North American startups raised $92 billion in seed- through growth-stage funding during the third quarter of 2026, but the headline number hides a sharp change in the market. Funding fell 35% from the prior quarter, even as it stood 50% above the same period a year earlier and remained far below the all-time peak.
The drop came as OpenAI and Anthropic did not produce the huge financing rounds that had lifted earlier totals. AI investment still dominated the quarter, taking roughly two-thirds of all startup funding in the United States and Canada. The money is still flowing, but investors are placing larger bets in fewer places.
AI Still Controls the Biggest Checks
Late-stage and growth companies collected $66.45 billion in Q3, while early-stage startups received $20.6 billion. The largest rounds went to Databricks, which raised $5 billion, followed by Crusoe at $3.9 billion, The Boring Co. at $3 billion, and Cognition at $2 billion.
The public markets offered little relief. The IPO market stayed sluggish during the quarter, with no blockbuster technology debuts. M&A activity was more active, highlighted by Nvidia’s September acquisition of Hugging Face.
That contrast matters for startups and their backers. Private companies can still secure enormous checks when they sit close to the AI infrastructure boom, but the lack of major new public listings shows that investors have not opened every exit route. A strong private funding round does not guarantee an easy path to the stock market.
The Infrastructure Boom Keeps Spending Rising
Even as startup funding cooled, companies spent more than $100 billion on computers and related equipment during the second quarter of 2026, up 60% from a year earlier. Data center construction has increased more than fivefold since early 2022, with annualized spending reaching $85 billion in August 2026.
The costs behind that expansion are rising too. Business computing equipment prices climbed 11.4% in Q2 from a year earlier, while consumer computing equipment prices rose 8.4% in August. Average hourly wages for electrical construction workers increased 7.1% over the same period.
That creates a construction bottleneck. Permitting delays and other limits are preventing data centers from being built fast enough to meet demand, even while AI companies continue to commit money to chips, servers, electricity, and real estate.
Anthropic’s financing shows how large the infrastructure bills have become. A record $60 billion debt package was being arranged to finance the company’s use of Google-designed chips. The package included $42 billion in senior secured loans backed by Broadcom and $18 billion in unsecured subordinated debt, with Blackstone committing about $9 billion.
Anthropic may also issue up to $42 billion in convertible notes to Broadcom. Because the subordinated debt lacks Broadcom’s guarantee, lenders would face Anthropic’s credit risk. The structure shows how AI expansion is moving beyond venture capital and into large-scale corporate borrowing.
Investors Are Questioning the Next Dollar
AI spending is growing, but companies still struggle to predict what their investments will cost. Only 11% of 396 companies surveyed could forecast their AI spending within a 10% margin of error.
Lower prices do not always mean lower costs. In tests, the cheaper AI model cost more than the expensive model in 32% of cases. Google’s more expensive reasoning model completed one task in 85 steps for $1, while a cheaper model took 952 steps, cost $14, and failed to finish.
That gap is forcing investors to look past simple price comparisons. A model that charges less for each step may require far more steps, use more computing power, or fail often enough to erase the savings.
A Bank of America strategy team said generating excess returns from AI beneficiaries is becoming harder because markets have already priced in much of the expected increase in AI spending. The same pressure is visible beyond technology: during the past 12 months, the S&P 500 consumer staples index rose 4.6%, while the consumer discretionary index fell 3.3%. Shares of Lululemon Athletica and Nike each lost about half their value.
Markets Still Reward AI Leaders
Technology markets continued to climb despite the higher costs and more demanding outlook. The Nasdaq Composite rose 286.45 points, or 1.05%, to close at a record 27,477.31. The S&P 500 gained 0.66% to 7,773.95, and the Dow Jones Industrial Average rose 0.18% to 51,267.90.
Nvidia gained 2.1% and closed at a record $238.90, lifting its market capitalization to $5.76 trillion. Roughly 100 technology billionaires among the world’s 500 richest people added a combined $845 billion through September 30, 2026. Elon Musk added $310 billion, accounting for roughly 40% of the index’s total gains.
Still, the wider wealth picture is less steady. The combined wealth of the world’s 500 richest people has fallen about 6% from its mid-June peak of $13.4 trillion to $12.6 trillion.
Bond markets are also sending a warning. The 10-year U.S. Treasury yield reached 5.347%, its highest level since April 2002, while the 30-year yield reached 5.702%, its highest since May 2002. Higher yields raise the cost of financing the companies and data centers driving the AI buildout.
The biggest question now is whether AI demand can support the spending already underway. The ASEAN+3 Macroeconomic Research Office warned that weaker AI demand could cut regional economic growth in 2027 by as much as 1.5 percentage points from the current forecast of 4.1%.
For North American startups, the message is clear: capital has not disappeared, but it has become more selective. AI remains the market’s strongest funding magnet, yet investors are demanding proof that each new model, chip order, and data center can turn enormous spending into durable business value.
Based on
