Andreessen Horowitz (a16z) is fundamentally reorienting its AI investment strategy away from pure software and toward physical infrastructure, defense, and national security applications. This shift reflects a growing recognition across venture capital that the real constraint on AI development is no longer model capability, but rather electricity, semiconductors, and the ability to deploy intelligence into the physical world.

Jess Leão, who recently joined a16z as a Partner on the American Dynamism team, articulated this thesis in a major newsletter relaunch. The American Dynamism team focuses on backing founders building in the national interest across aerospace, defense, public safety, education, housing, semiconductors, supply chain, industrials, and manufacturing. This represents a deliberate departure from the venture capital industry's historical focus on software-first, venture-scale businesses.

Why Is Physical AI Becoming the New Frontier?

The binding constraint on artificial intelligence today is not algorithmic innovation, but rather infrastructure. Leão noted that electricity, megawatts, transformer capacity, and interconnection queues are now the decisive factors determining which companies will lead the AI race. These decisions are being made not in Silicon Valley boardrooms, but in capacity auctions, permitting hearings, and county board meetings about data center expansion.

The implications are profound. AI is climbing out of the chat interface entirely and into robots, autonomous vehicles, and factories that learn. Manufacturing, energy, and robotics are no longer separate from AI development; they represent the same intelligence pushed further into the physical world. As Leão explained, "The real prize was never automating your software workflows. The entire physical world is up for grabs".

What Specific Challenges Are Driving This Shift?

Several concrete developments underscore why a16z and other venture firms are pivoting toward deep tech and physical infrastructure. In July 2026 alone, the industry faced multiple critical bottlenecks:

• Energy Constraints: PJM's capacity auction came up 6,831 megawatts short of demand, with prices hitting the Federal Energy Regulatory Commission (FERC) cap at $16.4 billion total, yet still insufficient to meet supply needs.
• Regulatory Barriers: New York became the first state to implement a statewide moratorium on new data centers, with polling showing 2-to-1 public support for the ban, directly constraining AI infrastructure expansion.
• Nuclear Energy Solutions: The Department of Energy's Reactor Pilot Program achieved a major milestone when four reactors reached criticality by July 4, 2026, beating the White House's own deadline and signaling a potential path to solving AI's electricity demands.

These developments illustrate that venture capital's traditional playbook of backing software-first startups is increasingly misaligned with the actual constraints on AI deployment. The companies that will dominate the next decade are those solving infrastructure, energy, and manufacturing challenges, not those optimizing model inference speed.

How Is a16z Positioning Itself in This New Landscape?

a16z's American Dynamism thesis, articulated in 2022, was contrarian at the time. The firm argued that the most important companies would be built in the national interest, in atoms as much as bits. That thesis now looks prescient. Leão noted that "a thesis that was contrarian when a16z articulated it in 2022 now looks a lot like where the whole industry is headed".

The shift reflects a broader recognition that the AI race is increasingly a contest between the United States and China, pulling in chip fabrication, export controls, rare earth minerals, and defense applications where autonomy is changing how wars are fought in real time. Venture capital is following suit, with more and more funds recognizing what deep tech means for the next decade.

This reorientation also signals a maturation of the venture capital industry's understanding of AI's true economic impact. Rather than chasing the next Instagram-style social media application powered by AI, firms like a16z are backing founders building shipyards, reactors, semiconductor fabs, and autonomous systems that will determine which nation leads the next generation of economic and military power.

Steps to Understanding a16z's New Investment Thesis

• Recognize the Infrastructure Bottleneck: AI's primary constraint is no longer model capability but electricity supply, semiconductor availability, and physical infrastructure capacity, making deep tech investments more valuable than pure software plays.
• Track National Security Implications: a16z's focus on defense, aerospace, semiconductors, and supply chain reflects the reality that AI development is now inseparable from geopolitical competition between the US and China.
• Monitor Regulatory Developments: State-level bans on data centers and federal initiatives like the Reactor Pilot Program are reshaping where and how AI infrastructure can be built, making regulatory expertise a competitive advantage for venture firms.

The venture capital industry's pivot toward deep tech and physical AI represents a fundamental recalibration of where the next trillion-dollar companies will be built. a16z's American Dynamism team is betting that the future belongs not to the next great AI model, but to the companies that solve the physical, infrastructure, and national security challenges that make AI deployment possible at scale.