Forty Unicorns in Thirty-One Days

Forty companies joined the Crunchbase Unicorn Board in July 2026 — the highest monthly total in more than four years. Three entered above $10 billion.

Crypto.com, Kling AI, and Ant International together added $49 billion in valuation over the month.

By geography: 19 of the new unicorns are American, just under half. China contributed eight, the U.K. three, Singapore two. Leading sectors by count were financial services, robotics, AI orchestration, multimodal AI, energy, and semiconductors.

The sector list is the interesting part

A unicorn count is a crude instrument — it measures the number of times a private round cleared a round-number threshold, which is as much a function of how funds are structured as of how good the companies are.

The sector composition is less crude, because it shows where the capital chose to go when it had options.

Financial services leading is a return, not a novelty; fintech dominated the last unicorn boom and has spent three years being repriced downward. Its reappearance at the top suggests the reset has completed and a new cohort is being funded on different assumptions.

Robotics and semiconductors at the top is the genuinely new information. Both are capital-intensive, hardware-dependent categories with long development cycles and physical supply chains — precisely the profile venture capital spent a decade avoiding in favor of software margins. Their presence indicates money is flowing toward businesses that require factories.

AI orchestration appearing as a distinct category from multimodal AI is the subtler signal. Orchestration is the layer that runs agents, routes between models, manages context, and controls cost. That it is now a separate line item in a unicorn count means investors have concluded the layer above the model is a defensible business rather than a thin wrapper — the same thesis a company like Writer is arguing when it markets a harness rather than a model.

Four years is a specific comparison

"Highest in more than four years" places the prior peak in early 2022 — the top of the zero-rate venture cycle, immediately before the repricing that cut private valuations, froze exits, and stranded a generation of companies at marks they could not grow into.

The comparison invites an obvious concern. Unicorn creation peaked, then the market broke. A count matching that peak looks like the same conditions returning.

But the mechanics differ in ways that matter. The 2021–22 cohort was minted largely by crossover funds and non-traditional investors deploying into late-stage software at revenue multiples that assumed permanently cheap capital. This cohort is being minted in a higher-rate environment, concentrated in categories with physical constraints and government demand, by investors who watched the last correction happen.

That does not make the valuations right. It makes them differently wrong if they are wrong — overpaying for hard technology with real capital requirements produces a different failure mode than overpaying for software with negative unit economics. The first burns money slowly and leaves assets. The second burns money quickly and leaves nothing.

The geographic split

19 U.S., eight China, three U.K., two Singapore.

The U.S. share — just under half — is lower than the American venture industry's self-image usually assumes, and China's eight is notable given the persistent narrative that its startup funding has not recovered. Eight new billion-dollar companies in a month is not a dormant ecosystem.

Kling AI appearing among the largest additions is instructive on that point. It is a Chinese video generation company that reached its valuation on the strength of a product with global users, not on a domestic-market thesis. The capital came from Chinese sources — Alibaba and Tencent among them — but the demand did not.

The Singapore entries reflect what has become a durable pattern: capital and corporate structure routing through a neutral jurisdiction while operations sit elsewhere. Two in a month is a small number that indicates a stable channel rather than a surge.

What a unicorn count does not tell you

It does not tell you the terms.

A $1 billion valuation with a 1x liquidation preference and no participation is a genuine mark. The same headline number with structure — participating preferred, ratchets, guaranteed multiples on exit — is a debt instrument wearing an equity valuation, and the headline is close to meaningless.

The 2021 cohort taught this lesson expensively. Companies that announced billion-dollar rounds discovered at exit that the preference stack consumed most of the proceeds, and common holders — founders and employees — received a small fraction of what the valuation implied.

Crunchbase counts valuations, because valuations are what get disclosed. Structure rarely is. Forty companies crossed a threshold in July; how many of those marks would survive a down-round test is not in the data and will not be for years.

The read

The count is real and the composition is encouraging. Robotics, semiconductors, and energy leading a unicorn cohort means venture capital is funding physical infrastructure again, which the last cycle largely refused to do and which the AI buildout badly needs.

The four-year comparison is the caution rather than the story. Peak unicorn creation has historically been a late-cycle indicator, not an early one — the count peaks when capital is most abundant and least discriminating, which is also when the marks are least reliable.

The honest position is that both are true simultaneously. Forty companies crossed a billion dollars in a month because there is genuine demand for what several of them build, and because there is more capital chasing that demand than there are companies worth funding at these prices. Which factor dominated will be visible in about three years, in the exit data, and not before.