shortages make the status quo untenable. Humanoids are not yet ready to replace factory workers, and the industry cannot wait. Second, VCs are betting that generalist physical AI will capture more value than narrow tools because configurability and intelligence compound in ways that task-specific hardware does not. The thesis holds that once a foundation model can operate across environments, each new deployment generates data that improves the model for every other use case, creating a flywheel that specialized robots cannot replicate.
The question is whether scaling laws that worked for language will transfer to the physical world, where reliability requirements are higher and real-world data is harder to collect.
The capital commitments are forcing a reckoning around business models and timelines. Prometheus is spending heavily on compute, a strategy that echoes the playbook of large language model developers but with unclear paths to revenue given the long development cycles in physical engineering. Theker is starting with retail and logistics, sectors where task variability is high but consequences of failure are lower than in heavy manufacturing. Physical Intelligence is focused on building the brain before the body, a risky approach that requires convincing hardware partners to integrate its models before proving commercial traction. All three are burning capital faster than revenue can scale, a dynamic that historically has ended poorly for robotics companies but that investors now view as a necessary cost of building general systems.
The bet is that general beats specialized, even if it costs billions to find out.
Defense and maritime applications are also drawing attention. Saronic, a defense maritime robotics company, appeared alongside Theker and others in recent optimism roundups, signaling that generalist physical AI is not limited to commercial manufacturing. The defense sector offers higher margins, longer procurement cycles, and customers willing to pay for reliability, making it an attractive parallel track for companies building adaptable autonomous systems. Standard Bots, another generalist factory robot startup, has appeared in multiple funding roundups, though specifics on its raise remain undisclosed. The breadth of capital flowing into this category—from retail automation to defense—suggests that investors believe the generalist thesis applies across domains, and that whoever builds the most adaptable intelligence layer will capture outsized value as physical automation scales.