If a company remains stuck in the third tier of the embodied AI sector, its days are numbered.

This conclusion stems from a layered analysis of the sector’s evolving dynamics and the shifting logic of capital allocation. By early 2025, the embodied AI landscape had already begun to take shape, with leading players emerging and a preliminary competitive framework established. The Embodied AI Society and several frontline investors deeply involved in hard technology posed a critical question: Can true dark horses still emerge in this sector? And if so, what should their founders look like?

At that time, investors from various institutions largely converged on a single judgment: Dark horses will inevitably emerge, most likely from within the current head embodied AI companies—those core executives with hands-on experience in technology deployment, supply chain integration, and scenario expansion, backed by significant capital.

Unlike the 2023 trend of betting on “scientists,” if capital were to adopt a “people-first” approach again, the ideal candidate would be someone who has personally overseen the technology iteration from 0 to 1, navigated the challenges of large-scale deployment, and deeply understands industrial boundaries and market demands.

A year later, this prediction has been strongly validated. Beneath the surface of the embodied AI sector, a new wave of entrepreneurial activity driven by executive leadership is surging. According to the Embodied AI Society, some investors are even actively encouraging core talents from leading embodied AI companies to launch their own ventures, offering highly attractive starting packages.

However, new contradictions have also emerged sharply: The capital pie is, after all, limited.

By 2026, the concentration of financing in the sector has become even more pronounced. This new wave of startups, founded by executive entrepreneurs, has become highly sought after by investors due to their stellar teams and industrial resources. On one side is the “siphon effect” of the leading players, and on the other is the “squeezing posture” of the new forces.

Caught in the middle, third-tier companies lack both the technological and capital barriers of the leaders and the industrial resources and flexible strategies of the new forces. Their survival space is being continuously compressed.

This is not to deny the necessity of current third-tier players, but the reality is stark: If they continue to linger in the third tier, maintaining a clear gap with the second tier in key areas such as technological breakthroughs, commercialization, and industry influence, and fail to find a differentiated path to break through, they may miss out on the historic dividends of the embodied AI industry.

Looking back, the heated discussions about “embodied AI opportunities” at the beginning of 2025 now seem almost unbelievable, given the explosive growth and substantive deployment that followed.

This is not merely a pessimistic narrative or a misjudgment from a narrow perspective but rather a cautious probing, even a self-questioning.

At the beginning of 2025, embodied AI, previously hidden behind the industrial curtain, was suddenly thrust into the spotlight due to a national-level debut, instantly being labeled as the “next tech revolution” and placed in a context of “moral imperative.” The slogan of the “embodied AI era” resonated throughout the industry. But beneath the hype, the sector’s foundations were just being laid: Technologically, core bottlenecks such as insufficient generalization, the embodied gap, and the difficulty of Sim2Real deployment remained unbroken, with most products still stuck in the “specific scenario demonstration” stage. Commercially, a stable profit model was lacking, and the widespread dilemma of “holding a hammer and looking for a nail” persisted, with truly scalable scenarios that could be deployed being few and far between.

The massive influx of traffic and capital in a short period did not necessarily nourish the industry to maturity but instead left most practitioners deeply worried. It felt more like an irrational “hype to death.”

When a batch of startups with immature technologies and vague business logics, seemingly “ready to fall apart after a few steps without wind,” appeared alongside massive funds ranging from hundreds of millions to billions, this severely imbalanced ecosystem inevitably raised questions: Was this just another standard act of “blowing bubbles” in the tech industry?

So at that time, let alone “opportunity theories,” there was even a wave of discussion about “retreating.”

A year on, the “opportunity theory” of embodied AI has undergone a complete evolution.

The sector, once seen as “overhyped,” has slowly emerged with verifiable deployment scenarios under the leadership of top players. Business models once questioned as “castles in the air” have begun to see large-scale deployment.

Capital, which once chased concepts indiscriminately, has also become more restrained, shifting toward hardcore assessments of technology, supply chain, and commercialization capabilities. Investors no longer just look at technical parameters on PPTs but demand live demonstrations of practical scenarios like “opening a mineral water bottle.” It’s no longer about “profiting by riding the wave” but “carefully selecting high-quality players for long-term success.” It’s no longer about “whoever rides the concept gets a piece of the pie” but “only a few who truly solve industry pain points and possess core barriers will make it to the end.”

Nowadays, few investors or industry insiders publicly talk about an embodied AI bubble anymore. Instead, more voices are calling for “patience.” This long-slope, snow-covered track finally seems to have found a more appropriate pace of development.

