Signals Inbox·August 30, 2026·AI Infrastructure

What is the Machine Age fund by a16z, exactly?

Machine Age is a16z's $1.1 billion bet on the physical side of AI: chips, memory, networking, data centers, power infrastructure and robotics, plus software when it directly changes how that stack is built or operated.

Q1What did a16z actually launch with the Machine Age Fund?

a16z has raised $1.1 billion for Machine Age, its first dedicated hardware-infrastructure fund and its clearest commitment yet to investing in the physical systems underneath AI.

The fund covers chips, memory, networking and storage, according to a16z's launch announcement. But the mandate keeps going beyond computer components. Data centers, robotics and home AI appliances are explicitly included, while the firm also talks about cooling, electrical equipment, materials and the real-estate buildout required to support increasingly dense AI systems.

The Wall Street Journal described Machine Age as a16z's first dedicated hardware-infrastructure fund. That description is useful, although "hardware" alone can give the wrong impression. We are looking at a fund that can invest from a new processor architecture all the way to the infrastructure needed to power thousands of those processors together.

Five senior partners signed the announcement: Ben Horowitz, Martin Casado, Raghu Raghuram, David Ulevitch and David George. They span a16z's Infrastructure, Growth and American Dynamism teams, which already gives us a clue about how broad Machine Age is supposed to be.

Q2Why is a16z launching the Machine Age Fund now?

a16z is launching Machine Age now because AI is starting to run into physical limits faster than the hardware and power industries can expand around it.

The numbers in a16z's own launch post are unusually concrete. The firm says compute density per rack has increased 28-fold between an Nvidia H100 rack and a Rubin-era rack. Rack power has moved from roughly 5 to 10 kilowatts to about 100 to 250 kilowatts for current high-density systems, and a16z expects 1-megawatt racks within three years. Large data-center developments are also moving from tens of megawatts toward hundreds of megawatts, with some campuses reaching gigawatt scale.

That creates problems well outside the GPU itself. More compute requires more memory bandwidth, faster interconnects, better cooling, transformers, switchgear, land and electricity. a16z says the hardware supply base is accustomed to increasing output by roughly 20% to 30% a year, while parts of AI demand now need much faster expansion.

Founders have noticed the same change. Hardware startups have grown from a small part of a16z's deal flow a couple of years ago to more than 20% currently, according to the firm. Creating a dedicated fund looks like a response to what a16z was already seeing in its inbox.

Q3What does a16z mean by the “Machine Age”?

The Machine Age is a16z's name for the physical buildout required to keep AI improving and to move AI from computers into machines.

We should take the word "machine" quite broadly here. A Machine Age company could make a processor, redesign the network connecting thousands of accelerators, build electrical equipment for AI data centers or put AI into industrial robots. Those businesses sit in different markets, but each one deals with a physical constraint created or intensified by AI.

The range becomes clearer when we map the mandate by layer.

Machine Age mandate by layer

This explains why Machine Age cannot be reduced to one market such as semiconductors or humanoid robotics. a16z is grouping together the bottlenecks that appear when AI needs more physical capacity.

Q4Is Machine Age really a hardware fund, or can software companies fit too?

Software can fit into Machine Age when the software directly changes how physical AI infrastructure is built, financed, controlled or used.

a16z itself talks about "computer infrastructure" and complete systems rather than restricting Machine Age to companies manufacturing physical components. That leaves room for systems software, cloud infrastructure and software-heavy platforms sitting directly on top of hardware.

Volta shows how wide that definition can become. The company combines cloud operations with project finance to make AI compute available to customers that cannot sign the same five-year infrastructure commitments as hyperscalers. a16z co-led Volta's $300 million Series A, alongside Altimeter, with Nvidia, Dell and Matter also participating.

So the practical test is fairly simple. A conventional SaaS application using an AI model sits outside the core Machine Age thesis. A software-heavy company that changes the economics or operation of a large AI computing system can sit right inside it.

Q5Is a16z actually changing strategy with Machine Age?

Machine Age gives a name and a dedicated budget to a shift that was already happening inside a16z.

The firm's history with physical technology goes back much further than this fund. a16z led Skydio's Series A in 2016, wrote its first check into Anduril in 2019 and participated in Waymo's 2020 financing. SpaceX has also been part of the portfolio.

The pace has picked up sharply lately. Before announcing Machine Age, a16z had already backed six companies that it now uses as examples of the strategy: Unconventional AI, Nexthop AI, Volta, Atoms, Heron Power and Mind Robotics. Together they cover alternative compute, networking, data-center infrastructure, electrical equipment and industrial automation.

The organization has been moving in the same direction. Martin Casado runs Infrastructure, David Ulevitch leads American Dynamism and David George leads Growth. Raghu Raghuram spent decades around VMware and data-center systems, while Guido Appenzeller previously served as CTO of Intel's Data Center Group. Machine Age pulls people who were already making related investments into one explicit hardware push.

