North America Startup Funding Fell 35% in Q3 2026

North American startups raised $92 billion across seed through growth-stage rounds in the third quarter, a 35% drop from the prior quarter but a 50% increase over year-ago levels, according to Crunchbase data as of early October9. On its face, that sequential decline looks like a correction — a market catching its breath after two years of AI-fueled exuberance. But the more accurate reading is that the quarter's headline number is an artifact of a single missing data point: OpenAI and Anthropic, the two companies that pulled in $110 billion and $65 billion in financings earlier this year, simply didn't raise another megaround between July and September. Strip out the giants, and the underlying venture market looks historically healthy, heavily AI-weighted, and quietly repositioning itself for the most consequential IPO window in a generation.

Where the money went

Late-stage and growth deals absorbed the bulk of the quarter's capital: $66.45 billion, up roughly a third from a year earlier but down sharply from the first half of 20269. The largest rounds went to Databricks, which pulled in $5 billion, followed by AI infrastructure unicorn Crusoe at $3.9 billion, Elon Musk's tunneling venture The Boring Co. at $3 billion, and AI coding startup Cognition at $2 billion9. More than a dozen startups closed late-stage or growth rounds of at least $1 billion apiece — evidence that while the absolute dollar count slipped, the appetite for very large checks into mature companies remained intact9.

Early-stage funding told a similar story of retrenchment from a peak rather than genuine weakness. Investors put $20.6 billion into Series A and B rounds in the quarter, down sequentially but still well above year-ago levels, with round counts holding fairly steady9. The biggest early-stage checks were striking for their size and sector: open-source AI company River AI raised a $1.1 billion Series A, nuclear startup Valar Atomics locked up a $660 million Series B, and chip developer Fab2 raised a $500 million Series A9. Nine-figure Series A rounds are no longer anomalies in AI and deep tech — they are becoming the default for companies building foundational infrastructure.

Seed-stage dealmaking stayed busy, with at least $5 billion going to seed, angel and pre-seed rounds, a figure that typically rises as deals get added to the dataset in subsequent months9. Even at the earliest stage, AI dominated: physical AI startup Walden Robotics picked up $300 million at seed, and Veeda AI secured $90 million9.

AI's share hits two-thirds

The most striking structural fact of the quarter is concentration. Roughly two-thirds of all North American startup funding went to AI-focused companies, per Crunchbase — about $61 billion in total, down sharply from the prior two quarters (which were inflated by the OpenAI and Anthropic rounds) but still one of the highest AI tallies on record9.

That level of concentration has been building for a while and is worth taking seriously as a systemic feature rather than a passing fad. A year ago, 46% of global Q3 funding went to AI companies, and nearly a third of it went to a single company — Anthropic's $13 billion Series F11. Megarounds of $100 million or more consumed a record 70% of U.S. venture capital in that period11. In other words, the Q3 2026 numbers don't reveal a new concentration problem; they confirm an old one. Venture capital has become, functionally, a financing mechanism for a small cadre of AI companies plus a long tail of everyone else — a bifurcation that has pushed sectors like biotech to a 20-year low share of funding11.

The quiet IPO quarter before the storm

The exit picture in Q3 was paradoxical: sluggish in practice, electrifying in prospect. Just 17 venture-backed North American companies went public on major U.S. and Canadian exchanges, collectively raising just under $4 billion9. The largest debuts came from biotech — Adarx Pharmaceuticals, Braveheart Bio and Electra Therapeutics — with advanced nuclear fuel developer Standard Nuclear and scooter platform Lime among the other notable offerings9. Quarter-over-quarter comparisons were always going to be ugly given that Q2 featured SpaceX's record-setting market entry, but even against a typical quarter, the tech IPO pipeline ran dry9.

The real story is what didn't happen. Anthropic is reportedly eyeing a public listing as early as November, and OpenAI filed confidentially for an IPO in June with a debut expected in 20279. This follows a year in which the AI IPO era moved decisively from speculation to execution: Anthropic confidentially filed on June 1 at a $965 billion valuation built on roughly $47 billion in annualized revenue, and OpenAI followed a week later, targeting as much as $1 trillion at listing1. Anthropic has since selected Nasdaq, with reports suggesting a valuation that could reach as much as $2 trillion and capital raised potentially exceeding $100 billion3.

The groundwork was laid in May, when AI chipmaker Cerebras went public at an IPO price of $185 per share, raised at least $5.6 billion, and closed its first day up 68% at a valuation near $67–95 billion — the largest U.S. tech IPO since Snowflake24. SpaceX, which absorbed Musk's xAI in a February merger valued at a combined $1.25 trillion, is pursuing an IPO at a targeted $1.75 trillion valuation that would raise up to $75 billion, surpassing Saudi Aramco's record4.

M&A filled the gap

With IPOs quiet, strategic acquirors did the heavy lifting. Nvidia's September acquisition of open model platform Hugging Face for $12.93 billion was the quarter's biggest deal9. AMD bought World Labs, Fei-Fei Li's AI research lab, in a stock deal valued around $8.2 billion, and Stripe acquired OpenRouter, which routes prompts across different AI models, for a reported $7.5 billion9. In total, 11 North American startups were acquired at reported prices of $1 billion or more during the quarter9.

A slowdown in numbers, not in spirit

So how should the 35% sequential decline be read? The reporting itself draws a clear conclusion: the dip is largely due to the absence of new OpenAI and Anthropic megarounds and doesn't reflect a broad weakening in the venture investment climate9. The deal volume data supports this — round counts at late and early stage held close to prior-quarter levels9. What contracted was the outlier, not the base.

My own reading is that Q3 2026 is the setup chapter, not the story. The two most valuable private companies in the world are on the cusp of becoming public companies, and that transition — not the quarterly funding tally — is what will define the next phase of venture markets. OpenAI's most recent rounds valued it at $852 billion on roughly $25 billion in annualized revenue, and Anthropic's revenue run-rate went from about $1 billion to $47 billion in under 18 months1. Those are the multiples public markets are about to stress-test for the first time.

There are genuine risks in the concentration. Bankers have warned that trillion-dollar offerings from SpaceX, Anthropic and OpenAI — potentially $150 billion in combined fundraising — could soak up institutional capital that would otherwise flow to smaller deals4. Some observers have raised the uncomfortable question of whether nearly half of venture capital being tied up in AI constitutes a bubble11. And OpenAI has told investors it doesn't expect to be profitable until 2030, a fact that will be scrutinized closely in an S-17.

But for venture investors, the maturation of OpenAI and Anthropic into pre-IPO giants is precisely the outcome they've been funding for years. Q3's decline looks less like a bear market and more like a market marking time before the most consequential test of AI-era valuations ever staged.