The Next AI Race Will Be Physical 

(And Armenia Is Already on the Map)

In 2018, when my firm began investing in artificial intelligence, it was still the future. Models were clever but narrow, training cycles were slow, and “generative” was a term more common in art school than in tech. Back then, the bet was on the inevitability of AI eating more and more of the software stack—and it paid off. Today, AI is everywhere.

But ubiquity changes the game. When something becomes table stakes, the edge moves. For AI, the next decisive frontier is not in abstract intelligence but in embodied intelligence—the ability to sense, compute, and act in the physical world. I call this the era of physical AI.

That’s where the bottlenecks are now. A large language model can write an essay, but it can’t navigate a drone through GPS-denied airspace or guide a robotic arm in a chaotic factory floor. These problems require tight integration of sensors, custom silicon, and decision-making algorithms—all in real time. They demand hardware.

For a decade, venture capital recoiled from capital-intensive hardware. That era is ending. Defense contracts are underwriting early prototypes; national security concerns have shrunk perceived risk; and the market appetite is there. Companies building photonic chips, advanced packaging, and autonomous systems are closing nine- and ten-figure rounds. We’re seeing a rebalancing from bits back to atoms.

This hardware comeback isn’t just a business cycle, it’s geopolitical. The war in Ukraine, tensions in the Taiwan Strait, and the weaponization of supply chains have pushed governments to rebuild domestic manufacturing and diversify sources of critical technology. In this new world, countries and companies that can deliver integrated physical AI systems have strategic leverage far beyond their GDP ranking.

Which brings me to Armenia.

If you don’t work in semiconductors or AI, you might not know that Armenia has one of the densest concentrations of deep-tech talent anywhere. Not in sheer numbers—its population is under three million—but in per-capita output and in global connectedness. Armenia hosts the largest Synopsys R&D center outside the United States, alongside major footprints for AMD, Siemens and Cisco. Earlier this year, Nvidia chose Armenia for a new high-performance computing center that will run on its latest Blackwell GPUs, an asset so coveted that engineers from Silicon Valley have discussed relocating to get access.

This is not an accident of geopolitics or a sentimental Soviet legacy. Armenia’s engineering culture predates the Cold War and was never just a hand-me-down from Moscow. The country’s talent base has been shaped by rigorous math education, a diasporic network deeply embedded in global tech, and a startup culture that is default-international from day one. An Armenian founder will often incorporate in Delaware before hiring their first employee, not because they are chasing foreign glamour, but because the local market is too small to be the primary target.

The results show up in the scoreboard. By a consistent methodology—counting both Armenia-born companies and those with a major Armenian operational base—we have four unicorns: Picsart, ServiceTitan, Digitain, and SoftConstruct. That’s more than one unicorn per million citizens, putting us in rare company worldwide. And there’s a pipeline of “soonicorns” on track to join them. For comparison, India would need more than a thousand unicorns to match that ratio.

What’s most important is that these wins feed back into the ecosystem. Founders are angel-investing locally, diaspora investors are taking fresh interest, and the talent base is compounding. With physical AI and advanced hardware poised to dominate the next decade, Armenia’s mix of deep R&D, cross-border capital, and geopolitical relevance makes it a natural node in the emerging global network.

In our own portfolio, we’re already seeing the shape of what’s coming. Robotics companies marrying multimodal models to custom chips. Sensor platforms for autonomous systems that must operate in contested environments. Brain–computer interface research that, in the long run, could eliminate the need for screens entirely. Each of these domains requires exactly the kind of multidisciplinary expertise—in math, physics, embedded systems, and AI—that Armenia produces in abundance.

This is not to romanticize the challenges. No country of Armenia’s size can scale manufacturing alone. Supply-chain resilience will require partnerships—with Europe, the United States, and Asia—and deliberate integration into larger markets. Infrastructure must keep pace with ambition: the Nvidia center is a leap forward, but to be a true hub, we need redundant energy supply, export-friendly regulation, and a capital market deep enough to finance scaling, not just seed rounds.

The upside is clear. Physical AI will define the next industrial wave. The ability to deploy AI into the physical world — safely, securely, and at scale—will determine not only which companies win, but which countries retain technological sovereignty. Armenia has already proven it can punch far above its demographic weight in software. With the right investment and partnerships, it can do the same in hardware.

For investors, this is both a diversification play and an arbitrage. The big-name AI clusters like Silicon Valley, Shenzhen, Tel Aviv are all saturated and expensive. Armenia offers an underpriced, underrecognized alternative, plugged into those same networks, but with room to grow. For policymakers, it’s a chance to anchor a trusted partner in a strategically vital region with technology that matters.

In 2018, betting on AI meant believing it would one day be everywhere. That day has arrived. The next bet—the one we are making now—is on the systems that will take AI off the screen and into the world. The race is on, and Armenia is already running.