AI architecture startup Dream launches with $17M in funding

Dream Technologies Inc., a startup using artificial intelligence to generate building designs, launched today with $17 million in funding.

Andreessen Horowitz led the seed round. It was joined by SV Angel, XYZ, Zigg Capital and several others.

San Francisco-based Dream was founded last year by former Palantir Technologies Inc. executives Andy Elder and Arjun Mathur. Elder headed the software maker’s product design team, while Mathur was the senior vice president of its private sector business. Dream’s leadership team also includes former engineers from Okta Inc. and Siemens AG.

Crafting a building plan manually can take a team of architects several months. Dream has developed a series of AI models designed to speed up the task. According to the company, its algorithms can generate hundreds of design variations in a few minutes. Dream says its software thereby not only accelerates construction projects but also lowers costs.

The company trained its models on data about 100 million square feet of U.S. real estate. The properties in the dataset include commercial buildings, apartments and single-family homes. Dream stores the blueprints its models process using a custom schema, a file that defines how a set of records should be organized. The schema is designed to ease the task of verifying architectural calculations.

Architecture teams craft building designs through an iterative process that involves numerous revisions. Those revisions are time-consuming because implementing even a small change often requires engineers to modify several blueprint elements. For example, adding an extra cooling system to a server hall might necessitate upgrades to the room’s electrical wiring and ventilation system.

Dream says its models speed up such changes. Additionally, they reduce the amount of work involved in adapting an existing building design to a new location. A retailer looking to open a store in a new city could reuse one of its existing venues’ designs and have Dream adapt it to local building codes.

The company’s models generate blueprints through a multistep process. First, they analyze the site where a building is set to be constructed and review the associated operational constraints. Dream uses those data points to generate a layout that defines which building elements should be placed where. Its models then define more granular details such as the dimensions of individual structural components.

“A hallucination in an LLM is a typo. A hallucination in a building is a structural fault, with potentially catastrophic consequences,” said Elder. “We built Dream to be exact — down to how a duct routes through a wall — because precision at that level is what actually shortens timelines and takes cost out of a project.”

The company is also working on a tool that can speed up internal design projects. According to Dream, the tool combines the U.S. real-estate dataset that underpins its models with information about 73 popular design and architectural styles.

Dream doesn’t sell its software but rather provides ready-to-use blueprints to construction and architecture firms. The company acts as an architect of record, which means that it takes legal responsibility for its plans. Dream disclosed today that it’s already supporting 10 construction projects in California and Texas with a combined value of $20 million.

Image: Dream