Andreessen Horowitz has closed a $1.1 billion fund dedicated exclusively to hardware, signaling a major strategic bet on the physical infrastructure that powers artificial intelligence. The Machine Age Fund, announced on Friday, will focus on the entire physical layer of AI, from semiconductor chips to the buildings that house them. It marks one of the largest venture vehicles ever created for hardware-only investments and reflects a broader shift in how the technology industry views the constraints facing AI growth.
The fund's mandate is unusually broad. It covers chips, memory, networking, storage, and then extends further into complete systems. Those systems are defined loosely enough to stretch from data centers to robotics to AI appliances designed for the home. That breadth suggests a16z is not simply placing bets on individual components, but on the entire stack of physical technology that makes AI possible. The firm appears to be treating hardware not as a niche asset class, but as a core part of the AI investment thesis going forward.
Five senior partners
The launch is backed by five named partners: Ben Horowitz, Martin Casado, Raghu Raghuram, David Ulevitch, and David George. Having five partners attached to a single vehicle is highly unusual in venture capital. It indicates that the firm considers the thesis central to its future strategy rather than an experimental side project. Each of these investors brings deep technical and operational experience, ranging from enterprise infrastructure to cybersecurity and consumer technology. Their combined involvement is a strong signal that a16z intends to be a serious and sustained player in hardware investing.
The physics behind the fund
The case for the Machine Age Fund rests on physics rather than traditional market sizing. Compute density has climbed enormously in a short period. Between Nvidia's H100 generation and its upcoming Rubin racks, compute density has risen 28-fold. That jump has dramatic consequences for power consumption. A rack that once drew five to ten kilowatts now draws between 100 and 250 kilowatts. Within three years, that figure is expected to reach a megawatt per rack. This is not an incremental change; it is a transformation in how data centers need to be built, powered, and cooled.
Individual data centers are already moving from tens of megawatts to hundreds of megawatts. Some campuses are approaching gigawatt scale, which is roughly the output of a small nuclear power plant. These requirements are sending builders to banks for financing on an entirely different order of magnitude. The demand for physical infrastructure is now so large that it is reshaping the economics of data center construction, energy procurement, and supply chain management.
The bottleneck has shifted
For years, the biggest challenges in AI were algorithmic and software-driven. Model design, training data, and compute efficiency dominated the conversation. But the a16z team argues that the bottleneck has moved out of the models and into the world. Every layer of the AI stack is now hitting the limits of what the supply chain can produce, and in some places, the limits of physics and computer science themselves. Chips need to be manufactured, memory needs to be integrated, networks need to move massive amounts of data, and facilities need to dissipate enormous amounts of heat.
The shift can be seen in a16z's own deal flow. Hardware once accounted for only a marginal share of the deals the firm saw. Now it represents more than 20% of the pipeline. That is a striking change for a firm known primarily as a software and consumer internet investor. It reflects a broader recognition in the venture industry that the next wave of AI innovation will require substantial physical infrastructure investments, not just lines of code.
Hardware's venture challenge
Hardware has historically been a difficult sell in venture capital. The conventional wisdom is that hardware takes longer to develop, costs more to prototype, and scales less cleanly than software. Software companies can often reach global distribution with low incremental costs, while hardware companies face manufacturing, inventory, and logistics hurdles at every stage. But the fact that hardware now represents a fifth of a16z's pipeline suggests that these objections are no longer decisive. The sheer scale of AI-driven demand has made hardware projects more predictable and more attractive than in previous cycles.
The portfolio the fund builds on spans far more than semiconductors. It includes companies such as Unconventional AI, Nexthop, Volta, Atoms, and Mind Robotics, along with Skydio, SpaceX, Anduril, and Waymo. This list is revealing. Drones, launch vehicles, defense hardware, and autonomous vehicles all qualify under the fund's definition of systems. That places the fund a considerable distance from a conventional deeptech mandate, which might focus narrowly on scientific hardware or advanced materials. Instead, the fund appears to be targeting every meaningful area where physical infrastructure intersects with AI.
Token intensity and demand
On the demand side, the argument turns on token intensity. Both the volume of AI work and the amount of compute each unit of work consumes are rising by orders of magnitude. The firm puts that growth in triple digits. As AI models become more capable and more widely deployed, the number of tokens they process grows rapidly. Each token requires substantial compute, memory, and bandwidth. The result is an explosion in demand that cannot be satisfied by software optimization alone. It requires physical hardware, and that hardware requires physical infrastructure.
The specific areas of interest are narrower and more revealing. Memory and interconnect improvements are one priority. Power-efficient edge devices are another. And then there is the surrounding infrastructure: cooling, materials, electrical systems, and real estate. These are the unglamorous components that determine whether an AI data center can actually function at scale. They are also the components where the constraint has quietly settled. A rack full of accelerators that cannot be fed data fast enough is simply an expensive way to generate heat. The industry has spent the past two years discovering how often that is the actual limiting factor.
Infrastructure as a venture category
Real estate, power distribution, and cooling have not historically been venture capital categories. Venture investors typically look for high-growth, scalable technology companies, not utility-like assets. But the scale of AI infrastructure needs is changing that calculation. Siting has already become a binding constraint in many regions. According to recent reports, 63% of new capacity is now going to locations outside the five established data center hubs. This decentralization creates demand for new types of real estate, new power distribution models, and new cooling technologies.
The fund arrives on top of an already unusual year for a16z. In January, the firm announced more than $15 billion across new funds, including a $1.7 billion Infrastructure Fund 2 and a $1.18 billion American Dynamism Fund 2. How the Machine Age Fund relates to those vehicles has not been made clear. There is no word yet on limited partners, cheque sizes, or stage focus. What is visible, however, is the pattern of recent investments. A Series A into Netris, a company that automates the networking that slows down GPU clouds, points in the same direction as the Machine Age thesis. Networking is a critical part of the physical layer, and a16z clearly sees it as a place where value can be created.
The pitch, stripped of the language around it, is that the scarce thing has changed. For most of the past decade, the most valuable resources were talent and distribution. Companies with the best engineers and the strongest go-to-market strategies tended to win. Now, a16z is betting $1.1 billion that the scarce resources are transformers, substations, and thermal design. The physical world is once again the frontier of technological competition, and the Machine Age Fund is positioned to back the companies that build the infrastructure of that new era.
Source:TNW | Investors-funding News