AI Unicorns Surge as May Adds 29, Led by Deployment and Robotics

A month defined by putting AI to work

May 2026 did not produce a breakthrough model. It produced 29 new billion-dollar companies, and most of them are in the business of getting artificial intelligence to work inside real organizations. Crunchbase counted 29 entrants to its Unicorn Board for the month and found that the clearest pattern was companies that help enterprises deploy AI, along with a group of robotics startups that tie software intelligence to physical machines.11 Crunchbase reads the cohort as a sign of where investors now expect returns: turning model advances into working applications and pairing AI with automation.11

Two things happened at the same time, which is what makes May unusual. The flow of new unicorns kept up its record pace, while the three biggest private companies in the world started moving toward public markets. Taken together, those events say a lot about the AI capital cycle four months later.

The labs become consultancies

The two largest new entrants were spinoffs from the frontier labs themselves. OpenAI Deployment Company raised $4 billion in a private equity round led by TPG, with Advent International, Bain Capital and Brookfield as co-leads, at a $14 billion valuation.11 OpenAI holds a majority stake. It also bought Tomoro, gaining about 150 forward-deployed engineers to work directly with enterprise customers, and says the new money will go toward growth and further acquisitions.11 Anthropic's version, the Anthropic Applied AI joint venture, raised $1.5 billion to build a services business that integrates Claude into client operations. Anthropic, Blackstone and Hellman & Friedman each put in $300 million, and Goldman Sachs and Sequoia Capital also invested.11

This matters because it is a strategic shift. The two companies that set the pace in foundation models are now using private equity money to build something close to a consulting and systems-integration operation. That suggests selling API access alone has not been enough to drive deep enterprise adoption. Customers apparently need engineers on site, and both labs have decided to own that work rather than leave it to the traditional consulting firms. Before the end of summer, OpenAI's deployment arm was the third most valuable new unicorn of the first half, behind only DeepSeek and OKX.14

Below the two giants, May's AI software cohort is made up of tooling for agents. Exa, which builds search for AI agents, raised a $250 million Series C led by Andreessen Horowitz.11 Blitzy raised a $200 million Series A from Northzone at a $1.4 billion valuation for software that maps legacy codebases and runs multi-week autonomous engineering projects.11 OpenRouter, which sends application traffic to more than 400 models, raised $113 million led by CapitalG, with corporate venture arms from Nvidia, ServiceNow, MongoDB, Snowflake and Databricks also in the round, at a $1.3 billion valuation.11 Socket, a security product for open-source packages in automated coding workflows, reached $1 billion after a $60 million Series C led by Thrive Capital. It counts Anthropic, xAI, Cursor and Replit among its users.11

The data trackers report slightly different figures for some of these deals. Crunchbase puts Exa at $2.2 billion, while TechCrunch, citing PitchBook, lists it at $1.95 billion.1115 Crunchbase also records defense parts maker Amca at $1 billion, against PitchBook's $1.1 billion.1115 The gaps are small and don't change the overall picture. They are still a useful reminder that a "valuation" in this market often depends on who is counting and on what terms.

Robotics, and China's growing share

The robotics entries show how much early-stage hardware rounds have grown. Hark, a new company from Figure founder Brett Adcock, raised a $700 million Series A led by Parkway Venture Capital at a $6 billion valuation. It plans to build its own models, training stack and hardware for personalized robots.11 The other three robotics unicorns are Chinese. Tianji, a dual-arm robot maker, raised $147 million from Hillhouse and Meituan. Agilink, a dexterous-hand spinout from Agibot, reached $1 billion after four rounds since January. Botshare, a robot rental platform, did the same.11

This is not specific to May. Crunchbase reported that robotics and frontier labs also led April's count, when six humanoid companies, five of them Chinese, passed the billion-dollar mark.13 Across the first half, China minted 38 new unicorns, up sharply from 10 in all of 2025.14 Crunchbase's sector data describes record robotics funding driven by investor interest in embodied AI, a field venture firms once treated as a capital-hungry hardware bet.17 Acquisitions point the same way. Meta bought San Diego-based Assured Robot Intelligence in May and moved the team into its Superintelligence Labs, and Skild AI bought Zebra Technologies' robotics unit in April.17

Not only AI, but mostly AI

May was more varied than some recent months. New unicorns came from healthcare, quantum, aerospace, manufacturing, e-commerce and energy. Seventeen were U.S.-based, four each came from China and the U.K., two from Canada, and one each from India and Brazil.11 Even so, AI shapes much of the non-AI list. Cowboy Space raised $305 million at a $2 billion valuation to launch orbital data centers. Armada is building modular data centers for military and remote sites. Vi Labs sells healthcare AI agents built on clinical data.11 The largest non-AI deal was MiRus's $1.5 billion corporate round from Boston Scientific, which also gets an option to acquire the company's heart valve technology.11

The longer trend backs up the concentration argument. By early August, 195 companies had joined the board in the first half, already more than the 193 minted in all of 2025. Robotics and new AI labs led the count.14 One analysis argued that frontier labs and physical AI are pushing other sectors out, rather than simply leading in a few individual months.18 That reading looks correct. The non-AI unicorns in May are real, but more and more of them describe themselves in terms of AI infrastructure.

The exits: May's promise, October's reality

In May, the Unicorn Board reached $9.9 trillion in total value, and Anthropic passed OpenAI to become the second most valuable private company after SpaceX.11 Anthropic's late-May round raised $65 billion at a $965 billion post-money valuation. It said on June 1 that it had confidentially filed for a U.S. IPO, ahead of OpenAI, which most prediction-market traders had expected to file first.4 OpenAI announced its own confidential filing on June 8, without setting a timeline.10 D.A. Davidson's Gil Luria described the two labs as racing to list before Wall Street's capacity ran out. In his view, the combined capital needs of SpaceX, OpenAI and Anthropic could disrupt markets, so whoever went first would have an advantage.4

Forecasts in the spring were very optimistic. Several outlets described a sequence of SpaceX in June, OpenAI in September and Anthropic in October, with a combined value of about $3.6 trillion.9 Reporting at the time disagreed on how ready OpenAI was. CFO Sarah Friar reportedly warned colleagues that the company might not be ready for a 2026 listing if compute spending kept growing faster than revenue.2

The cautious view has turned out to be right so far. SpaceX priced 555,555,555 shares at $135 and raised about $75 billion, the largest IPO on record.5 Its trading since then has been bumpy: shares closed at $154.60 on June 22, down 31.5% from an intraday peak of $225.64.5 In September, Sam Altman told Fortune that OpenAI would not go public in 2026 and said the current safety climate made it a bad moment to list.5 Anthropic is now the only one of the two labs still reported to be aiming for a listing this year, although that timing comes from press reports rather than the company.5

Cerebras offered an earlier and more encouraging signal. Its May IPO valued the chipmaker at $56.4 billion, compared with a $23 billion private valuation only three months before.11

What it adds up to

The May cohort is best read as the start of a second phase. The model-building race is consolidating into a few very large companies, and those companies are now heading for public markets on uneven schedules. Below them, venture money is going into the deployment layer: agent tooling, routing, code automation, security and, increasingly, robots. The labs moving into services with private equity partners is the clearest sign that the money is shifting from inventing AI to installing it.

The risk is that the exit timeline is what supports these private valuations. SpaceX's post-IPO slide and OpenAI's delay show that public markets will not automatically accept trillion-dollar private marks. If Anthropic lists this fall and holds its valuation, the deployment and robotics unicorns created in May will look early and smart. If it doesn't, many companies valued on assumptions about future public-market appetite may face a harder second round than their first.