Welcome back to Upside, where Dan Bowyer of SuperSeed and Lomax Ward of Outsized Ventures, joined by Harry Destecroix, MBE, of SCVC unpack the forces shaping European venture, deep tech and capital.
This week’s conversation reflects a system shifting: Europe is writing bigger checks, physical AI is moving into focus, and the economics of AI are starting to change.
The question is no longer where innovation happens.
It is where value accrues.
The stack isn’t just scaling. It is being contested.
What’s Covered
00:00 Intro and the week’s themes
02:00 Europe’s €15B fund-of-funds and the capital gap
08:00 Seed vs growth: where Europe is actually underfunded
14:00 Bezos’ $100B physical AI strategy
20:00 Roll-ups vs rebuilds in the industrial economy
25:00 China’s token model and the cost collapse of AI
31:00 Security, sovereignty and model choice
36:00 Innovate UK and founder-led policy
41:00 Capex vs revenue: the emerging imbalance
47:00 Predictions and market direction
52:00 Deals of the week
Europe Writes the Check — But Not Yet the System
The EIF’s €15B fund-of-funds is a clear signal. Europe understands it needs scale.
The gap is real. There are materially fewer growth rounds in Europe than in the US. Companies that start comparable do not stay comparable.
But the issue is not only growth capital.
It is capital formation across the stack.
Seed funds remain too small.
Companies are still undercapitalised too early.
Fragmented rounds continue to slow progress.
This creates a compounding disadvantage.
More growth capital helps.
It does not fix the pipeline.
The Allocation Problem Remains
Europe still relies heavily on public capital.
That introduces:
slower deployment
structural constraints
repeated allocation to the same managers
Private capital remains largely absent.
Pension funds—the backbone of venture in the US—are still effectively missing in Europe.
Until that changes, scale will remain episodic rather than structural.
The €15B is progress.
It is not resolution.
Physical AI Moves Into Focus
While Europe debates capital, Bezos is deploying it.
Project Prometheus is a $100B attempt to modernise industry:
acquire legacy assets
apply AI and automation
rebuild productivity
This is not venture. It is industrial strategy at scale.
The Next Layer of AI
The digital economy has already been built.
The physical economy—manufacturing, logistics, infrastructure—has not.
That is where AI is now moving.
Two models are emerging:
Upgrade incumbents
faster
constrained
Rebuild from scratch
slower
structurally cleaner
Both will exist. Only one will dominate.
Europe’s Exposure
Europe sits directly in the path of this shift.
Manufacturing is a larger share of GDP
Robotics talent is strong
Productivity growth is weak
Physical AI is not optional.
If the transition happens elsewhere, Europe absorbs the downside without capturing the upside.
The Economics of AI Are Changing
China is pushing a different model.
AI is not positioned as premium.
It is being driven toward commodity pricing.
Inference costs are falling rapidly.
Tokens are becoming the unit of value.
This reframes the market.
Cost, not capability, increasingly drives adoption.
The Trade-Off
The choice is becoming clearer:
trusted, higher-cost systems
cheaper, less controlled alternatives
Security matters.
Cost often decides.
This tension will define the next phase of adoption.
Policy Starts to Shift
In the UK, Innovate UK is being reshaped with a more founder-led approach.
The shift is from:
funding projects
tobacking companies
If executed well, it reduces friction and improves outcomes—particularly in deep tech.
If not, it reinforces the existing system.
Execution remains the constraint.
Capital Is Moving Ahead of Revenue
One imbalance is becoming clearer.
Infrastructure investment is accelerating:
hundreds of billions annually
trillions projected
Revenue is growing.
But not at the same pace.
Part of this gap is narrative:
projected run rates
early enterprise usage
experimental demand
The risk is not collapse.
It is misalignment.
Capability vs Adoption
AI capability continues to advance.
The constraint is shifting to:
adoption
integration
monetisation
Markets are pricing potential.
The economy is still catching up.
That gap is real.
The Competitive Landscape
Three models are emerging:
US
Scale, capital, integration
China
Cost, control, commoditisation
Europe
Talent, research, fragmented capital
Each is coherent.
Only one currently compounds efficiently.
Closing Reflection
The shift is clear.
The AI race is no longer about building the best model.
It is about capturing value.
Europe is moving.
But unevenly.
Capital is increasing.
But not yet optimally deployed.
The opportunity remains open.
The system will determine whether it is captured.
Listen on Apple Podcasts or Spotify — and if you’re building in AI, deep tech or enterprise infrastructure, this one’s worth queueing.
