Europe’s late-stage thaw: AI and defence fuel a valuation mini-boom

• Growth/secondary deals are back in Europe, led by AI apps and defence; some prices look punchy versus fundamentals.

• I’m leaning “quality growth at a fair-ish price”: proven ARR and capital efficiency still matter more than hype.

• Watch secondaries vs primaries, defence procurement realism, and whether 2025 H2 deal flow sustains H1’s pace.

What happened

On 5 September 2025, the FT reported a broad rebound in European start-up valuations, with AI, fintech and defence names drawing competitive term sheets and secondary demand. Examples cited: ElevenLabs exploring an employee sale above $6bn, Mistral eyeing ~$10bn, workflow player n8n entertaining >$2bn, Framer at $2bn, defence names like Quantum Systems/Cambridge Aerospace in market, and Revolut running a staff secondary at $75bn.

Context & data

• Europe VC totals have stabilised: H1’25 investment was c. $30.9bn (Q1 $16.3bn, Q2 $14.6bn) per KPMG’s Venture Pulse. Mega-rounds are picking up even as round counts stay muted.

• AI secondaries creeping up: ElevenLabs is letting employees sell shares at an implied ~$6.6bn valuation (roughly 2x January’s primary). Secondary, not new capital.

• European frontier-adjacent AI still raising: FT says Mistral is in talks for a new round at ~$10bn (separate from the later ASML partnership headlines).

• Fintech bellwether: Revolut opened an employee sale marking $75bn headline valuation; recall FY24 results showed strong profitability, so this isn’t just “story stock” momentum.

My take

I see two currents: (1) real operating momentum (e.g., Revolut’s FY24 profit engine; AI apps with tangible ARR) and (2) FOMO-driven secondaries where price discovery is happening off small floats. I’d mark to unit economics and cash conversion rather than sentiment. For AI apps, I want hard KPIs (ARR growth, net retention, gross margin after model-inference costs, and payback). For defence, I want contracted backlog, TRL maturity and visibility on MoD/EU procurement cycles—funding headlines are easy; deliveries, testing and export approvals aren’t.

Relative to US comps (where Anthropic just closed $13bn), European prices are still a discount on absolute scale—but the gap is narrower in application layers and dual-use. The risk is that secondary marks bleed into primary pricing, stretching LTV/CAC maths and elongating payback under higher-rate regimes. I’d pay up where moats are compounding (distribution, data flywheels, proprietary models/infra, or sovereign demand), but fade froth where growth is inorganic or compute-cost leverage is unproven.

Risks & watch-list

• Secondaries vs primaries: employee liquidity at premium prints can anchor expectations and crowd out disciplined new money.

• Execution: AI gross margins hinge on inference cost curves + vendor terms; watch model/infra mix and capex commitments.

• Defence realism: budgets are up, but procurement is lumpy and milestone-driven; slip risk on trials, integration and export licences.

• Macro/exit: narrow IPO window; trade sales likely for sub-scale assets. If H2’25 VC slows, flat rounds/structured terms could return.