The burgeoning artificial intelligence landscape has witnessed an unprecedented surge in mergers and acquisitions throughout the past year, a trend so pervasive that many significant deals now scarcely register. Both AI titans, Anthropic and OpenAI, have been actively engaged in strategic acquisitions, targeting developer tooling companies, AI service providers, and product testing startups. Their objective is clear: to rapidly translate cutting-edge model capabilities into tangible enterprise revenue and expand their market dominance. It was against this backdrop of high-stakes M&A activity that a weekend rumor regarding Anthropic’s potential acquisition of the robotics startup Physical Intelligence garnered significant, and rapid, attention, even in the face of an initial denial from Physical Intelligence’s CEO.
The intensity of the rumor’s spread can be attributed, in large part, to the prominent figures involved and the impressive trajectory of Physical Intelligence. This is not a nascent or obscure robotics venture. Physical Intelligence was co-founded by Lachy Groom, an investor-operator whose influence and standing in Silicon Valley have seen a marked ascent in recent years. The startup had already secured over $1 billion in funding and was reportedly in discussions for an additional $1 billion funding round in the spring, aiming for an $11 billion valuation. Furthermore, its proprietary "robot brain" model is recognized as one of the more widely adopted in robotics research, suggesting a foundational technology with significant potential.
As it has since transpired, the rumor, while initially denied, was not entirely without basis. Subsequent reporting from The Information revealed that Anthropic and Physical Intelligence did indeed engage in acquisition talks during the spring. This development lends credence to the initial report by tech blogger Robert Scoble on X, which, despite potentially mischaracterizing the specifics, accurately indicated that a significant interaction between the two entities had occurred. The speed at which the rumor propagated underscores the intense scrutiny and anticipation surrounding the strategic moves of major AI players as they prepare for their highly anticipated public debuts.
Physical Intelligence’s initial response to the widespread speculation was, by all accounts, less than a robust rebuttal. According to The Information, CEO Karol Hausman informed employees of the reports’ inaccuracy via a Slack message, reportedly accompanied by a GIF of a character from the popular television series "The Office" shaking their head in negation. Lachy Groom, a co-founder of Physical Intelligence, did not respond to inquiries for comment from TechCrunch on Monday evening.
The M&A activity surrounding AI leaders is reaching new heights. Anthropic has completed four known acquisitions this year, while OpenAI has been even more aggressive, acquiring at least 17 companies since the beginning of 2023. These acquisitions are occurring as both companies are actively preparing for their initial public offerings (IPOs). Anthropic confidentially filed for its IPO on June 1st, with OpenAI following suit a week later. These impending public offerings are poised to be among the most significant stock market debuts in U.S. history, placing immense pressure on both companies to demonstrate strong growth and market leadership.
The strategic rationale behind an AI company’s interest in a robotics startup like Physical Intelligence, particularly at this juncture, points towards a critical frontier in AI development: the integration of advanced intelligence with the physical world. The prevailing hypothesis is that achieving true superintelligence may necessitate a profound understanding of physical environments, a capability that cannot be fully replicated through text-based data alone. This is a significant departure from the early days of AI development, which often focused on abstract reasoning and data processing.
OpenAI’s own historical engagement with robotics provides a compelling precedent for this strategic pivot. The company demonstrated early prowess in robotics with a robotic hand capable of solving a Rubik’s Cube. However, this dedicated robotics division was subsequently disbanded in 2021. Co-founder Wojciech Zaremba later articulated that the approach at the time was missing crucial elements required for the development of true superintelligence. The team was later reconstituted in 2024, quietly establishing a humanoid robotics lab in San Francisco. This initiative was officially announced by CEO Sam Altman in late May, marking the return of "OpenAI Robotics." The company has outlined a near-term focus on robots for infrastructure tasks, with a long-term vision of developing a personal robot for every individual.
Anthropic, while not having publicly showcased hardware development on the scale of OpenAI’s lab, has consistently emphasized safety and frontier capability testing through its internal research groups. A notable example is "Project Fetch," initiated in November, where Anthropic staff explored how its Claude AI model could assist non-experts in programming a robot dog. A subsequent phase of this project in June revealed that a newer iteration of the model could complete the same tasks approximately 20 times faster than the previous year’s best human-plus-Claude team. This research indicates a growing focus on bridging the gap between AI’s cognitive abilities and physical execution.
Acquiring an established robotics team with deep expertise, such as that possessed by Physical Intelligence, would allow Anthropic to significantly accelerate its progress, potentially bypassing years of internal development and research. However, this potential acquisition is complicated by the intricate investment landscape surrounding Physical Intelligence. Founded approximately two years ago in San Francisco by Lachy Groom, former Google researchers, and professors from Stanford and Berkeley, Physical Intelligence’s early investor base mirrors that of OpenAI. Key investors include Khosla Ventures and Thrive Capital, both significant backers of OpenAI. Furthermore, Founders Fund, another major OpenAI investor, was reportedly involved in Physical Intelligence’s most recent funding round earlier this year.
The most significant complication arises from the fact that OpenAI itself is an investor in Physical Intelligence. This makes OpenAI not merely an observer but a stakeholder in a company that its primary competitor, Anthropic, was reportedly in recent acquisition talks with. This situation raises critical questions about the nature of OpenAI’s initial investment. It is plausible that such early-stage strategic investments, particularly in a competitive field, may include protective clauses such as information rights or a right of first refusal on sales to competitors. These provisions are designed to safeguard a strategic investor’s interests in precisely such scenarios.
Consequently, if Physical Intelligence is indeed a company currently exploring sale options, OpenAI, already a shareholder, a known associate of Groom, and actively pursuing its own robotics ambitions to surpass Anthropic, may possess a more compelling claim to the company than Anthropic. When approached for comment on these specific questions regarding their investment and potential rights concerning Physical Intelligence, OpenAI did not respond.
The broader implications of these potential acquisitions extend beyond the immediate competitive dynamics between Anthropic and OpenAI. They signal a critical inflection point in AI development, where the focus is shifting from purely digital intelligence to embodied intelligence. The ability of AI systems to interact with and manipulate the physical world is increasingly seen as a necessary precursor to achieving more sophisticated and generalized artificial intelligence. This necessitates not only advanced algorithms but also robust hardware integration and a deep understanding of physical constraints and dynamics.
The race to acquire or develop in-house robotics capabilities underscores the understanding that while vast amounts of data can train AI models for complex tasks, the nuanced understanding of physical reality – gravity, friction, object permanence, and delicate manipulation – remains a distinct and challenging frontier. Companies that can effectively bridge the gap between digital intelligence and physical embodiment are likely to hold a significant advantage in the development of future AI systems, whether for industrial automation, personal assistance, or more advanced applications. The coming months will likely reveal the full extent of these strategic maneuvers and their impact on the trajectory of AI innovation.
About the Author
