Andreessen Horowitz has made its name writing large cheques into the software layer of industries that seemed too regulated, too technical, or too entrenched to be disrupted by code. Biotech. Defence. Crypto. Now, with a $50 million Series A into Endra, a16z is placing that same bet on MEP engineering — the mechanical, electrical, and plumbing design work that sits at the technical core of every building ever constructed, and that has been done manually, slowly, and expensively for as long as buildings have had wiring.
The round, announced June 1, 2026, brings Endra’s total raised to $75 million in approximately thirteen months. The company closed a $20 million seed round in December 2025, reported by Fortune. Notion Capital and Norrsken VC, both seed investors, returned for the Series A. That velocity — doubling and a half in capital in just over a year — is one signal. The investor leading the round is the louder one.
What Endra Does
MEP is the unglamorous centre of building design. Before a structure can be occupied, someone must design every HVAC system, every electrical circuit, every water and drainage run, and make sure those systems do not collide with each other or with the structural and architectural elements around them. The coordination is complicated. The documentation is enormous. And the tooling — primarily Revit and AutoCAD — was built to make the work possible, not to make it fast. A complex MEP scope on a commercial building can take a team of engineers weeks or months.
Endra’s product generates coordinated MEP designs, calculations, and documentation in minutes. The company describes it as automating work that previously required specialised engineers to model each system by hand, resolve clashes in 3D space, and produce the drawings and calculations that permit applications and construction teams require. According to the company, one fire alarm design that would have taken 300 hours manually was completed in 30 minutes using the platform. Endra says its weakest use case — the discipline where the efficiency gain is smallest — still delivers a 12x improvement. And it says 100% of customers who have run a pilot have converted to paid deployment.
These are company-stated figures, not independently audited results. But the direction of the evidence is hard to dismiss entirely when the customer list is what it is.
The Customer List
The names Endra has confirmed as customers — AtkinsRealis, Buro Happold, Hoare Lea, Ramboll, AFRY, Tetra Tech, and Rejlers — are not construction-tech early adopters trying out a compelling demo. Buro Happold, Ramboll, and AtkinsRealis are among the most respected engineering consultancies in the world. They work on complex, high-consequence projects — stadiums, airports, hospitals, data centres — where a design error is not a learning opportunity. Firms like these do not run pilots on MEP design without engineering-grade confidence that the outputs are trustworthy.
The most telling relationship in that list is AFRY. The Swedish engineering group is both a paying customer and a formal strategic partner, confirmed via an AFRY press release. A company that is simultaneously buying the product and signing a partnership agreement is not hedging. It is signalling that the technology is real enough to build a commercial relationship around, not just to evaluate.
That combination — the conversion rate, the efficiency claims, and the customer names — is the answer to the question a16z had to answer before writing a $50 million cheque into a company that automates something as consequential as building systems design.
Why MEP, Why Now
MEP engineering has resisted software-driven automation for reasons that are structural, not accidental. The design space is enormous: every building is different, every code jurisdiction has its own requirements, and the systems that make a building function must coordinate with each other and with the work of other disciplines. AI that could navigate that complexity had to wait for models capable of reasoning about spatial constraints, engineering rules, and regulatory requirements simultaneously. That capability has arrived in the last two to three years.
What makes MEP especially attractive as a market is that it is enormous and unavoidable. Every building needs it. The engineering consultancies that deliver it are under the same margin pressure as the rest of the industry, and they face the same shortage of experienced engineers. A tool that can compress a months-long scope into days is not an efficiency improvement; it is a capacity multiplier for firms that cannot hire fast enough to meet demand.
The construction design market has seen a wave of AI investment in 2025 and 2026 — but most of it has targeted the architectural and conceptual end of design, or the downstream world of estimating and field coordination. Endra is working in the technical middle: after the architect has committed to a form, before the contractor begins to build. That is a less glamorous position than generative architecture, but it is arguably more load-bearing. You can build without a stunning facade. You cannot build without coordinated MEP.
What to Watch
Endra’s thesis depends on something that is difficult to establish publicly: that the outputs are reliable enough for licensed engineers to stamp and submit. A fire alarm design completed in 30 minutes is impressive if it is correct; it is a liability if it is not. The 100% pilot-to-paid conversion rate is the company’s strongest evidence that engineers who have tested the platform believe the outputs clear that bar — but the volume of pilots behind that figure is not disclosed, and early customers tend to be the most forgiving.
The strategic risk is incumbency. Autodesk owns Revit. Any MEP workflow automation that reaches scale is a product Autodesk can build toward, acquire, or partner against. Endra’s defensibility over the next two years depends on how deeply embedded its outputs become in the engineering workflows of firms like Ramboll and Buro Happold — and on whether those firms become the kind of reference customers that make every competitor engagement start from behind.
The $75 million raised will fund expansion across Stockholm, New York, and San Francisco. That geographic spread — Scandinavia for engineering talent and early enterprise relationships, the US for market scale — is a reasonable configuration for a company trying to establish itself as infrastructure for a global profession before the incumbents notice how large the gap has become.
Bricks & Bytes, which covered the funding announcement, noted the round’s pace as significant context. The point stands. A16z does not typically lead a $50 million round into a company that is thirteen months old and has $20 million in prior capital unless the signals from customers are unusually clear. In MEP engineering, those signals have been almost impossible to get. Endra appears to be getting them.
