a16z's $1.1B Machine Age Fund: The Venture Giant Goes All-In on AI's Physical Layer

On Friday, August 28, Andreessen Horowitz announced the Machine Age Fund, a new $1.1 billion vehicle that is the firm’s first fund dedicated entirely to hardware. The mandate, in the firm’s own words, is to “open the throttle and accelerate the physical buildout of AI” — investing in everything from AI chips, memory, and networking gear to data centers, robotics, power infrastructure, and even home AI appliances.

For a firm built on the thesis that “software eats the world,” the pivot is striking. a16z made its name — and its returns — backing software platforms, from Facebook and Twitter to Airbnb, Stripe, and more recently the leading AI application layer. The Machine Age Fund inverts that logic: the binding constraint on AI in 2026 is no longer model quality or clever code, but the physical infrastructure the models run on. As a16z put it in the announcement post, every layer of the AI stack is “hitting the wall of today’s supply chain capability, and the limits of physics and computer science.”

What the numbers say

The firm backed its thesis with unusually concrete engineering figures:

• Compute density: rack-level compute increased 28x from an H100 rack to a Rubin rack.
• Rack power: moved from roughly 5–10 kW to 100–250 kW today, heading toward 1 MW per rack within three years.
• Data center scale: shifting from tens of megawatts to hundreds of MW, with gigawatt-scale campuses now being planned.
• Power sourcing: moving beyond grid-only supply to grid-plus-behind-the-meter or fully captive generation.
• Supply-side growth: the hardware industry is structured to grow 20–30% per year at most — not the triple-digit growth needed to catch up with AI demand.

That mismatch — demand compounding at AI speed against a supply chain built for gradualism — is the core investment argument. Someone has to rebuild the memory hierarchy, the interconnects, the cooling, the materials, the electrical systems, and the real estate. a16z wants to own a piece of whoever does.

A hardware team hiding in a software firm

The fund is led by partners Martin Casado (who also runs the firm’s infrastructure software practice), Raghu Raghuram (former VMware CEO, who will make investments from the new fund alongside Casado), David Ulevitch and Erin Price-Wright (who lead hardware and US manufacturing deals through the American Dynamism practice), Shangda Xu and David George (investors across the AI infrastructure stack from silicon to compute platforms), and Guido Appenzeller, former CTO of Intel’s Data Center Group. Ben Horowitz co-authored the announcement.

The firm is quick to note this isn’t a total departure. a16z led Skydio’s Series A in 2016, invested early in SpaceX, wrote its first check into Anduril in 2019, and was among the first venture investors in Waymo’s 2020 raise. Recent hardware bets include Unconventional AI, Nexthop, Volta, Atoms, Heron Power, and Mind Robotics. Hardware deal flow, the firm says, has grown from a trickle to over 20% of its pipeline over the last couple of years.

Context: the capital rush into AI’s physical layer

The Machine Age Fund lands amid an extraordinary wave of capital targeting AI infrastructure. Global AI-related debt raised for infrastructure has exceeded $400 billion, and 2026 has seen neoclouds like Lambda securing billion-dollar debt facilities just to buy chips, Broadcom reportedly exploring an AI financing package approaching $100 billion, and hyperscalers committing to gigawatt-scale campuses. Venture capital has largely watched that game from the sidelines — the checks were too big and the collateral too physical. With $1.1 billion, a16z isn’t trying to outspend Microsoft or Amazon; it’s positioning to fund the startups that supply, optimize, or disaggregate pieces of that buildout.

It’s also a hedge every AI investor is now making in some form: if the model layer commoditizes and the application layer gets crowded, the scarce assets — power, silicon, memory bandwidth, robotic actuators — capture the value. “Machine intelligence is going vertical,” the firm writes, noting that as AI evolves from chat to reasoning to coding and other knowledge work, both the demand and the token intensity of the work “increase by orders of magnitude.”

What to watch

Three open questions will determine whether this fund is visionary or vintage-topping:

1. Exit paths for hardware. Chip and infrastructure startups are capital-intensive with long timelines. Can a16z’s GTM, talent, and marketing machine — built for SaaS — actually serve founders whose products take years to ship?
2. Competition with strategic capital. When Nvidia, Broadcom, and sovereign funds are all writing infrastructure checks, a $1.1B fund is a mid-sized player. The edge has to be deal selection and founder support, not scale.
3. Timing.† If AI infrastructure spending cools — as some analysts warn amid rising AI debt loads — hardware startups funded at peak valuations will feel it first. If demand keeps outrunning supply, as the 28x rack-density figure suggests, the fund will look prescient.

One thing is certain: the boundary between “tech investing” and “industrial investing” has officially dissolved. When the firm that coined “software eats the world” raises a hardware fund to build “what intelligence runs on,” it’s a signal about where the next decade of value creation sits — not in the code, but in the machines that run it.