a16z Podcast: AI Infrastructure Is Driving the Next Investment Cycle

SPEAKER_1 (0:01)

Let's dive into the latest state of markets presentation from A16Z. The team—David George, Sarah Wang, Alex Emerman, and Santiago Rodriguez—really unpacked where power, capital, and AI are colliding right now. And their first big takeaway is refreshing, honestly. They're arguing this isn't fantasy or hype. The market rally is actually supported by earnings, not just exuberance.

David George made a striking point: "The market is up about 20% while multiples are down 20%." That's a crucial distinction. So prices are rising because companies are actually performing, not because investors are irrationally optimistic? Exactly. That's why they say this looks very different from the dot-com era. But then the conversation shifts to something even more interesting—scale.

We're seeing a massive buildout in AI and infrastructure. Hyperscalers are pouring cash into chips, power, cooling, construction, and labor. Right, and they frame this as more than just a digital story. They say "We are entering a new age of atoms." It's not just about data centers anymore, it's about power grids, roads, water, transit—all the physical systems needed to support the digital economy. Which brings us to what they call "the next chapter of software eating the world." But now the real bottleneck is compute, not code. The five largest hyperscalers are projected to spend at extraordinary levels, and supply chains are already stretching into 2028

And they argue these assets have long useful lives, which justifies the scale of investment. What about enterprise AI adoption? Where does that stand? Still early. David George notes that 69% of S&P 500 companies have live deployments, but only 2% have integrated AI in ways that produce measurable long-term impact. Most are just experimenting with tools like Copilot. So the bigger opportunity is helping models understand a specific company, not just the world at large. And falling costs are making that more practical, right?

Absolutely. As model prices fall, agents can do longer, more complex work—checking their own output, using tools repeatedly. Companies like ServiceNow and Databricks are already seeing major savings. Better routing, caching, and fine-tuning are making AI faster, cheaper, and more reliable. What about consumer AI? That's got to be even earlier stage.

It is. Only 2% of U.S. households pay for AI today, but retention patterns are strong. These products often show "smile curves" in engagement—they improve over time as they learn and become more useful. So where's capital flowing next?

AI-related companies now account for 86% of U.S. VC deal activity, but the opportunity is broader. Consumer agents, robotics, autonomy, bio-pharma, personal health, and defense. Their bottom line? This wave may be remaking not just software, but the entire industrial economy.