Welcome back to Upside, where Dan Bowyer and Mads Jensen of SuperSeed and Lomax Ward of Outsized Ventures unpack the forces shaping European venture, technology and capital.
This week wasn’t driven by a single headline. It felt more like a system shifting beneath the surface.
Nvidia is mapping a path toward a trillion-dollar AI market. Hyperscalers are deploying hundreds of billions in capital. And yet, the real battleground is no longer just models — it’s distribution, enterprise control, and monetisation.
The stack isn’t just scaling, it’s being reordered.
What’s Covered
00:00 Intro and the week’s themes
02:00 Nvidia GTC and the inference step-change
07:00 The trillion-dollar AI demand question
12:00 Hyperscaler capex and the return of leverage
18:00 The shift from model performance to distribution
23:00 OpenAI’s monetisation dilemma
28:00 Enterprise AI as the real battleground
34:00 Market concentration and systemic risk
39:00 Capability vs adoption gap
45:00 Where humans still capture value
50:00 Startups vs hyperscalers
55:00 Europe policy signals and what they mean
Nvidia Builds the Spine — But Not the Outcome
Nvidia continues to execute at a level that almost resets expectations by default. The latest roadmap shows step-function improvements in inference performance and efficiency, reinforcing its role as the backbone of the AI economy.
But the question is shifting.
It is no longer just whether Nvidia can keep delivering. It is whether the ecosystem built on top of it can justify the scale of what is being built. Nvidia may own the infrastructure, but it does not automatically own the value created by it.
The Capex Surge Is Becoming a Balance Sheet Story
Hyperscalers are now investing at a scale that would have seemed implausible just a few years ago. What is changing is not just the size of that investment, but how it is being financed.
Increasingly, this is no longer purely funded by operating cash flow. Debt is entering the system.
That matters.
Because the narrative shifts from growth to balance sheet. From expansion to sustainability. When infrastructure is built ahead of monetisation, risk does not disappear — it accumulates.
The AI Race Has Moved Up the Stack
For the last two years, the narrative was simple: better models win.
That narrative is breaking.
The leading labs are converging on the same prize — enterprise — but taking very different routes. OpenAI is reorganising around enterprise and coding. Anthropic is leaning into distribution partnerships. Mistral is enabling enterprises to build on their own terms.
The common thread is clear.
Distribution, not just capability, determines who captures value.
OpenAI’s Strategic Fork
OpenAI sits in the most interesting position in the market.
It has scale, brand, and one of the largest user bases in technology. But it still faces a fundamental question: how to monetise that scale.
That creates a fork.
If it cracks consumer monetisation, it becomes a generational company. If it is forced into enterprise competition, it enters a market where others may have structural advantages.
Either way, the next phase of AI runs through this decision.
Risk Is Creeping Back Into the System
For a long time, the conversation has been dominated by growth.
Now, risk is starting to reappear.
Markets are increasingly concentrated in a small number of AI-driven companies. Capital is flowing faster than ever. And leverage is beginning to enter the system.
The concern is not that AI is overhyped.
It is that capital may be getting ahead of monetisation.
Capability Is Racing Ahead of Adoption
AI capability continues to accelerate at a remarkable pace.
But adoption is not keeping up.
There are clear breakthrough use cases, from coding to healthcare. Yet many organisations are still early, still experimenting, still trying to define ROI.
This creates a gap.
Markets are pricing what AI can do. The real economy is still figuring out how to use it.
Where Value Actually Sits
The assumption that value spreads evenly across the AI stack is unlikely to hold.
Instead, it concentrates.
Right now, one of the clearest areas of concentration is the final layer — the last mile of work. Judgment, verification, integration, and execution.
AI can do most of the work.
But the final decision still carries the most weight.
Startups Are Being Squeezed From Below
Products like Google Stitch highlight a structural shift. Hyperscalers are no longer just infrastructure providers. They are moving up the stack and competing directly with application-layer companies.
For startups, that changes the game.
It is no longer enough to build a better product. You need to build something that cannot be easily absorbed by the platform you depend on.
Europe Shows Signs of Movement
There were also signals of progress in Europe.
The UK is moving to limit non-compete clauses, which could unlock talent mobility. The EU is pushing forward with a unified corporate framework under EU Inc, aiming to simplify company building across the bloc.
These are meaningful steps.
But as always in Europe, the constraint is not ambition.
It is execution.
Closing Reflection
This episode ultimately comes down to one shift.
The AI race is no longer about who builds the best model.
It is about who captures the value.
Infrastructure is scaling. Capital is flowing. Capability is accelerating.
But ownership of distribution, workflows, and monetisation will define who wins.
And that is where the real competition now sits.
Listen on Apple Podcasts or Spotify — and if you’re building in AI, enterprise software or deep tech, this one’s worth queueing.
