Tech Is Globalizing. The Money Is Lagging.

Dispatches from my travels. Spain, France and Germany edition.

Photo by OhTilly on Unsplash

First published in my Forbes column here.

In the final days of SuperReturn in Berlin, SpaceX priced the largest IPO in history, raising $75 billion at a valuation near $1.75 trillion, and later growing to over $2 trillion. I had spent that week in Berlin and the week before at South Summit in Madrid.

For a decade I have argued that talent is global and capital is provincial.

Two weeks of conferences left me with a sharper version of the same point: tech has globalized, but the money has not (yet) caught up.

Talent Has Gone Global And AI Is Accelerating This

Annie Hamill of Adams Street opened SuperReturn with a keynote called “State of Global Venture 2026,” which made the point directly.

Europe’s startup ecosystem is now worth roughly $5.6 trillion, larger than several of the continent’s own public markets, and a record 27,000 founders started European companies in 2025. There are over 200 unicorns in Europe today. By comparison, when I started investing in venture in 2013, the term “unicorn” had only just been coined, for a small club of companies that was almost entirely American.

Taking a European lens as an example, Michael Collins of Alumni Ventures made the point: there are great founders everywhere, and the Bay Area holds no monopoly on grit or vision. People still move there, but the diffusion is real and accelerating. Alexander Schmitt of Lightspeed pointed to European labs taking categories outright from Europe, ElevenLabs in voice and Legora pressing Harvey in legal. Importantly, he argued the old valuation upside ceiling has lifted, that companies on the continent can now build past the $10 billion mark that used to cap their ambition.

Unfortunately, The Money Has Not (Yet) Followed

More than 90% of venture dollars now go to AI-native companies according to Adams Street, and the overwhelming share sits in a handful of US names. As has been well documented, value is pooling in a few private giants worth hundreds of billions each. Because of the ever-present power law, returns come from a thin tail of outliers, and for now allocation is piling into those, still largely in the US.

LPs Are Looking For Exits

The loudest theme of the week was liquidity (or at least the desire for it!). Venture distributions in 2025 ran at about half their 46-year average. Fewer than three dozen US tech companies went public last year, down from more than 120 four years earlier.

A majority of venture funds now take 14 years or more to fully wind down, and the bar to go public has stretched from roughly four years in the late 1990s to twelve today.

So the industry rebuilt the plumbing. Secondaries have become a structural approach. By figures cited on stage, around $20 billion of venture liquidity now flows through them, close to a third of the total. Continuation funds, long a private-equity instrument, are migrating into venture to crystallize value that has nowhere else to go.

The exit environment is seemingly improving. SpaceX just went public, OpenAI and Anthropic have both filed confidentially to follow, and Adams Street put the next twelve months of debut exits at more than $4.2 trillion (including SpaceX), on the order of Germany’s GDP and roughly double the entire prior decade. At the time of this piece, SpaceX agreed to acquire Cursor for $60 billion, the largest acquisition of a venture-backed startup on record.

But look at where the value sits: SpaceX, OpenAI, Anthropic, Cursor, Cerebras, Databricks. Three or four trophy names are not (yet) a market reopening. Yes, the door has cracked open for companies everyone already knows. For the other 24,000-odd venture-backed AI companies, and for the funds holding aging positions, the math has not changed.

Broader trends evolving rapidly

Strip away the panels and three other trends popped out again and again.

First, AI is collapsing the cost of building, which lowers the barrier to scale anywhere. This was a throughline of a panel I joined the week before at South Summit in Madrid, “Building Outliers: Scaling Companies in the AI Era,” alongside JC Glancy the cofounder of ZenBusiness and now Candosa (a former portfolio company at a previous firm, and current one at Fluent Ventures, the VC firm I work at, respectively). Entrepreneurs in emerging ecosystems can get to early product market fit with limited capital, and get to profitability much earlier. We call this camel seed strapping. But what it means is an increasing distribution of innovation globally.

Second, the AI roll-up trend in US is scaling globally (and may actually outstrip it globally). This was a theme on multiple panels at SuperReturn. The services industry that AI can attack is worth more than $15 trillion, dwarfing the software market venture has historically funded. The strategy of rebuilding fragmented professional services at the AI layer is real, and American capital is already crowding into it. Not every roll-up clears the bar; the ones that work need real margin expansion and plentiful affordable targets, as I have argued. The deeper risk, named all week as the central one, is commoditization. When the model is the only edge, an open-source release erases it, so the moat has to be proprietary data and workflow. That is where geography helps. In the US, a hundred funded teams chase each category. In most other markets, far fewer do.

Third, the opportunity is also a responsibility. Will Porteous of RRE made the point that venture sits at the far edge of the risk spectrum, and that position carries responsibility, because “what gets funded gets built, and what gets built can change everything”. I would push it one step further from the geographic lens. If the talent, the adoption, and the cost curve have all gone global, then deploying the overwhelming majority of venture capital inside a fifty-mile radius is a choice, not a law of nature. The encouraging part is that the responsibility and the return point the same way. The underserved problems and the least-crowded entry points are increasingly the same places.

The founders have gone global. So has the adoption, and so has the cheap compute that now lets a few engineers in Lagos or São Paulo build what once took a hundred in Mountain View. Capital is the last input still behaving as if geography were destiny. For the allocators who nodded along in multiple panels, the question is what, exactly, they are still waiting for.