Why I Built a Map of the AI Ecosystem (And What I Found)

The AI market is moving faster than most people can track.

Not just the model releases: those are visible enough. The pace at which new companies are forming, old ones are being acquired, categories are being defined and then redefined, and capital is being deployed at a scale that would have been implausible two years ago.

Most people working in business: founders, operators, executives: are making decisions that depend on understanding this market. Which tools to use. Which vendors to trust. Which capabilities to build internally versus buy. Which threats to take seriously. Which opportunities are real and which are noise.

They're making these decisions with patchy information. A few newsletters, some LinkedIn posts, whatever came up in the last team meeting. The full picture of what's happening, who the players are, and how everything connects is not accessible in a single place.

That's why I built AI City.

What it is

AI City is an interactive map of the global AI ecosystem. 235 companies across 15 districts, organised into three zones: the software layer, the financial quarter, and the industrial zone.

The software layer is what most people think of when they think of AI companies: foundation model labs like OpenAI, Anthropic, and Google DeepMind; developer tools and coding assistants; agent platforms; creative and generative tools; open-source projects; infrastructure and APIs.

The financial quarter is the capital layer: the VC firms funding the race (a16z, Sequoia, Khosla), the corporate strategics deploying billions as strategic investors (Microsoft's $13B into OpenAI, Amazon's $4B into Anthropic), the PE and growth funds positioning for the consolidation phase, and the sovereign capital that is reshaping the infrastructure investment landscape entirely.

The industrial zone is the physical substrate: the chip makers, the data centre operators, the custom silicon programs at the hyperscalers, and the energy companies that are now more relevant to AI than they have been to any previous technology cycle.

Each district has a building on the map. Each building has companies. Click into a district and you see the key players, what they do, their valuation where known, and the daily intelligence on what's happening there. The whole thing updates automatically. Live model pricing. Benchmark scores. Funding news.

What I found building it

I started this as a tool for my own use. I wanted to understand the ecosystem properly: not just the names, but the structure. Who funds who. Who competes with who. Which companies are genuinely building and which are wrapping APIs. Where the real moats are forming and where the apparent leaders are more fragile than they look.

Building the map forced a clarity that reading about it doesn't produce.

A few things that struck me as I built it:

The capital concentration is remarkable. A small number of investors have backed almost every significant company in the foundation model space. Khosla, Sequoia, and a16z between them have positions in OpenAI, Anthropic, Mistral, and most of the significant agent platforms. This isn't unusual for venture: but the scale of the individual bets is. A $13B single-investor commitment (Microsoft/OpenAI) is not normal VC behaviour. The strategic calculation is different at this scale.

The sovereign capital story is underreported. Everyone knows about the hyperscalers. Fewer people have tracked the UAE, Saudi, and Japanese sovereign funds' entry into AI infrastructure. MGX, PIF, and SoftBank are not making portfolio bets. They are funding national strategic positions in the infrastructure of the next economy. The Stargate initiative: $500B committed to US AI infrastructure: is the largest capital commitment in technology history. It barely registers in most coverage.

The energy constraint is the story nobody is telling loudly enough. AI data centres are drawing power at a scale that is restructuring the energy market. Microsoft signed a 20-year nuclear PPA specifically to power AI infrastructure. Helion has a fusion power purchase agreement with a 2028 delivery commitment. Oklo is building small modular reactors designed for data centre co-location. This is not a future problem being planned for. It is a present constraint being worked around at enormous cost.

The open-source vs proprietary race is genuinely contested. This was not obvious twelve months ago. DeepSeek's R1 showing frontier-level performance at a fraction of US training costs changed the calculation. The assumption that the frontier would always be proprietary and that open models would always lag is no longer safe.

Who it's for

AI City is built for people who need to understand this ecosystem to make better decisions in it.

If you're a founder deciding which AI infrastructure to build on, the map shows you who controls the stack you're depending on and who funds them. If you're an operator evaluating which tools to deploy, it gives you the context to assess vendors beyond the marketing. If you're an investor or advisor, it's a live view of where capital is concentrating and which categories are becoming real.

It's free to use. The full intelligence: live pricing, benchmarks, district detail: is available with a free account.

It's at [leonhardwick.com/ai-city](/ai-city).

---

*AI City is updated automatically every day at 6am UTC. The weekly digest: AI City Weekly: is available by subscription. Sign up from the map.*