a16z Launches $1.1B Machine Age Fund for AI

a16z raised $1.1 billion for its Machine Age Fund to accelerate the physical buildout of AI. The firm frames AI as alchemy that turns sand into thought, and thought at scale now demands a rebuild of American compute infrastructure.

The video pitch is blunt. The United States is short power, short compute, short chips, and the entire compute stack is being rebooted for AI.

Why hardware, why now

The fund will back the full physical stack on which AI runs. That includes chips, memory, networking and storage, plus complete systems from data centers to robotics to home AI appliances.

The authors are Ben Horowitz, Martin Casado, Raghu Raghuram, David Ulevitch and David George. Casado leads infrastructure, Raghuram is managing partner on Growth and Infrastructure after decades running data center system software, Ulevitch leads American Dynamism, and George leads Growth.

The timing rests on physics, not narrative. Compute density per rack is up 28x from an Nvidia (NASDAQ:NVDA) H100 rack to a Rubin rack, rack power has moved from 5 to 10 kW to 100 to 250 kW today, and the firm expects 1 MW racks within three years.

Data center scale is moving from tens to hundreds of megawatts to gigawatt campuses, with power increasingly drawn from grid plus behind-the-meter sources. Copper interconnects inside the rack are already at their limits.

What the fund will chase

a16z lists what must change: cheaper and higher-bandwidth memory across the hierarchy, faster and more scalable interconnects between nodes, power efficient edge devices, and the cooling, materials, electrical and real estate to support them.

The supply side is not built for this rate. Hardware vendors are accustomed to 20% to 30% annual growth, the firm wrote, while AI infrastructure needs triple-digit growth to catch demand.

Deal flow explains the shift. Hardware startups were a small share of a16z pipeline a couple years ago. They are now over 20%.

Recent bets cited include Unconventional AI, Nexthop, Volta, Atoms, Heron Power and Mind Robotics. Earlier hardware marks include leading Skydio Series A in 2016, SpaceX, Anduril in 2019 and Waymo in 2020.

Why this matters for AI builders and investors

This is not a contrarian hardware turn. It is a follow-on to a12z's American Dynamism thesis and to a market where memory, high-bandwidth memory and power have become the binding constraints, not model weights.

Competitors have moved the same way. Lux Capital, DCVC and Founders Fund have enlarged deep tech and industrials books, while hyperscalers and neoclouds have locked up gigawatt-scale capacity through vendor-financed campus deals and long-term power purchase agreements.

For founders, the bar changes. a16z says its go-to-market, talent and marketing platform is now tuned for hardware, with Guido Appenzeller, former CTO of Intel (NASDAQ:INTC) Data Center Group, plus Shangda Xu and Erin Price-Wright helping bridge customers, suppliers and manufacturers.

That signals where diligence will focus: manufacturability, supply chain, and power and cooling integration, not just silicon performance. Startups selling only a faster chip without a path to rack-scale deployment and contracted power will struggle to clear investment committees.

The test is near term and measurable. Watch 2027 rack deployments, HBM pricing and lead times, and whether 100 to 250 kW rack designs actually ship at volume without derating. If 1 MW rack timelines slip, capital will rotate back to software efficiency plays.

a16z is betting the bottleneck stays physical, and that someone has to fund it before thought at scale can be abundant.