Startup funding 2025: everything labeled AI is going up

Actualizado: 2026-05-03

First-quarter 2025 data from Crunchbase and CB Insights is clear: global startup funding has rebounded after two-plus years of correction. The problem is that the rebound is not evenly distributed. Almost all new money goes to companies presenting themselves as AI, while the rest of the ecosystem is flat or declining. It is worth separating signal from noise before drawing conclusions about whether the bull cycle is back.

Key takeaways

• Global Q1 2025 funding has grown vs the same quarter in 2024, but AI concentration is historic.
• Excluding AI-company rounds, aggregate growth is modest or negative in some segments.
• Consumer fintech, quick commerce and crypto remain in adjustment with double-digit declines.
• A non-trivial fraction of AI labeling is opportunistic: products whose advantage depends on the underlying model, not proprietary technology.
• Buying critical software from startups today warrants reasonable financial diligence on the vendor.

What aggregate numbers say

Per Crunchbase’s quarterly summary, global Q1 2025 funding has grown significantly vs the same quarter in 2024, with megacap rounds in the billions pulling the average up. CB Insights reports a very similar pattern.

The important detail is concentration. When AI-categorized companies are excluded, aggregate growth drops to a much more modest number. A handful of very large rounds to foundation model labs and compute-infrastructure companies account for most of the volume. The median round for other AI startups, those building on top of existing models, is more restrained but has also risen versus 2024.

What is happening in non-AI segments

Segments that concentrated capital in 2021 and 2022 — consumer fintech, quick commerce, crypto — remain in full adjustment:

• Double-digit declines in volume and round count.
• Several previous-wave unicorns closing internal rounds at valuations that, without saying so openly, are significant markdowns.

More traditional verticals like B2B SaaS, climatetech or digital health sit in an in-between position: they haven’t fallen as hard, but neither enjoy the AI rebound. In Europe the picture is even more moderate, and Spain stays in the group with contained growth, no megarounds but reasonable seed and pre-seed activity.

The AI-labeling distortion

A now more visible phenomenon is opportunistic labeling. Companies that in 2022 presented as vertical SaaS are repositioning as AI startups, sometimes with real technical basis and sometimes with a thin wrapper over a third-party model. Investors know this and filter accordingly, but inflated labeling affects aggregate data.

The practical consequence is that AI funding figures are less solid than they appear. A non-trivial fraction corresponds to products whose competitive advantage depends on the underlying model. If OpenAI, Anthropic or Google change pricing or restrict the API, those startups lose the business. This provider-dependency risk connects directly with the lessons of FinOps applied to AI, where token-cost exposure also depends on the third-party model.

Implications for founders

The read for founders raising now is clear:

• With a defensible AI angle: the best market in years. Seed and Series A valuations are comparable to 2021, and closing times have shortened. The risk is closing at high multiples that later force unsustainable growth rates.
• Without an AI angle: raising is slower, investors demand much more substantive traction, valuations remain compressed. Extending runway with bridge financing or chasing profitability over growth are reasonable paths.

For technical teams, pressure to add some AI story to the product is strong. The advice is not to force it. If there is a real application that improves the product, go ahead; if it is purely decorative to please an investor, it usually derails the roadmap and creates product debt. The same trends are visible in Y Combinator’s 2025 cohorts, where vertical agents with measurable value clearly dominate.

Implications for software buyers

A less-discussed point is that of enterprise customers buying software from startups. In an environment of extreme concentration, vendor financial health is very uneven:

• Well-capitalised AI startups have two to three years of runway.
• Those in other categories often do not, and discontinuity risk is real.

Anyone buying critical software should demand reasonable financial visibility before signing multi-year contracts: months of runway, whether they are at breakeven, who the investors are, and whether they are committed for the next round.

My read

Q1 2025 does not mark the return of a generalised bull cycle. Rather it marks a divergence. New money is betting near-monotonically on AI, with historic concentration on a handful of very large bets. The rest of the ecosystem remains in the correction cycle running since 2023.

Two scenarios are plausible: the AI wave keeps generating returns sufficient to justify valuations and eventually drags other categories up; or a few large bets miss expectations, multipliers compress, and capital retreats sharply. Not a 2022-style correction, because underlying productivity is more solid and measurable. Rather an adjustment in the speculative portion, with clear survivors justifying valuations and clear casualties.

For today’s founders: build with financial discipline even when the market offers generosity. For investors: differentiating real competitive advantage from AI wrapper is the work of the coming quarters.