Venture capital didn’t recover in 2025.

It split in two.

One-half got $100M rounds, soaring valuations, and an infinite runway.

The other half got ghosted.

UK venture funding hit $23.6bn — up 35% vs 2024, and the third-highest year on record. But the real story isn’t that the market came back.

It’s where it came back.

And how brutally selective that comeback really was.

Venture is back… for the winners.

And increasingly, those winners are being defined by AI — either because they’re building it, selling it, or being reshaped by it.

The market didn’t broaden — it weaponised capital.

A healthy venture market funds many sectors, rewards multiple paths to growth, and lets plenty of “good businesses” raise capital.

2025 didn’t look like that.

It looked like capital concentration.

UK growth was driven by 36 mega-rounds of $100m+, with Revolut raising the largest at $2bn. Globally, funding rose even as deal volume continued falling.

Translation:

• fewer cheques
• more capital per cheque
• more power-law outcomes
• capital as a weapon for the leaders

For every Revolut raising $2bn, hundreds of “good businesses” got nothing.

So 2025 wasn’t a broad-based reopening.

It was the market saying:

“Show me inevitability, and I’ll overfund you.”

“Show me momentum, and I’ll wait.”

 

The middle got hollowed out.

AI didn’t just perform well — it absorbed the market

The clearest explanation for 2025 is also the simplest:

AI captured 48% of total venture funding in 2025 — the highest share ever recorded.

When half the market’s risk capital deploys into one theme, the rules change.

• Good is no longer fundable
• Great is still not enough
• Only category leadership clears the bar

This is the fundamental shift.

It’s not that the market “likes AI.”

It’s that AI became the default destination for conviction capital — and that changes how every other company gets judged.

“AI winners” increasingly means “AI agents” (software that acts, not assists)

The momentum isn’t clustering around “AI features.”

It’s clustering around agents and copilots, especially:

• coding agents
• legal agents
• voice AI
• developer tooling
• agent frameworks + integration layers

The market’s next belief is apparent:

AI doesn’t win by being smart.

AI wins by taking work away from humans.

And the companies that win won’t be the ones with the slickest demos.

They’ll be the ones that own workflows — because workflow ownership is where durable economics live.

The next wave is Physical AI.

The frontier is moving from digital intelligence to physical capability.

In 2025, robotics reached a record $40.7bn in funding, up 74% year-over-year.

That’s not a curiosity.

That’s a capital reallocation signal.

Physical AI has the same winner-take-all mechanics as software:

• platform-level data advantages
• hardware-software integration
• compounding learning loops
• immense distribution moats

Once these flywheels start spinning, they don’t just create better products — they create unfair companies.

It’s not just AI: other sectors are re-accelerating

There’s an important nuance heading into 2026:

Other sectors are also gaining momentum:

• Fintech (+35% YoY)
• Digital health (+18%)
• Defence tech (+128%)
• Quantum tech (+211%)

But here’s the real pattern:

AI is becoming a horizontal force, not a vertical category.

The biggest companies may not be “AI companies.”

They’ll be category leaders in:

• defence
• manufacturing
• health
• logistics
• finance

…that win because AI exists.

The rich get richer (and it’s not subtle)

When the market reprices a small set of companies upward, it does two things at once:

1. It makes the leaders impossible to catch
2. It makes everyone else look second-tier by comparison

And in venture, perception becomes reality faster than it should.

Capital attracts talent.

Talent attracts distribution.

Distribution attracts customers.

Customers attract capital.

That’s the flywheel.

And 2025 poured fuel on the winners’ flywheels.

Why this might be wrong (and why I don’t think it is)

“Maybe this is just a cycle — and breadth returns in 2026.”

Possibly. But the structural forces here aren’t temporary: AI’s platform effects, data moats, and winner-take-all economics make this different. When AI becomes the operating system for industries, being second doesn’t mean lower margins. It means irrelevance.

“Maybe mega-rounds create fragility.”

They do. But in a category concentration environment, being underfunded is worse. The fragility of too much capital beats the certainty of too little.

“Maybe regional markets diverge.”

The pattern is global: fewer deals, larger rounds, AI dominance. The geography changes. The mechanics don’t.

What this means for founders: the 2026 playbook

The strategy isn’t “raise because AI is hot.”

The strategy is:

• Build proof of dominance (even in a narrow wedge)
• Win distribution early (partners > paid ads)
• Show compounding economics (not just usage)
• Reduce dependency on future funding
• Sell outcomes, not features

Create inevitability.

Because in this market, hope doesn’t raise money.

Momentum does.

Closing thought

2025 was the concentration year.

2026 is the reckoning.

Companies that are based on narrative will need to show revenue.

Companies that rely on revenue will need to show dominance.

And the ones left out?

They’re building in a market that forgot they exist.

The only question that matters now:

Are you building inevitability — or are you building hope?

Have a great week!