a16z Raises $1.1B Machine Age Fund for Physical AI

Andreessen Horowitz is treating the physical limits around artificial intelligence as an investable category of their own. The venture firm has raised $1.1B for the Machine Age Fund, a new vehicle spanning chips, memory, networking, storage, data centers, robotics, AI appliances, power, cooling, materials, electrical systems, and real estate.

The announcement makes hardware an official a16z investment motion. That turns the fund into more than another pool of capital: it is a wager that the next layer of AI value will be built by founders working against supply chains, electrical capacity, heat, materials, construction schedules, and the stubborn refusal of physics to accept a software update.

What Andreessen Horowitz Raised

The Machine Age Fund is a completed $1.1B raise announced on August 28, 2026. The official announcement is signed by Ben Horowitz, Martin Casado, Raghu Raghuram, David Ulevitch, and David George, bringing together leaders from the firm's Infrastructure, Growth, and American Dynamism practices around one broad hardware mandate.

Andreessen Horowitz says the vehicle will invest across the computer infrastructure on which AI runs, including semiconductors, memory systems, networking, and storage. Its definition of the stack continues outward into full systems such as data centers, robotics, and home AI appliances, then into the physical support around them: cooling, materials, electrical equipment, power delivery, and real estate.

The firm has not disclosed the Machine Age Fund's limited partners, check sizes, stage allocation, geographic limits, deployment period, or economics. It also has not explained how the vehicle overlaps with the Infrastructure, American Dynamism, Growth, and other funds inside the more than $15B a16z announced in January 2026. Those omissions matter because a dedicated thesis can still share deals, reserves, or ownership targets with adjacent pools, and the public announcement does not tell founders how that internal map will work.

Why AI's Bottleneck Is Becoming Physical

The Machine Age Fund begins from a problem software cannot abstract away. As AI systems handle more reasoning, coding, and other knowledge work, the demand for computation grows alongside the intensity of each workload. More useful intelligence therefore requires more memory bandwidth, faster interconnects, denser racks, larger data centers, and a power system capable of serving them without turning reliability into collateral damage.

a16z says compute density increased 28X from an NVIDIA H100 rack to a Rubin rack. The firm also says rack power moved from roughly 5-10 kW to 100-250 kW and could reach 1 MW over the next three years, while some data-center campuses move toward gigawatt scale. These are a16z-reported figures, but the power constraint is visible outside the announcement: NVIDIA's 800 VDC architecture is designed to support 1 MW IT racks beginning in 2027 and addresses copper, conversion loss, cooling, safety, and serviceability at that density.

That shifts the AI investment conversation. Model quality and application demand still matter, but neither can outrun the physical system beneath them forever. A rack of accelerators without enough memory bandwidth, power, or cooling becomes an expensive demonstration of heat, while a data-center plan without grid capacity remains a spreadsheet waiting for a substation.

What the Fund Is Actually Buying

The broad mandate suggests a16z is not trying to pick one hardware layer. It is investing in the handoffs between layers, where a better chip still needs memory, networking, storage, power conversion, cooling, packaging, manufacturing capacity, and a system that customers can deploy. Those interfaces are where technical progress often becomes an operating problem, and where a company can own value that is difficult to reproduce with code alone.

a16z points to recent investments in Unconventional AI, Nexthop, Volta, Atoms, Heron Power, and Mind Robotics as evidence that the thesis was already forming before the fund received a name. The firm also cites earlier exposure through Skydio, SpaceX, Anduril, and Waymo. The portfolio list is intentionally wide, reaching from semiconductors and networking into defense, aerospace, autonomous systems, and robotics, which shows that “machine age” is being used as an infrastructure frame rather than a narrow chip fund.

The announcement also says hardware startups have grown from a small portion of a16z's deal flow to more than 20% over the last couple of years. That is a firm-reported pipeline measure, not fund performance, but it helps explain the timing. A dedicated vehicle gives the partnership a way to organize conviction, staffing, and founder outreach around a category that was already consuming more attention.

The Institutional Bet Behind the Capital

Andreessen Horowitz says it had more than $100B under management as of April 30, 2026 and invests from seed through growth across a wide range of technology sectors. The Machine Age Fund is still a notable institutional turn because a16z built much of its public identity around software and the venture-platform model: capital paired with recruiting, marketing, policy, legal, and go-to-market support.

Hardware founders need that operating platform to solve a different set of problems. Customer introductions matter, but so do supplier relationships, manufacturing talent, infrastructure partners, certification pathways, site access, and the patience required when a physical build cannot be patched after launch. a16z says its network can reach the customers, suppliers, and other parties relevant to hardware reinvention, making that service promise part of the fund's investment thesis.

The claim deserves attention because it can be measured. If the platform works, portfolio companies should spend less time searching for industrial relationships that a software company rarely needs at the same depth. If it does not, the Machine Age Fund will still have $1.1B to deploy, but its differentiation will look much closer to capital with a familiar brand attached.

What Founders and Operators Should Watch

The first signal will be portfolio construction. Check sizes, stage concentration, follow-on reserves, geography, and the balance between components and complete systems will reveal what “all of the computer infrastructure” means once investment committees begin choosing among real companies. The disclosed mandate is broad enough to support a semiconductor startup, a grid technology company, a robot manufacturer, or a data-center systems business, but broad language does not guarantee balanced deployment.

The second signal will be whether a16z can make its platform model useful inside physical supply chains. Hardware founders often need commercial and technical relationships to arrive together because a customer may depend on manufacturing yield, deployment capacity, power availability, and supplier reliability before a purchase becomes credible. The fund's value will be easier to see when those handoffs shorten, not merely when another financing announcement appears.

The Machine Age Fund gives a16z a formal stake in AI's physical economy. Its durable record will be built deal by deal, as electrical limits become companies, suppliers become strategic partners, and a venture firm famous for software learns how much of the next computing cycle must be assembled in the physical world.