- Andreessen Horowitz launched its $1.1 billion Machine Age Fund on August 28, 2026, a direct bet that AI’s next constraint is physical infrastructure rather than model capability.
- The fund targets the full hardware stack, chips, memory, networking, data centres and robotics, with 7-10 year investment horizons that reflect hardware’s longer build cycles and thinner margins compared to software.
Andreessen Horowitz quietly built its reputation on software bets. Its $1.1 billion Machine Age Fund, launched August 28, 2026, is a different kind of wager: that the next decade of AI returns will be made in silicon, cooling systems and data centre architecture, not code. For founders trying to raise serious capital in AI hardware right now, the fund’s structure and stated thesis offer a useful map of what top-tier VCs actually want to see.
Find the Real Bottleneck
Securing hardware VC starts with identifying an acute, quantifiable problem in the AI infrastructure stack. Andreessen Horowitz‘s thesis for the Machine Age Fund states that AI infrastructure is hitting capacity limits across chips, memory and power simultaneously. That framing matters for founders because it defines the scope of what the firm considers investable.
Rack power requirements are the clearest example of where the physics is outrunning the engineering. Compute racks have gone from drawing 5-10 kW to 100-250 kW, with projections of 1 MW per rack within three years, according to the fund’s stated thesis. That single data point opens viable startup territory in advanced cooling, power delivery and data centre architecture designed to sustain those loads. The fund’s mandate runs wide: chips, memory, networking, data storage, robotics and data centres. Custom ASICs, high-bandwidth memory improvements, faster node interconnects and power-efficient edge devices all sit within scope. A pitch that maps a specific bottleneck to a specific performance or efficiency gain over existing methods is far stronger than one that gestures at the general hardware shortage.
The Team Investors Actually Fund
In deep tech, team composition is the primary diligence signal at early stages. The Machine Age Fund is led by general partners Martin Casado and Raghu Raghuram, a former VMware CEO, a pairing that combines systems-level technical depth with large-scale commercial execution experience. That combination is not accidental, and it telegraphs what the firm expects to see in founding teams.
The strongest hardware founding teams pair chip designers, electrical engineers and materials scientists with founders who have run manufacturing operations or hardware supply chains. Technical depth alone is not enough. Investors need confidence that the team can navigate fabrication timelines, advanced packaging constraints and the go-to-market dynamics of selling physical products to enterprise buyers, all of which differ substantially from software sales cycles. The ability to attract specialists in areas like systems architecture or advanced packaging after the initial raise further strengthens the case.
IP and the Full-Stack Requirement
A16z’s stated position is that AI’s constraints are systemic, not localised to a single component. That view has direct implications for what kind of IP attracts capital. Incremental improvements on existing hardware designs are unlikely to clear the bar. The fund is looking for companies that can re-architect parts of the stack from first principles, or build integrated systems that redefine how AI infrastructure operates.
The specific gaps the fund has named are instructive: cheaper, higher-bandwidth memory across the memory hierarchy; faster and more scalable interconnects between nodes. Startups with defensible IP addressing either of those will stand out. As a reference point, Nvidia‘s move from H100 to its Rubin rack architecture reportedly delivered a 28x increase in compute density, the kind of step-change improvement that illustrates what “order-of-magnitude” means in this context. A prototype that demonstrates measurable advantage, not just theoretical capability, is the minimum threshold for serious early-stage conversations.
A full-stack product vision matters too. Hardware that ships with software tooling and clear system integration is more attractive than a component looking for a home. The fund has stated it will invest in complete systems spanning chips, memory, networking, storage, data centres, robotics and home AI devices. Founders who can show how their hardware fits and performs within a larger deployment context are better positioned than those pitching a single component in isolation.
Manufacturing strategy is where many hardware pitches fall apart. A16z acknowledges that hardware carries longer build cycles and thinner margins relative to SaaS, and has stated it will offer patient capital on 7-10 year horizons, but that patience is conditioned on confidence in physical execution: identified manufacturing partners, access to fabrication capacity, which remains highly concentrated globally, and a credible plan for advanced packaging. The firm has also stated it will offer its U.S. manufacturing network to portfolio companies, which is a non-trivial resource given current fabrication constraints.
Business Model and Go-to-Market
Technology is necessary but not sufficient. Investors evaluate whether a hardware business can convert a technical lead into durable revenue. For AI hardware, that often means layering usage-based or infrastructure-as-a-service pricing over the hardware itself, or targeting enterprise buyers with a clear total-cost-of-ownership argument.
Early customer validation carries significant weight: pilot programmes with named enterprise clients, letters of intent, or disclosed contract values all reduce the market risk in a hardware pitch. Vertical focus sharpens the case further. Founders targeting healthcare, logistics or industrial automation with deep domain knowledge can demonstrate product-market fit more concretely than those pursuing a horizontal play. Proprietary datasets that improve hardware performance in a specific vertical also create defensibility that is harder to replicate than the hardware alone. A16z has stated it will provide go-to-market support, talent resources and broader network access to hardware portfolio companies, a meaningful commitment given how difficult enterprise hardware sales cycles are to break into from a standing start.
Pitching for Strategic Fit
The Machine Age Fund is not a general hardware vehicle. It has a stated thesis: accelerating AI’s physical layer is a social and national imperative, consistent with a16z’s broader “American Dynamism” investment theme. Founders who frame their pitch around compute efficiency, supply chain resilience or U.S. manufacturing capability are speaking the fund’s language directly.
A16z’s hardware portfolio already includes Skydio, Anduril and Mind Robotics, companies at the intersection of hardware and defence or industrial autonomy. A16z’s own investment in Volta, an AI cloud infrastructure provider that raised $300 million in early-stage VC in August 2026 with a16z co-leading alongside Altimeter Capital, illustrates the fund’s range. Elsewhere in the sector, Starcloud, a space-based data centre startup, raised a separate $250 million round the same month, though that round was led by Manhattan West with Nvidia and Cisco Investments participating, not a16z. The broader AI hardware investment wave gives this kind of capital a tailwind, but that also means more competition for the same pool of deep-tech founders. A pitch that connects specific IP to a specific infrastructure gap, backed by a team with credible manufacturing experience, is what separates fundable hardware companies from the rest of the field.
