Andreessen Horowitz raises $1.1 billion Machine Age fund for AI hardware infrastructure

Aug 28, 2026

Venture capital firm Andreessen Horowitz has closed its $1.1 billion Machine Age Fund, marking a major shift toward physical AI hardware and infrastructure. The fund targets critical bottlenecks caused by a 28-fold surge in AI rack power density, which is projected to reach 1 megawatt per rack within three years. As physical constraints in copper cabling, memory bandwidth, cooling, and electrical grid access threaten to slow AI scaling, venture capital is rotating back into hardware, with physical AI startups raising $47.4 billion in the first half of 2026.

Machine Age Fund launch

▪The Machine Age Fund's investment scope spans semiconductor chips, memory systems, networking, storage, robotics, home AI appliances, and power-efficient edge devices.
▪Andreessen Horowitz closed the $1.1 billion Machine Age Fund, dedicated exclusively to AI hardware and physical infrastructure, on August 27, 2026.
▪The Machine Age Fund was announced by Andreessen Horowitz general partners Ben Horowitz, Martin Casado, Raghu Raghuram, David Ulevitch, and David George.
▪The Machine Age Fund sits alongside other Andreessen Horowitz vehicles, including a $15 billion group of funds closed in January 2026.

AI rack power density surge

▪Andreessen Horowitz projects that AI data center rack power requirements will reach 1 megawatt of electrical power within three years.
▪AI rack power density jumped 28-fold between NVIDIA's H100 generation and its Rubin GPU racks.
▪AI data center racks that previously drew 5 to 10 kilowatts now consume between 100 and 250 kilowatts.

Hardware infrastructure bottlenecks

▪Traditional cooling systems engineered for ten-kilowatt racks cannot contain the heat generated by modern 250-kilowatt AI racks.
▪Memory bandwidth and cheap, high-bandwidth memory across the hierarchy have become critical bottlenecks for running AI models.
▪Copper cabling has hit its physical limits at high densities, making optical or alternative interconnects a necessity for AI clusters.

Data center grid constraints

▪Goldman Sachs forecasts that U.S. data center power demand will grow from 41 gigawatts in 2026 to 66 gigawatts in 2027.
▪Securing electrical grid connections for data centers currently takes five to seven years, while physical data center construction takes 12 to 18 months.

Venture capital hardware shift

▪The top five cloud and infrastructure companies are projected to spend over $600 billion in capital expenditure in 2026.
▪Physical AI startups raised $47.4 billion globally across 521 deals in the first half of 2026, representing an 80% year-over-year increase.
▪Hardware opportunities have risen to comprise over 20% of Andreessen Horowitz's incoming deal flow.