Key Takeaways
- • Ramtin Naimi, founder of seed firm Abstract, argues on the Invest Like the Best podcast that AI venture has become a capital-markets problem rather than an idea problem, with mega-fund capital timing and structure squeezing founders and investors.
- • Global venture funding reached nearly $680 billion in the first three quarters of 2026, but about a third flowed into just 27 companies raising $1 billion or more, while North American deal count fell 16% year over year as dollars rose 50% (per Crunchbase).
- • Mega-funds are deploying roughly three times faster than modeled and returning capital years early, with some term sheets signed before the cash is held, forcing winners' rounds into two tranches with the second priced at three to five times the primary.
- • OpenAI's latest round drew about $110 billion of a $300 billion target at a $1.25 trillion valuation, below an expected $1.4 trillion, a cold signal that public and late-stage investors are beginning to price the same constraints.
- • Naimi identifies debt, not equity, as the binding constraint on AI compute, noting neolabs need debt at five to ten times equity and that CoreWeave's five-year CDS spread moved past 1,100 basis points.
Summary
Venture capital in late 2026 is constrained by capital structure rather than by a lack of ambitious ideas, according to Ramtin Naimi, founder of the seed firm Abstract, speaking on the Invest Like the Best podcast. He describes mega-funds deploying roughly three times faster than they modeled, returning capital to LPs years ahead of schedule, and in some cases signing term sheets before holding the cash. He cites $150 million term sheets that ask to be funded roughly three months out, and a $50 million term sheet where the lead would wire only $5 million at closing and needed eight to ten weeks for the rest.
The data partly supports this view. According to Crunchbase, global venture funding hit nearly $680 billion in the first three quarters of 2026, yet a third went to 27 companies raising $1 billion or more. Naimi says the losing bidders in hot rounds now pay three to five times the primary price in a second tranche, executed in tandem with the primary round, compared with roughly twice in 2021. He reports more term sheets renegotiated in the past six to eight weeks than at any point since 2021, and he frames the tranche as a structural necessity given how capital-intensive these companies are.
Naimi also draws a distinction between the current market and 2021. He argues that growth is real, with companies growing 10 to 20 times year over year and portfolio companies growing 200 to 700% and moving from negative 15% margins to positive 35%, but that many revenue streams are not clean SaaS. He lists data labeling at 15 to 20% margins, foundation-model-dependent apps, and annualized data sales as examples, and notes public SaaS comparables sit near 13 times revenue. He has not yet seen vertical AI companies replace labor, he says, and the bull case rests on addressable spend such as paralegal budgets.
On the LP side, Naimi says concentration in SpaceX, Anthropic, OpenAI and Databricks has made liquidity for the asset class dependent on a few names. He cites a $2.5 billion fund that began deploying in July 2026 and expects to return to market by January 2027, and another fund that is seeking $5 billion after a $3 billion raise nine months earlier.
The public-market backdrop is weaker. In early November, a broad tech selloff erased more than $1 trillion in market value in one session, according to the article. The compute layer is where Naimi sees the binding constraint: he says debt at five to ten times equity is needed, and CoreWeave's five-year credit default swap spread passed 1,100 basis points, with CoreWeave, Nebius and Iren each falling at least 30% in October. He also estimates there are 80 to 90 neolabs that have raised $1 to $5 billion each, versus an earlier expectation of two or three winners.
Naimi argues that only two to three companies beyond SpaceX, Anthropic and OpenAI will reach trillion-dollar outcomes, and that proactively backing non-obvious founders is the remaining source of alpha as the obvious trade crowds.
The data partly supports this view. According to Crunchbase, global venture funding hit nearly $680 billion in the first three quarters of 2026, yet a third went to 27 companies raising $1 billion or more. Naimi says the losing bidders in hot rounds now pay three to five times the primary price in a second tranche, executed in tandem with the primary round, compared with roughly twice in 2021. He reports more term sheets renegotiated in the past six to eight weeks than at any point since 2021, and he frames the tranche as a structural necessity given how capital-intensive these companies are.
Naimi also draws a distinction between the current market and 2021. He argues that growth is real, with companies growing 10 to 20 times year over year and portfolio companies growing 200 to 700% and moving from negative 15% margins to positive 35%, but that many revenue streams are not clean SaaS. He lists data labeling at 15 to 20% margins, foundation-model-dependent apps, and annualized data sales as examples, and notes public SaaS comparables sit near 13 times revenue. He has not yet seen vertical AI companies replace labor, he says, and the bull case rests on addressable spend such as paralegal budgets.
On the LP side, Naimi says concentration in SpaceX, Anthropic, OpenAI and Databricks has made liquidity for the asset class dependent on a few names. He cites a $2.5 billion fund that began deploying in July 2026 and expects to return to market by January 2027, and another fund that is seeking $5 billion after a $3 billion raise nine months earlier.
The public-market backdrop is weaker. In early November, a broad tech selloff erased more than $1 trillion in market value in one session, according to the article. The compute layer is where Naimi sees the binding constraint: he says debt at five to ten times equity is needed, and CoreWeave's five-year credit default swap spread passed 1,100 basis points, with CoreWeave, Nebius and Iren each falling at least 30% in October. He also estimates there are 80 to 90 neolabs that have raised $1 to $5 billion each, versus an earlier expectation of two or three winners.
Naimi argues that only two to three companies beyond SpaceX, Anthropic and OpenAI will reach trillion-dollar outcomes, and that proactively backing non-obvious founders is the remaining source of alpha as the obvious trade crowds.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.