a16z Launches $1.1B Machine Age Fund for AI Infrastructure
TL;DR
Andreessen Horowitz has launched the Machine Age Fund, a $1.1 billion vehicle dedicated entirely to the physical infrastructure of AI: chips, memory, networking, storage, power systems, data centers, and robotics. Five general partners are named on the announcement — Ben Horowitz, Martin Casado, Raghu Raghuram, David Ulevitch, and David George — and the firm frames the fund around a simple thesis: the binding constraint on AI progress has shifted from model quality to the physical capacity to build and power the machines that run it. It matters because it's the clearest signal yet that the largest pool of venture capital in the industry believes the next multi-year investment cycle in AI runs through silicon, steel, and megawatts, not just software.
Key Takeaways
The bottleneck has moved from algorithms to physics. a16z's own framing leans on hard infrastructure numbers: a roughly 28x jump in compute density moving from H100 to Rubin-class racks, power draw per rack climbing from 5-10kW toward 250kW and eventually approaching a full megawatt, and data center footprints scaling from tens of megawatts to gigawatt campuses. That's not a software roadmap, it's a utility-scale engineering problem, and it's where a16z is now pointing a nine-figure-plus check size.
Five GPs, not one champion, are attached to this fund. Fund launches anchored to a single partner's pet thesis are common; naming five sitting GPs, including co-founder Ben Horowitz himself, signals this is a firm-wide capital allocation decision rather than a side bet. That matters for founders evaluating whether a fund's stated thesis will survive partner turnover or a change in market sentiment.
The stack bet is vertically broad, which cuts both ways. Machine Age spans chips and memory at one end and home AI appliances at the other, with data centers and robotics in between. Broad mandates let a16z chase the best deal wherever it appears in the stack, but they also mean founders in any one sub-vertical are competing with founders across five or six adjacent categories for the same pool of capital.
This lands on top of, not instead of, a16z's core funds. Machine Age follows the firm's roughly $15 billion raise earlier in 2026 across its flagship vehicles. Layering a dedicated infrastructure fund on top of that is a statement about pace: a16z is choosing to deploy capital into physical AI infrastructure faster than its general-purpose funds alone would allow.
Fund Overview
Fund Name: The Machine Age Fund
Fund Size: $1.1 billion
Stage: Primarily growth and infrastructure-stage, spanning early bets on emerging hardware categories through capital-intensive scale-ups
Check Size: Not disclosed publicly; the fund's mandate and target company profile (data centers, chip startups, robotics manufacturers) implies large individual checks, likely eight to nine figures
Geography: Global, with a US-centric deal pipeline consistent with a16z's broader portfolio
Focus: The physical infrastructure stack underneath AI — chips, memory, networking, storage, data centers, robotics, and AI-native hardware for the home
Key LPs: Not disclosed in the firm's announcement
Why This Fund Matters
For the last three years, the AI funding narrative has been dominated by model labs and application-layer startups. Machine Age is a bet that the more durable, less commoditized opportunity now sits one or two layers down the stack, in the physical systems that make frontier models trainable and inferable at scale. That's a defensible read: model weights get open-sourced or leapfrogged in a matter of months, but a gigawatt data center campus or a novel power-delivery architecture takes years to replicate and carries real barriers to entry — permitting, capital intensity, supply chain relationships, and engineering talent that isn't easily poached.
The specific numbers a16z is citing are worth taking seriously on their own terms. A 28x jump in rack-level compute density between generations, power draw pushing toward a full megawatt per rack, and data center buildouts scaling from tens of megawatts to gigawatt campuses describe an industry that is running into hard physical limits faster than most public commentary has acknowledged. Cooling, grid interconnects, and chip packaging are becoming as strategically important as the model architectures themselves.
It also reflects a broader shift among the largest multi-stage VCs: rather than betting purely on which application will win, capital is increasingly flowing to whoever controls scarce physical capacity, since that capacity is what every application layer startup ultimately depends on. Expect more of the largest funds to stand up dedicated infrastructure vehicles over the next 12-18 months as this thesis gets validated or challenged by results.
