AI Is Compressing 3-Month Infrastructure Timelines to Hours

01 /The Concept

Hyperscalers have more capital to deploy into data center construction than they have shovel-ready sites to deploy it into. The bottleneck is not permitting or construction — it is the four-to-twelve weeks required to determine whether a site is viable.

Build was co-founded in 2024 by James Stirrat-Ellis, an architect who designed large-scale cultural infrastructure at Heatherwick, including work on Singapore's Changi Airport Terminal 5, and Ben McClusky, an AI researcher. After moving to California, Stirrat-Ellis taught himself to code and led product at one of the early agentic AI startups before co-founding Build. The company's platform targets the infrastructure development cycle: site sourcing, technical due diligence, power assessment, permitting, and early-stage design. These workflows currently require manually aggregating data from dozens of agencies, consultants, and proprietary databases — sequential processes that represent the primary timeline bottleneck in infrastructure project development.

Build's platform draws on more than 1,600 data sources — planning, environmental, power, and political — and runs agentic analysis across them in parallel. The company reports a reduction in project timelines of more than 95% across 100+ projects in 15 countries, working with governments, Fortune 500 companies, and institutional real estate groups including Tishman Speyer. Index Ventures led Build's €7.4 million ($8.5 million) seed round in June 2026, alongside Pebblebed, Puzzle Ventures, and Tiny.vc, with angels including Sarah Friar (CFO, OpenAI) and John Stecher (CTO, Blackstone).

02 /Why It Matters

The insight Build operates on is that the information required to make infrastructure siting decisions already exists — it just isn't linked, normalized, or queryable at the speed that $100 billion in annual data center investment now demands. By running the aggregation and analysis as parallel AI agents rather than sequential manual research, the platform compresses the information-gathering phase from months to hours. Human judgment doesn't get eliminated; it gets moved earlier in the process, where it can act on more complete information rather than waiting for consultants to finish collecting it.

The data center investment wave has created a specific version of this bottleneck. Hyperscalers and private equity groups have more capital committed to new facilities than they have sites that have cleared due diligence. Assessing whether a parcel can support a 50MW data center requires cross-referencing zoning regulations, grid interconnection capacity, fiber availability, water access, environmental restrictions, and permitting timelines across dozens of agencies. Each of those queries has traditionally been a phone call or a consultant assignment. Build runs them concurrently against 1,600+ normalized data sources.

Stirrat-Ellis's architectural background is not a marketing detail. Infrastructure project assessment outputs must match the format and decision criteria that professional development teams use to commit capital. Generic data aggregation repackaged as reports doesn't survive a site review with an experienced infrastructure engineer. A platform designed by someone who has done the work being automated produces outputs that match the format of the decisions its customers need to make.

03 /Opportunities

The platform Build assembled for data center site selection has an obvious generalization: any infrastructure investment decision requiring multi-source due diligence — energy infrastructure, industrial real estate, transmission corridor permitting, logistics facilities — faces the same information aggregation problem. Data centers are the beachhead, not the ceiling.

The specific adjacent opportunity is supply chain intelligence for the infrastructure development process after site selection. Once a site is identified and due diligence complete, the bottleneck shifts to long-lead equipment procurement: transformers, cooling systems, and backup generators carry 12-to-18-month lead times that can stall a project regardless of permitting speed. The data center development community currently has no centralized visibility into equipment availability, lead time forecasts, or supplier capacity. A platform extending Build's data aggregation model from site assessment to equipment procurement tracking — on the same infrastructure intelligence substrate — would address the next point of failure in the development pipeline, at the moment when the companies experiencing it have not yet articulated the problem as a software need.