According to Fortune, at their recent Brainstorm AI conference, top investors Steve Jang of Kindred Ventures and Cathy Gao of Sapphire Ventures tackled the Silicon Valley question of whether we’re in an AI bubble. Jang stated, “I think it is a bubble, but bubbles are good for innovation,” framing it as a necessary phase for a new technology wave that happens every five to seven years. He argued this heat is needed to attract the world’s best talent away from giants like Google and Meta and to fund their startups. Gao agreed valuations in some areas have “far outstripped” fundamentals but cautioned the growth curves are unprecedented, making the total market size hard to calculate. Both outlined strategies for surviving the eventual market correction, with Jang betting on AI infrastructure and Gao on complex enterprise workflows.
Bubbles Are The Engine
Here’s the thing: Jang’s argument flips the script. We’re trained to think of a bubble as pure froth, a sign of irrational excess that always ends in tears. But he’s basically saying, “Yeah, and?” His point is that you can’t get the monumental shift—the “whole stack” change he talks about—without that period of manic, maybe-irrational investment. It’s the venture capital version of “no pain, no gain.” The capital flood and the talent exodus are two sides of the same coin. You need the crazy money to pull engineers from their cushy FAANG jobs. Otherwise, why would they leave? So in a weird way, the media constantly asking “Is this a bubble?” is a pressure release valve. It keeps everyone just skeptical enough to maybe avoid a total catastrophe.
Survival Strategies Divide
Now, their survival blueprints are fascinating because they’re so different, and it shows where each investor is placing their bets. Jang is going deep on the picks-and-shovels play: chips, GPU marketplaces, frontier models. He’s not wrong about the pricing power. The companies providing the raw compute, whether it’s NVIDIA or the cloud giants, sit at the top of the food chain. Every app company is a customer. That’s a powerful place to be. But Gao’s take on the application layer is brutal and probably correct. “AI is no longer a differentiator.” Ouch. She’s saying the “AI for X” wrapper is dead on arrival. The winners will be the ones who stop talking about AI and start solving a messy, embedded workflow problem for a business. It’s a shift from feature to foundation. That’s a much harder thing to build, which is why she thinks first-mover advantage in enterprise software still matters.
The Robotics Reckoning
Jang’s “spicy” prediction on robotics is the real wake-up call, though. Think about it. He’s comparing most current robotics startups to the GPT-3.5 phase. That feels incredibly generous. The gap between a demo in a lab and a reliable, safe, affordable product in the physical world is a chasm. And he’s right about the adoption cycle. Governments, insurers, and public acceptance move at a glacial pace compared to software updates. A whole bunch of these humanoid robot companies are building beautiful, intricate solutions for a technological reality that doesn’t exist yet. When the foundational models do leap forward—the GPT-4 or GPT-5 moment for robotics—a lot of that intricate hardware and software will be instantly obsolete. It’s a recipe for heartbreak. The companies that survive will likely be the ones with the deepest pockets and the patience to wait out a decade-long adoption curve.
The Real 2026 Problem
So Gao’s final point about 2026 is the sleeper hit. Better models won’t solve the enterprise sales problem; they might make it harder. Why? Because as AI gets more powerful and embedded, the questions of trust, visibility, and accountability get louder, not quieter. If an AI is managing your customer support or your supply chain, you need to know *why* it made a decision. We haven’t solved explainable AI at scale. We also haven’t solved the data governance and security nightmares. The sales pitch moves from “Look at this cool feature” to “Trust us with the core operations of your business.” That’s a completely different, and much more difficult, conversation. The winners in the application layer won’t just have the best workflow tool. They’ll have built a fortress of trust around it. And honestly, how many startups are even thinking about that?
