Manhattan's office boom is the visible symptom. The real story is that applied AI is clustering by vertical, city by city, and the map matters for every enterprise vendor decision you make.

The Wall Street Journal ran a property story last week about AI startups fueling a Manhattan office boom. Landlords signing leases with AI companies. Founders choosing loft spaces over remote-first setups. The subtext of the piece was cultural: these companies want people in a room together. Read it as a real estate story and you will miss the more important signal underneath it.

Applied AI is not distributing evenly. It is clustering, city by city, vertical by vertical. And where a company is built shapes what it is built to solve.

New York Is Not Trying to Be San Francisco

According to Tech:NYC's 2025 Ecosystem Snapshot, New York hosts more than 2,000 AI startups, over 1,000 of which have raised venture capital, collectively attracting more than $27 billion since 2019. Manhattan now hosts more seed and Series A companies than San Francisco, with 543 early-stage rounds closed in a single year compared to 486 in the Bay Area.

Those numbers invite a comparison that misses the point. San Francisco is still the dominant hub for foundation model development. The Bay Area accounted for the vast majority of US AI funding in 2025. New York is not competing for that category. It is winning a different one.

Enterprise AI and business-to-business software account for roughly 35% of total capital raised in the New York ecosystem. Financial technology AI is the second-largest category. Health technology AI, legal AI, and AI infrastructure round out the major clusters. Startups like Kalshi in prediction markets and Hebbia in document intelligence are building products that would be harder to conceive anywhere else, because the customer base and the regulatory complexity they are solving for are concentrated in lower Manhattan.

That is what the office demand story is actually measuring. Startups building for Wall Street, the media industry, and large hospital systems want proximity to their customers. The lease is a business development strategy.

Europe Is Not One Ecosystem

The reflex is to group European AI into a single narrative and contrast it with the United States. That framing is too blunt to be useful.

European AI funding reached a record $21.8 billion in 2025, up 58% in a single year. In the first quarter of 2026, AI claimed more than 50% of all European venture capital for the first time, with AI startups pulling in $9.2 billion in the quarter alone, per Crunchbase data. The funding is real and accelerating.

But Europe is a collection of distinct hubs, each with a different strength.

The Structural Problem Europe Has Not Solved

Capital is arriving in Europe. The talent has always been there. Nearly 50% of late-stage funding for European deep tech spinouts still comes from outside the continent, primarily from the United States. Europe produces the research. It has not yet retained the capital formation that converts research into scaled companies.

There is a second constraint that is harder to fix. A European AI startup in 2026 is built on American cloud infrastructure, distributed through American platforms, and discovered through American search and advertising systems. Before a single line of proprietary code creates value, the company is paying rent to landlords who are also increasingly its competitors. This is not a regulatory problem. It is a structural dependency that regulation alone cannot resolve.

The European Innovation Council's Step Scale-Up program, with a 300 million euro budget in 2026, is a direct attempt to address the scale-up gap. Whether it moves fast enough to retain the companies it funds is the actual test.

What This Means for Enterprise Vendor Decisions

CIOs and CTOs sourcing AI vendors are usually evaluating a product, a price, and a roadmap. Geography rarely enters the conversation. It should.

A New York-built AI company has been designed around financial services compliance, media workflows, or healthcare data requirements from day one. Its assumptions about customer behavior, regulatory environment, and integration complexity are different from a San Francisco company building for developer infrastructure or a Berlin company building for government and defense. The ecosystem a company grows up in shapes the problems it considers normal.

That is not a soft consideration. It is a configuration and support cost that shows up in implementation.