Build: Automating the Built World

"We're going to be shipping thousands — hundreds of thousands — of buildings across the world."

That was the cornerstone of the seed conversation the Build co-founders, James and Ben, had with the Pebblebed team. It cemented our conviction in the magnitude of their ambition: constructing the digital substrate for the built world. Pam asked if their vision included "robots and everything," and the duo responded "yes" without skipping a beat. This is not your father's B2B SaaS.

The strength of their conviction comes from deep familiarity with the built world. James is that rare founder in whom domain expertise and technical chops live in the same brain. Before Build, James was CTO and CPO at a fast-growing AI agent company, and before that, he was an architect. He worked on Singapore's ~$10 billion national airport expansion¹, and so has seen firsthand the process of developing critical infrastructure for sovereign-level workloads. (That audacity carries over to his personal life; James is an avid skydiver, with over 200 jumps, and nearly every Build team member has submitted to the initiation rite of jumping out of a plane. The Pebbles have not taken the plunge – yet.)

Before any development can begin, the site needs to be sourced and vetted. Successful site selection takes a specialist team stitching together thousands of data sources by hand, from environmental risk to community sentiment. This takes months and costs teams anywhere from hundreds of thousands to millions of dollars. The range of services in the journey to creating a building constitute as much as 5-10% of a project's total cost. Mega projects like data centers today often cost several billion dollars.

More importantly, it throttles timelines. On things like Technical Due Diligence ("TDD"), Build instead compresses weeks into hours at a fraction of the cost with their purpose-built agent swarms. Customers pay not just for the high quality of the report vs legacy industry, but the speed itself is enormously valuable. However, naively chaining clankers is problematic: when every step has to be right, error compounds across steps (1-pᴺ). A hundred 99%-reliable agents still yield a report that's wrong two out of three times. Reliability is crucial here - imagine sinking millions and years only to find out that you have no power for your building.

You manage this with structure, and this is where Ben's research in multi-agent systems comes in. Build's first workers (their term for agents that run multi-agent workflows to achieve reliable results) runs hundreds of sub-agents in a four-tier hierarchy, each with its own tools, context, and guardrails, executed in parallel. From there the surface area only grows: Build has now expanded into more workflows and products across the building lifecycle, including design, acquisitions and even asset management.

Build's focus hits another important tailwind, inference which we believe to be the most important workload today. Last year, ~$350 billion was spent on computing infrastructure, and it is still growing. The four largest hyperscalers spent a record ~$410 billion in 2025 and are estimated to reach ~$725 billion in 2026 — up 77% in a year, the overwhelming majority of it data centers, chips, and the power to run them. Build uses AI in the best way possible, to build more AI with their partners. Some of the largest data center developers in the world are now Build's customers. These include the UK Government on its AI Growth Zones, where they are targeting £100 billion (~$130B) in data center development, and institutions that manage hundreds of billions in real estate assets.

In this age of AI, where cognition is becoming abundant and cheap(er), the economic surplus flows to the inputs that are now the binding constraints: land. Build is developing the systems managing those scarce resources for a world where AI is ubiquitous. We are proud to be partners with James and Ben in their seed round. And when the robots do show up to pour the foundations, remember we said yes without skipping a beat, too.

---

¹ Which Singapore-born GPs Pam and Tammie are grateful for.