The risk sits in timing and capital intensity. Infrastructure bets of this kind require far more follow-on capital per company than a typical software portfolio, and the payback period is measured in years of utilization, not quarters of revenue growth. If AI demand growth decelerates even modestly, or if a wave of new chip architectures makes today's buildouts obsolete faster than expected, a fund this concentrated in physical capacity carries real downside that a diversified software portfolio would not.
The Team
Ben Horowitz is a co-founder and general partner of Andreessen Horowitz, and his attachment to this fund signals it sits close to the firm's core strategic priorities rather than being a satellite bet. Martin Casado leads much of a16z's infrastructure and enterprise investing and co-founded Nicira, giving him direct operating experience in the kind of deep infrastructure category this fund targets. Raghu Raghuram joined a16z as a general partner after serving as CEO of VMware, bringing enterprise infrastructure and data center operating experience that maps directly onto Machine Age's thesis. David Ulevitch, founder of OpenDNS, brings a security and networking lens relevant to the fund's networking and systems bets. David George has been central to a16z's growth-stage investing, which is likely to be the primary check-writing motion for a fund targeting capital-intensive, later-stage infrastructure companies.
Early Portfolio
a16z's own announcement names Unconventional AI, Nexthop, Volta, Atoms, Heron Power, and Mind Robotics as companies associated with this thesis. Heron Power and Volta point toward power electronics and grid infrastructure, Mind Robotics and Atoms toward physical robotics, and Nexthop toward networking — a spread that maps cleanly onto the fund's stated stack, from power delivery through networking to physical robotics deployment.
What This Means for Founders
This is a fund for founders building physical, capital-intensive infrastructure rather than pure software: chip design and packaging, power delivery and cooling systems for data centers, networking hardware built for AI-scale traffic, and robotics companies building the physical layer of automation. If your company's moat is measured in supply chain relationships, manufacturing capacity, or engineering complexity rather than iteration speed on a model, this is squarely the kind of company Machine Age is designed to back.
The value-add case is straightforward: access to a16z's balance sheet and its enterprise relationships across hyperscalers, chipmakers, and data center operators, which matters enormously for infrastructure startups that need design partners and anchor customers as much as they need capital. Founders should expect a fund at this scale to write large checks into fewer companies, so the bar for conviction is likely to be correspondingly high — this is not a spray-and-pray seed vehicle.
Fund Momentum Take
We think this is directionally correct and arguably late rather than early. The physical constraints a16z is citing — power, cooling, chip supply, data center capacity — have been visible to anyone tracking hyperscaler capex for the past two years. What's notable is that a firm best known for software-first, capital-light bets is now writing a $1.1 billion check into an asset class that looks a lot more like infrastructure private equity than classic venture. That's either a sign the line between VC and infrastructure investing is permanently blurring in the AI era, or a sign that traditional VC economics — asymmetric software returns on modest capital — are being stretched to justify a very different risk profile.
Our bet: the winners in this fund will be the companies solving the least glamorous parts of the stack — power delivery, cooling, and interconnect — rather than the flashier robotics and appliance plays, because those are the bottlenecks that are actually gating hyperscaler buildouts today. The risk worth watching is capital intensity outpacing demand: if enterprise AI adoption plateaus before this capacity comes fully online, a fund this concentrated in physical buildout will feel the drawdown first and hardest.
Frequently Asked Questions
What is the Machine Age Fund?
A $1.1 billion fund from Andreessen Horowitz dedicated to the physical infrastructure underlying AI, spanning chips, power, networking, data centers, and robotics.
Who are the general partners behind the fund?
Ben Horowitz, Martin Casado, Raghu Raghuram, David Ulevitch, and David George are the general partners named on the announcement.
How is this different from a16z's other funds?
It's a dedicated vehicle focused specifically on physical AI infrastructure rather than a general-purpose venture or growth fund, layered on top of a16z's roughly $15 billion raised earlier in 2026.
What kind of companies will this fund back?
Chip and memory companies, power and cooling systems for data centers, AI-scale networking hardware, robotics manufacturers, and AI-native hardware makers.
Are the fund's LPs public?
No, a16z has not disclosed limited partners for the Machine Age Fund.
Have a fund closing to announce? Submit your fund here.
Need help raising capital? Check out our Fundraising Advisory services.