Analysis
Twenty-nine companies crossed billion-dollar valuations in August, adding roughly $63 billion in combined value to the private markets, according to Crunchbase News. AI software led the pack, followed by semiconductors with five new unicorns, and robotics and financial services tied with three each; data centers, security and energy each added two.
The most heavily valued new entrants show where the money concentrated:
- XPeng Robotics (China) -- $6.3B valuation on its first outside funding round of $900M+, for its IRON humanoid robot program.
- Lumilens (San Jose) -- $5.5B valuation on a $700M Series C for photonic AI infrastructure.
- River AI (Palo Alto) -- $5B valuation on $1.1B raised for a custom AI model training platform.
- Source Foundry (San Francisco) -- $5B valuation on $400M raised for semiconductor manufacturing tools.
- Twenty Technologies (security) -- $100M Series B that pushed it past the $1B mark on August 4 (Pulse).
“- Lumilens (San Jose) -- $5.5B valuation on a $700M Series C for photonic AI infrastructure.”
Where the growth is concentrated
The US accounted for 16 of the 29 new unicorns, more than half, with China adding four and South Korea, India, Singapore, the UAE, Switzerland, Germany, Turkey, Nigeria and Indonesia each contributing one. More than a third of the new entrants are under three years old, underscoring how fast today's best-funded startups reach ten-figure valuations compared with the multi-year climbs typical of the 2010s unicorn class.
The board isn't just growing, it's also churning. Nine companies exited in August: three via IPO, including humanoid-robotics maker Unitree Robotics, whose Shanghai debut popped 460% and pushed its market value to roughly $50 billion; and six via acquisition, including Hugging Face and Airtable. That churn rate -- roughly one exit for every three new entrants -- is a healthier signal for the venture ecosystem than pure unicorn-minting alone, since exits are what actually return capital to LPs rather than mark up paper valuations.
The AI concentration risk
The concentration risk in August's cohort is the same one that's defined 2026 broadly: AI software and semiconductors accounted for the majority of both new unicorns and aggregate value added, meaning the private markets' unicorn pace is now almost entirely a referendum on AI infrastructure demand holding up, not a broad-based venture recovery.
The semiconductor cohort deserves its own look given how concentrated AI capital has become around chip infrastructure this year: five new semiconductor unicorns in a single month is a higher rate than most full years produced during the 2010s, when semiconductor startups were seen as too capital-intensive for traditional venture returns. That's changed as AI compute demand has made chip-adjacent infrastructure -- from photonic interconnects like Lumilens to manufacturing tooling like Source Foundry -- look more like software-margin businesses to investors willing to underwrite the capital intensity.
August's pace also compares instructively to July, when Pulse's archive shows a similar cohort of new unicorns was led by frontier AI labs and DeepSeek-adjacent names rather than the more diversified robotics-and-semiconductors mix seen this time. That shift, from labs to infrastructure and hardware, suggests the marginal venture dollar chasing AI is increasingly flowing toward the picks-and-shovels layer rather than a fifth or sixth foundation-model challenger -- a rotation worth watching for any GP still deciding whether their next fund's AI allocation should chase model companies or infrastructure.
