AI captures 89% of VC dollars as overall funding falls

Key Takeaways

• Global AI investment fell to $149.5B in Q2 2026, down from Q1's record $237.6B
• Mega-rounds ($100M+) represented just 6% of deals but captured 89% of funding
• 37 new AI unicorns emerged in Q2, the strongest quarterly total since Q2 2022

AI venture capital funding dropped 37% in Q2 2026, falling to $149.5 billion from Q1's record $237.6 billion. The decline sounds alarming. It isn't. Strip away the five largest transactions and the market looks stable. What's actually happening is extreme capital concentration: a handful of massive rounds are swallowing nearly all the money while the broader ecosystem ticks along.

CB Insights published its quarterly analysis this week, and the numbers reveal a VC market that's simultaneously cooling in volume and doubling down on proven winners. Rounds of $100 million or more accounted for just 6% of all AI deals. Those same mega-rounds captured $132.5 billion, or 89% of every dollar invested in the sector. The math is stark: a tiny slice of companies is absorbing almost all the capital.

Who took the biggest checks?

Anthropic claimed three of the five largest financings of the quarter. The Claude maker raised $50 billion, $10 billion, and $5 billion in separate transactions. Project Prometheus, backed by Jeff Bezos, pulled in a $12 billion Series B at a $41 billion valuation. DeepSeek, the Chinese AI lab, closed a $7.5 billion Series A and emerged with a $59.2 billion valuation, the highest among Q2's new unicorns.

Anthropic's three rounds alone total $65 billion, or 43% of all Q2 AI funding. Remove that single company and the quarter looks markedly different. This is the new pattern: fewer bets, bigger checks, winner-take-most dynamics.

What happened to exits?

Liquidity events painted a mixed picture. M&A activity fell 10% quarter over quarter. IPOs slipped 6%. AI still led all categories for exits, with 447 transactions, but the pace is slowing. The companies that did exit successfully weren't desperate. They'd been scaling aggressively.

SpaceX grew its workforce 48% over two years to more than 20,000 employees before its record $1.78 trillion IPO. Cerebras, the AI chip maker, nearly tripled headcount before exiting at a $40.6 billion valuation. Biotech company Tubulis expanded staff 71% in a single year prior to Gilead's $5 billion acquisition.

These aren't fire sales. The hiring patterns suggest companies were scaling products and operations, not running out of runway. VCs appear willing to wait longer for exits if the underlying business is growing.

37 new unicorns in one quarter

The unicorn factory is running hot. Thirty-seven new AI companies crossed the $1 billion valuation threshold in Q2, up from 32 in Q1 and the strongest quarterly total since Q2 2022. The global roster of AI unicorns now stands at 671.

The U.S. still leads, producing 20 of the 37 new unicorns, or 54%. But that share is lower than America's existing 67% ownership of the cumulative unicorn population. The gap signals a geographic shift. Asia contributed nine new unicorns, representing 24% of Q2's cohort, well above its 15% share of the total unicorn base. DeepSeek's $59.2 billion valuation drove much of Asia's outperformance.

China's AI sector is producing high-valuation companies despite export controls on advanced chips. The constraints haven't stopped DeepSeek from building competitive models. If anything, the restrictions appear to be accelerating efficiency-focused R&D.

What does concentration mean for smaller startups?

The data implies a bifurcating market. Foundation model companies, AI chip makers, and embodied AI plays are commanding massive checks. Anthropic, OpenAI, and DeepSeek compete for capital measured in tens of billions. Smaller application-layer startups face a different reality.

Early-stage rounds haven't collapsed, but they're not growing either. Seed and Series A investors are more cautious about AI startups that depend on third-party model APIs. The reasoning: if OpenAI or Anthropic can ship a similar feature in six months, the startup's moat evaporates. Defensibility matters more than it did two years ago.

Teams building AI applications often need workflow automation to differentiate. Tools like Zapier, Make, and n8n let startups connect AI capabilities to existing business systems without custom code. The companies finding traction are those that embed AI into specific vertical workflows rather than offering generic chatbot wrappers.

Is the market cooling or consolidating?

Both readings have merit. Total dollars fell. Deal count stayed roughly flat. The median round size increased. These are signs of consolidation, not collapse. VCs are writing fewer checks but for larger amounts into companies with clearer paths to dominance.

The $237.6 billion Q1 figure was an outlier driven by extraordinary fundraises. Q2's $149.5 billion is still historically massive. Compare it to full-year totals from 2021 or 2022 and the quarter looks robust. The perception of decline depends on the baseline you choose.

Foundation model training costs keep rising. Only companies with access to billions in capital can compete at the frontier. This dynamic favors incumbents and makes new entrants increasingly rare at the top tier. The market is selecting for scale.

What comes next?

The second half of 2026 will test whether this concentration pattern holds. Anthropic's fundraising pace can't continue indefinitely. At some point, even the largest backers run out of capital to deploy. When that happens, either valuations compress or deal flow shifts to later-stage companies with clearer revenue paths.

The geographic shift bears watching too. If Asian AI companies continue capturing a disproportionate share of new unicorns, U.S. dominance in AI could erode faster than expected. DeepSeek's valuation already rivals Western competitors. The next DeepSeek may not need Western capital at all.

Frequently Asked Questions

Why did AI venture capital funding drop in Q2 2026?

Q1 2026 set a record at $237.6 billion, inflated by several extraordinary mega-rounds. Q2's $149.5 billion represents a return toward historical norms rather than a market collapse. The underlying deal count stayed relatively stable.

Which companies raised the most AI funding in Q2 2026?

Anthropic led with three separate rounds totaling $65 billion. Project Prometheus raised $12 billion at a $41 billion valuation. DeepSeek closed a $7.5 billion Series A at a $59.2 billion valuation.

How many new AI unicorns were created in Q2 2026?

37 new AI companies crossed the $1 billion valuation threshold, the highest quarterly total since Q2 2022. The global count of AI unicorns now stands at 671.

What share of AI funding went to mega-rounds?

Rounds of $100 million or more represented only 6% of deals but captured 89% of all AI investment dollars, totaling $132.5 billion out of $149.5 billion.

Is Asia gaining ground on the U.S. in AI startups?

Yes. Asia produced 24% of Q2's new unicorns despite holding only 15% of the cumulative unicorn population. DeepSeek's $59.2 billion valuation was the highest among new entrants.

Source: Crowdfund Insider