Houston/ Science, Tech & Medicine

The Woodlands Startup Mavria Hits $1B Valuation Building AI Data Centers In Factories

AI Assisted Icon
Published on November 17, 2026
The Woodlands Startup Mavria Hits $1B Valuation Building AI Data Centers In FactoriesSource: Eben Calhoun / Wikimedia Commons

A startup based just north of Houston says it has raised $130 million and reached a $1 billion valuation building AI data centers the way car makers build vehicles — on an assembly line rather than from scratch at every site. Mavria, headquartered in The Woodlands, Texas, says its first product, Terrain, is a factory-built data center platform meant to bring computing power online faster, with more predictable delivery and lower costs than traditional construction.

The company says its latest funding round was led by Paradigm, a technology investment firm, with participation from AMP PBC, Liquid 2, Critical Frontier and Dream Ventures, according to Mavria. “We've raised $130 million to date across pre-seed and seed rounds. Our latest round values Mavria at $1 billion,” the company said in its announcement. Mavria is described elsewhere as headquartered in The Woodlands, Texas, just north of Houston, according to DatacenterDynamics.

The company says it is already putting the new capital to work securing critical supply chains, scaling up manufacturing and expanding its team, and that it is hiring across engineering, manufacturing and operations. Mavria says its first system, Terrain 01, is in production in Houston right now — though outside reporting has yet to confirm details like a named customer, pricing or a delivery timeline for that unit, as detailed by citybiz.

A Factory Approach to a Years-Long Problem

Mavria's pitch centers on turning a construction project into a manufacturing process. The company says Terrain integrates power, cooling, controls and information-technology pods into a single standardized system built using repeatable manufacturing processes, with the goal of cutting deployment timelines, improving cost predictability and simplifying procurement for customers. Terrain is designed for frontier AI laboratories, cloud providers and data-center developers, per the seed facts behind citybiz's reporting.

That pitch responds to a real bottleneck. Building a data center can take roughly three to six years and cost an estimated $10 million to $37 million per megawatt, according to InnovationMap. Mavria co-founder and chief executive Adam Blake has said the company aims to deploy infrastructure as a product rather than custom-building each project, and that he and his co-founders previously worked together at the Department of Energy, where they identified gaps between the country's computing ambitions and its industrial infrastructure capacity.

There is a notable discrepancy over how much power Terrain's first unit can actually deliver. InnovationMap reported that the first Terrain system is a 5-megawatt unit, while TechTimes reported that Mavria had not published a capacity figure and that the 5-megawatt number instead came from Dealroom, a separate data provider, according to TechTimes. Neither account has been independently reconciled.

Why Factory-Built Capacity Is Suddenly a Hot Bet

Mavria isn't alone in betting that modular manufacturing can outrun traditional data-center construction. Romania-based Runware says it plans to scale its factory to produce 30 megawatts of data-center capacity per day by 2028 and aims to bring its modular pods online within 48 hours — down from roughly three weeks today — after raising more than $60 million, as reported by TechCrunch, which described an industrywide push to deploy data centers faster and more efficiently.

Even factory-built approaches face real-world limits once they leave the assembly line. TechTimes reported that NVIDIA's comparable DSX POD system moves cabling, power distribution and cooling piping into the factory, but the finished pod still has to arrive at a site that already has high-voltage power, a coolant loop and a network handoff ready to go — meaning grid readiness remains a constraint no factory can fully engineer around.

The Power Demand Driving the Race

The push to manufacture data centers faster comes as AI computing's electricity appetite draws increasing scrutiny. A Harvard-led study estimated that 403 U.S. hyperscale data centers together used between 68 and 99 terawatt-hours of electricity between May 2024 and April 2025, depending on operating scenarios, according to the Harvard T.H. Chan School of Public Health. That demand amounted to roughly 1.8% of total U.S. electricity consumption and was associated with about 25.7 million metric tons of carbon dioxide, the researchers found.

Fossil fuels generated roughly 54% of the electricity powering those hyperscale data centers, mainly natural gas at 38% and coal at 15%, per the same study. The researchers also cautioned that their data could not separate electricity used for general-purpose computing from electricity used specifically for AI workloads at each facility — a limitation that leaves the AI sector's precise share of that demand unresolved.

Deep Ties to Federal Energy Policy

Blake's résumé links Mavria's founding team directly to recent federal energy policy. He joined the Department of Energy in January 2025 as a senior adviser to Energy Secretary Chris Wright and went on to lead the department's Department of Government Efficiency team, with responsibilities that included implementing a federal directive to streamline regulatory pathways under the Atomic Energy Act. That directive targeted bringing three new non-commercial nuclear reactors to criticality by July 4, 2026.

Blake and his co-founders have said they identified constraints in the aging U.S. power grid and in industrial supply chains as key obstacles standing in the way of AI infrastructure demand, and that expanding America's industrial base will grow increasingly important to AI competitiveness. The founding team's stated focus is coordinated engineering, procurement and manufacturing rather than treating each data center as a one-off construction job.

Paradigm, which led the round, was founded in 2018 and is led by Matt Huang and Alana Palmedo, with offices in San Francisco, New York and Washington, D.C. The firm says it manages more than $11 billion in assets and invests across stages and geographies — from early-stage financing to public markets — in companies developing artificial intelligence, robotics, cryptocurrency and other emerging technologies. Its portfolio includes Zipline, True Anomaly, Stripe, Kalshi, Hyperliquid and Citadel Securities.

Houston-Science, Tech & Medicine