So yes, this is a real strategic change, just not a sudden one. a16z had already started moving down the AI stack before it put the Machine Age name on the strategy.

Q6What kind of startups is a16z already backing for Machine Age?

The clearest guide to Machine Age is a16z's recent portfolio: the firm is backing companies wherever AI demand has created an expensive physical bottleneck.

The six companies a16z highlights are associated with more than $3.6 billion of disclosed recent equity financings. That is the size of the full financing rounds, not the amount invested by a16z itself, since individual a16z check sizes were generally not disclosed.

What stands out is how little overlap there is between the actual products. One company is rethinking computation, another is building networking equipment, another transformers, another data centers and two are pushing AI into physical industry. The common thread is the problem they solve rather than the product they sell.

Recent a16z investments associated with Machine Age

Q7Why is electricity suddenly part of an a16z AI fund?

Electricity is now part of a16z's AI thesis because power availability is becoming one of the things that decides how much compute can actually be deployed.

The International Energy Agency estimates that global data-center electricity consumption could more than double to around 945 terawatt-hours by 2030. That is slightly more electricity than Japan consumes today. Its base case has total data-center demand growing around 15% a year, while electricity consumption from accelerated servers, the machines most closely associated with AI workloads, grows around 30% annually.

The U.S. numbers are even more striking. The IEA expects data centers to account for almost half of the country's electricity-demand growth through 2030.

a16z is already putting money behind this bottleneck. Earlier this year, its American Dynamism fund co-led Heron Power's $140 million Series B with Breakthrough Energy Ventures. Heron is developing solid-state transformers and says customers have expressed interest covering more than 40 gigawatts of potential demand. We should be careful with that figure because expressed interest is not the same as contracted orders, but the scale explains why Heron decided to expand manufacturing so aggressively.

AI infrastructure these days increasingly looks like an electrical-engineering problem as much as a computing problem. Machine Age is designed around that reality.

Q8Why are networking and memory suddenly so important to AI?

Networking and memory are becoming expensive AI bottlenecks because faster processors stop being useful when data cannot reach them quickly enough.

Modern AI clusters connect enormous numbers of accelerators. As those clusters get bigger, more time and money goes into moving information between GPUs, racks and storage systems. a16z says networking bandwidth inside a rack has increased so quickly that conventional copper links are approaching physical limits.

Nexthop AI shows how valuable investors think that problem could become. The company raised a $500 million Series B earlier this year at a $4.2 billion valuation, with Lightspeed leading and a16z joining as a major investor. Nexthop builds networking systems aimed specifically at AI and hyperscale infrastructure.

Memory has a similar problem. More powerful accelerators keep increasing the amount of data that needs to be supplied quickly enough to keep them busy. That is why a16z explicitly calls for cheaper, higher-bandwidth memory across the memory hierarchy rather than treating memory as a boring commodity attached to the GPU.

This is the Machine Age logic in one line: value shifts toward whichever component becomes the next constraint.

Q9How important is robotics to the a16z Machine Age Fund?

Robotics is a major part of Machine Age, although the current mandate is much wider than robots alone.

The investment market around physical AI is moving extremely fast right now. Crunchbase counted $47.4 billion of global venture funding across 521 physical-AI deals in the first half of 2026. That was almost four times the $12 billion invested in the second half of 2025 and roughly 80% above the first half of 2025.

The money is still coming. Generalist, which builds general-purpose AI models for robots, raised another $200 million only two months after a previous $400 million round, according to Axios. Nvidia is also pushing deeper into robots, vehicles and drones; The Wall Street Journal recently reported that the company's physical-AI activities already generate around $10 billion a year.

a16z's own bets lean toward real industrial work. Mind Robotics was created around factory automation and has Rivian manufacturing environments available for training and deployment. Atoms is pursuing automation across areas including food production, mining and transportation.

Robotics fits naturally into Machine Age. Once AI leaves the screen, models need sensors, actuators, machines, factories and power. That creates another large physical stack for a16z to invest in.

Q10Will Machine Age actually finance data centers and real estate?

Machine Age can reach into data centers and real estate, but the current portfolio suggests a16z is looking for venture-scale technology and business models rather than simply owning buildings.

The launch announcement explicitly includes data centers and even mentions the cooling, materials, electrical and real-estate buildout around them. Still, financing a conventional data center is very different from venture investing. A single large campus can consume billions of dollars, much of it financed with debt rather than startup equity.

Volta gives us a better picture of what a venture-style opportunity can look like. The company combines data-center development, compute infrastructure, cloud operations and project finance. Its goal is to open capacity to AI startups that struggle to make the long commitments normally required to finance large facilities.

This is likely the useful boundary for Machine Age. a16z can invest surprisingly deep into the physical buildout when there is a technology platform, new financing model or structural advantage capable of producing venture-scale returns. Owning ordinary infrastructure for its yield is a different business.

Q11Is $1.1 billion actually a big fund for a16z?

