Andreessen Horowitz (a16z) has closed a $1.1 billion fund dedicated entirely to the physical infrastructure behind artificial intelligence, the venture firm announced on August 28, 2026. The Machine Age Fund is its first vehicle built specifically for hardware — chips, memory, networking gear, power systems, cooling, robotics and the data centers that host AI workloads.

Partners Ben Horowitz, Martin Casado, Raghu Raghuram, David Ulevitch and David George said the fund reflects a shift in what limits AI progress. According to the firm, a two-year software lead can be erased by better models, but a large GPU cluster converts capital directly into capability — meaning physical capacity, not code, is becoming the harder constraint to build against.

Why hardware, why now

a16z pointed to compute density jumping roughly 28x between Nvidia’s H100 generation and its upcoming Rubin racks, pushing per-rack power draw from 5-10 kilowatts to 100-250 kilowatts today, with some projections reaching 1 megawatt within three years. Data centers built for AI have scaled from tens of megawatts to gigawatt-scale campuses, straining supply chains that historically grew 20-30% a year and now face triple-digit demand growth.

The firm said hardware-focused startups now make up more than a fifth of its deal pipeline, up from a negligible share a few years ago — evidence, it argues, that founders are increasingly targeting the physical layer rather than purely software products.

What it will back

The fund will invest across chips, memory, networking, storage, data centers, power and cooling infrastructure, materials, and consumer AI hardware, as well as the robotics systems that bring AI into physical environments. a16z cited existing bets in the space, including its compute startup Volta, which recently signed a multibillion-dollar deal with Anthropic, and Travis Kalanick’s robotics venture Atoms. The firm also pointed to earlier hardware bets — Skydio, SpaceX, Anduril and Waymo — as precedent for wagering on physical technology years before it went mainstream.

The fund marks a structural shift for a firm long associated with software-only bets, formalizing a wager that the next phase of the AI buildout will be decided by chips, power and steel as much as by model weights.