From the current stage, companies with more cutting-edge technologies and engineering capabilities have already formed a stable landscape, and it’s not a simple matter of “buying a ticket to board the train.” Moreover, this “ticket” has already exceeded the affordability range of most investors. In comparison, the second tier still maintains high growth potential and relatively friendly “ticket prices,” with most investors still believing in and waiting for a breakthrough from the second tier.

However, third-tier companies are finding it increasingly difficult to be seen or chosen. An investor once told the Embodied AI Society that they would rather wait for new players with differentiated characteristics to emerge.

Thus, it is evident that the opportunities in embodied AI have completely flowed to those more worthy of being bet on.

If the survival space for the third tier is being extremely compressed, a pressing reality must be asked: Who exactly is taking away the pie that originally belonged to them?

The answer is not outsiders appearing out of nowhere but a batch of new players naturally differentiated as the industry enters the “deep waters.” Unlike the evangelists in 2023 who often peddled the “ultimate form of general artificial intelligence,” the new entrants who can secure “new tickets” from investors in 2026 have extremely specific and pragmatic profiles.

The biggest opportunity lies with the executives of embodied AI companies. They represent the strongest force at present. As mentioned earlier, the “people-first” investment criteria of frontline investors have changed. They no longer just look at academic credentials but also at “scars.” Core executives who have navigated pitfalls and personally overseen mass production and delivery in top embodied AI companies are becoming hot properties for top-tier VCs.

Why are these individuals so popular?

Because they understand what “won’t work.” Over the past two years, these former tech experts, hardware or supply chain leaders, have experienced firsthand the ninety-eight difficulties from PPT blueprints to hand-assembled engineering prototypes and then to small-batch trial production. They know which upstream suppliers have extremely poor yield rates and understand which scenarios are purely “pseudo-demands” at this stage.

When these executives gather their old teams to start their own ventures, they carry no historical baggage. Moreover, they are unlikely to pursue stories like “all-powerful general-purpose humanoid robots” again. Instead, they will seek a more comfortable and pragmatic ecological niche within the embodied AI industry.

For capital, the “trust cost” of such teams is extremely low: Investors don’t need to question their technical capabilities (backed by deployment cases from top companies), worry about their supply chain resources (with mature partnerships), or doubt their commercialization abilities (with customer resources as a foundation).

Therefore, many executive entrepreneurship projects can secure hundred-million-yuan-level angel round financing with just a “team background + preliminary business plan,” while ordinary startups might need to develop mature products first to secure small seed round investments. This “head start” under capital support further amplifies their opportunities.

If executive entrepreneurship can be seen as a differentiation within the same lineage, then cross-industry players from major firms are completely upending the table.

Embodied AI is essentially about intelligence, which happens to be the logic that new energy vehicle companies and internet giants are playing. In the past year, a large number of engineering veterans with automotive backgrounds have entered the sector. These veterans’ greatest strength lies in their absolute control over the supply chain. On the other hand, internet giants are no longer content with just making peripheral investments and have begun to deeply deploy in embodied AI, not ruling out the possibility of establishing separate subsidiaries in the future. If they do, they will be a force to be reckoned with.

The last group of unignorable new forces is the “embodied-native” new generation of elite geeks from prestigious universities.

The current top tier of embodied AI includes many founders with prestigious university backgrounds or scientists with significant academic achievements. Essentially, these scholars are the “wave-makers” of embodied AI, but what they learned academically was still “robot learning” rather than embodied AI.

However, the current academic representatives in universities conduct research with “embodied AI” as the top-level architecture, belonging to “embodied-native” scientists.

This reserve force aligns more closely with the narrative of embodied AI and embodies the style of the new era, favoring a “recognize reality, abandon gaming” approach. They prefer to take a small-and-beautiful product route. For example, focusing on consumer-end elderly care robots, they use precise scenario positioning to compensate for shortcomings in supply chain and engineering capabilities. Although their mass production capabilities still need refinement, with their cutting-edge algorithmic advantages and precise scenario entry, they are also diverting opportunities in some niche areas, becoming another “pie-snatcher” for the third tier.

By placing these three profiles side by side, we can see the true rules of the game in the current embodied AI sector: Resources are flowing more rapidly in a specific direction.

This also explains why it is said that there is not much time left for companies that remain stuck in the third tier forever.

If a company neither has the mine-clearing experience of executive entrepreneurship teams nor possesses the supply chain integration capabilities of cross-industry veterans, and if it still clings to the obsession of “building a big robot” and refuses to engage in profitable activities, then under the dual pressure of the leaders’ siphon effect and the new forces’ squeezing posture, this kind of half-hearted persistence will eventually exhaust the last cash on its accounts in a series of unnoticed exhibitions.