At $1.1 billion, Machine Age is large enough to be a serious a16z strategy, although it is far smaller than the firm's biggest investment pools.

Earlier this year, a16z announced more than $15 billion across several strategies. That package included $1.7 billion for Infrastructure, $1.7 billion for Apps, $1.176 billion for American Dynamism, $700 million for Bio + Health, $6.75 billion for Growth and another $3 billion across other venture strategies.

Machine Age is therefore almost the same size as the entire American Dynamism vehicle and about two-thirds the size of Infrastructure. For a first dedicated hardware-infrastructure fund, that is substantial.

The Wall Street Journal reports that Machine Age sits within the broader capital a16z raised earlier rather than representing another $1.1 billion that should simply be added to the previously announced total.

Machine Age compared with a16z's other funds

Q12Is the market behind Machine Age actually big enough?

The market behind Machine Age is already enormous; the harder part is finding the pieces of that spending where a startup can capture exceptional margins and value.

Moody's currently expects the six largest U.S. hyperscalers to spend roughly $785 billion this year, after raising its earlier estimate of about $700 billion. The forecast covers Microsoft, Amazon, Meta, Alphabet, Oracle and CoreWeave and puts spending on course to move toward $1 trillion next year.

As seen above, Machine Age has $1.1 billion to invest. That is only around 0.14% of one year's projected hyperscaler capital expenditure. So a16z clearly cannot finance the AI buildout itself, and it does not need to.

The venture opportunity comes from owning equity in companies that become critical suppliers into that spending. If a new networking architecture, transformer design, compute platform or robotics system becomes standard infrastructure, the company selling it can grow much faster than the amount of venture capital originally invested.

There is already plenty of money flowing toward the category. The much harder question, and the one a16z now has to get right, is which bottlenecks become durable startup markets rather than temporary shortages eventually absorbed by Nvidia, Broadcom, Eaton, Vertiv, hyperscalers or other incumbents.

Q13What could make the Machine Age thesis go wrong?

Machine Age can be directionally right about AI infrastructure and still produce disappointing venture returns if a16z pays too much, backs technologies that take too long to mature or mistakes temporary shortages for permanent startup markets.

Capital intensity is the obvious problem. Several of the companies a16z is backing need hundreds of millions of dollars before their technology reaches wide commercial deployment. Hardware mistakes are also expensive. A software company can change a product quickly; a semiconductor architecture, transformer factory or large robotics platform can require years of engineering and manufacturing work before investors know whether the economics really work.

Customer concentration creates another risk. Many AI-infrastructure startups ultimately sell into a relatively small group of hyperscalers and giant data-center operators. Those customers can be enormous, but they also have enough purchasing power to squeeze suppliers, build technology internally or switch architectures.

Efficiency could change the demand curve as well. The IEA's Energy and AI work includes a high-efficiency scenario precisely because better models, chips and software can reduce the amount of infrastructure required for a given amount of AI work.

Moody's has also been increasingly explicit about the financial risk around today's buildout. Its recent research points to higher borrowing, weaker credit metrics and the possibility of overbuilding if AI revenue does not grow fast enough to justify the capital going in.

The physical buildout looks real. That does not mean every supplier wins. Machine Age gives a16z a huge opportunity set, but it also puts the firm into businesses where timing and capital discipline matter enormously.

Q14Is Machine Age replacing a16z Infrastructure or American Dynamism?

Machine Age currently sits across a16z's existing investment teams rather than replacing Infrastructure, American Dynamism or Growth.

The people running it make that fairly clear. Martin Casado comes from Infrastructure, David Ulevitch from American Dynamism, David George from Growth and Raghu Raghuram works across Growth and Infrastructure.

Their markets increasingly overlap anyway. A transformer company can be an American manufacturing investment and an AI-infrastructure investment at the same time. A networking startup naturally fits Infrastructure while also being central to Machine Age. An industrial robotics company can sit inside physical AI and American Dynamism.

Machine Age gives those investments a shared thesis and a dedicated capital pool. For founders, it also gives a16z a clearer front door for businesses that might previously have fallen awkwardly between several teams.

Q15So what is the Machine Age Fund by a16z, exactly?

The Machine Age Fund is a16z's dedicated bet that the next phase of AI will require a huge rebuild of the physical systems underneath it, from computing and networking to electricity, data centers and robots.

The fund makes much more sense once we stop treating "hardware" as a synonym for chips. a16z is following the constraints created by AI wherever they appear. Today that takes the firm into processors, memory, networking, electrical equipment, compute platforms, industrial machines and the infrastructure connecting all of them.

Our research also shows that the fund follows behavior already visible inside a16z. Hardware now represents more than 20% of the firm's startup deal flow, and its recent investments already cover most of the stack Machine Age is supposed to target. The dedicated fund turns those scattered bets into an official strategy.

In plain English, Machine Age is a16z's attempt to invest in the companies that will physically make more AI possible, and increasingly, in the machines that will let AI do useful work outside a computer screen.