a16z puts $1.1B behind AI's physical bottlenecks

Why it matters

AI's infrastructure shortage is creating a founder market around power, chips, cooling, networking and robotics. A16z is reserving $1.1B to compete for it.

a16z, presented by cofounder Ben Horowitz and four other investors, launched the Machine Age Fund on August 28, saying it raised $1.1 billion to back the chips, memory, networking, power systems, data centers and robots required to keep AI scaling.

The announcement, first reported by TechCrunch, gives a dedicated pool of capital to a hardware practice that a16z says already accounts for over 20% of its recent startup deal flow. That percentage lacks a disclosed time period or deal count, though the direction is clear: founders building physical infrastructure have become important enough to warrant their own fund inside a16z.

The move also rewrites the operating premise behind the venture franchise Horowitz and Marc Andreessen started in 2009. Andreessen's argument that software would eat the world gave a16z its organizing thesis. The Machine Age Fund starts from a constraint software cannot route around. AI models need physical capacity, from electricity and cooling to high-bandwidth memory and interconnects, and each layer takes time and heavy capital to build.

a16z's fund announcement describes AI development as a "social and national imperative" and says it wants to "open the throttle and accelerate the physical buildout of AI." The language is characteristically expansive. The allocation is concrete: $1.1 billion reserved for founders working where AI demand collides with supply chains, manufacturing limits and the laws of physics.

The operators behind the hardware turn

Horowitz presented the fund with Martin Casado, Raghu Raghuram, David Ulevitch and David George. The announcement separately identifies Guido Appenzeller, Shangda Xu and Erin Price-Wright among a16z's hardware and infrastructure specialists. Their resumes explain why a16z believes it can move beyond its software reputation without starting from zero.

Horowitz co-founded Loudcloud in 1999 and ran its transformation into Opsware, software built to automate data-center operations. Hewlett-Packard acquired Opsware in 2007. That experience gave Horowitz a close view of infrastructure cycles, including what happens when a technically sound product arrives before the hardware and connectivity around it are ready.

Martin Casado co-founded Nicira, which used software-defined networking to make network infrastructure programmable. VMware bought Nicira for $1.26 billion in 2012. Casado subsequently ran VMware's networking and security portfolio, which a16z says reached a $600 million annual revenue run rate by the end of 2015.

Raghu Raghuram joined a16z in October 2025 after nearly three decades at VMware, including a stint as CEO. His relationship with Horowitz reaches back to Netscape, where Horowitz hired him as a product manager. At a16z, Raghuram serves as managing partner and invests across growth and infrastructure.

David Ulevitch started the donation-funded DNS service EveryDNS while at Washington University in St. Louis and later co-founded OpenDNS. Cisco acquired OpenDNS for $635 million in 2015. David George, who leads a16z's growth practice, brings the later-stage financing side of the fund's strategy.

This is an unusually operator-heavy group for a hardware fund. Its members have built, sold or managed businesses spanning data-center automation, networking, cybersecurity and enterprise infrastructure. a16z is betting that those histories will help it evaluate founders whose products combine difficult engineering with manufacturing, supply-chain and sales risk.

AI racks are becoming industrial projects

a16z's fund announcement argues that computing infrastructure is being redesigned "all the way down to the electricity." According to a16z, rack power has climbed from roughly 5 to 10 kilowatts for H100-era systems to about 100 to 250 kilowatts for Rubin racks. a16z expects 1-megawatt racks within three years.

Those projections are a16z's own, and the fund announcement does not provide an independent forecast. They still illustrate the investment thesis. Denser compute creates demand for new cooling equipment, electrical systems, materials, networking and real estate. The economics look closer to industrial development than the low-capital software businesses on which venture funds built much of their recent performance.

That changes what founders need from an investor. Hardware teams face longer development cycles, larger early financing requirements and dependencies on suppliers and manufacturing partners. a16z says its existing go-to-market, talent, marketing and network capabilities are ready to serve hardware founders and provide access to relevant customers and suppliers. The announcement does not establish separate fund-specific staffing, service levels or operating commitments.

A16z has already backed businesses including Unconventional AI, Nexthop, Volta, Atoms and Mind Robotics, according to the announcement. The fund also points to earlier investments in Skydio, SpaceX, Anduril and Waymo as evidence that hardware has long existed inside the portfolio. a16z does not specify whether the recent investments named in the announcement came from the new fund.

The Machine Age Fund arrives as specialist investors target physical technology. Eclipse announced $1.3 billion across two funds aimed at modernizing physical industries. Lux Capital and DCVC also compete for deep-technical founders. a16z's advantage is its scale: Andreessen Horowitz reported over $100 billion under management as of April 30, 2026.

a16z has been adding personnel around the same thesis. RuntimeWire reported in July that a16z hired Connor Love to expand its defense and space investing, another capital-intensive corner where software, autonomy, manufacturing and government procurement increasingly meet.

A dedicated fund makes the commitment harder to reverse

Creating a named vehicle does more than give a16z another marketing banner. It reserves capital for companies that can require repeated financings before factories, robots or infrastructure networks reach commercial scale. It also gives hardware founders a clearer entry point into a16z, an organization historically identified with software investing.

The fund announcement does not identify its limited partners, investment period, check-size range, portfolio target or deployment schedule. Those details will determine how concentrated the strategy becomes and whether a16z intends to finance companies through the most expensive stages of production.

There is another governance question as a16z expands across connected layers of AI infrastructure. Axios reported on August 18 that the Justice Department was investigating partner board roles at potentially competing portfolio companies. The report concerned a16z's positions at Databricks and Fivetran, rather than the Machine Age Fund. A hardware portfolio spanning chips, networking, data centers and robotics could produce similar competitive overlaps that require careful board and information-rights decisions.

For founders, the immediate proposition is simpler. a16z has put $1.1 billion behind the view that AI's next constraints are physical. Software ate the world. Keeping it fed has become a hardware job.