In the venture capital world, the word “bubble” usually serves as a warning shot—a signal to pull back before a market correction wipes out portfolios. But at the recent Fortune Brainstorm AI conference, two top investors argued that when it comes to artificial intelligence, a bubble might be exactly what the industry needs.
During a panel moderated by Fortune’s Allie Garfinkle, Kindred Ventures founder Steve Jang and Sapphire Ventures partner Cathy Gao tackled the question dominating Silicon Valley: Are we in an AI bubble? The answer was, in short, maybe, but that’s the wrong question to ask.
“I think it is a bubble, but bubbles are good for innovation,” said Jang. He argued that the term “bubble” is often just finance shorthand for a “new technology wave” that occurs every five, six, seven years. According to Jang, this market heat is functionally necessary: “You need a bubble in technology and startups … to not only attract the world’s best talent to work on a certain set of problems but you also need the capital to fund them.”
Jang pointed to the exodus of top engineers from stable roles at tech giants like Google, Meta, and Uber to launch startups as a “good signal” rather than a warning sign. While admitting that “bubbles popping are bad,” Jang suggested that as long as the media continues to question the market, it helps “release pressure” and keeps the ecosystem healthy.
Gao agreed that in certain pockets, “valuations have far outstripped any sort of fundamental” metrics. However, she cautioned against dismissing the trend entirely, noting that the current growth curves “far outstrip the growth curves of companies we’ve ever seen before,” making the total addressable market difficult to calculate. “I don’t think we have a good sense of how big some of these companies can ultimately become.”
The Investment Playbook: Infrastructure vs. Workflow
Beyond the macroeconomic debate, the panelists outlined divergent strategies for surviving the pop, whenever it comes. Jang emphasized that in a true technology wave, “the whole stack changes,” creating opportunities from the bottom up. He noted that Kindred Ventures is focusing heavily on “accelerating and modernizing the AI infrastructure,” including chips, GPU marketplaces, and specialized frontier models. He observed that despite new entrants, margins remain high for cloud and chip providers, giving them “pricing power on all of the application layer companies.”
Gao, who focuses primarily on the application layer, offered a stricter framework for survival. “Let’s get real: AI is no longer a differentiator,” Gao said. She warned that “AI for X” companies are vulnerable. Instead, she said she looks for companies transitioning from simple features to complex workflows that embed deeply into an enterprise.
“In the future, it’s just going to be a customer support workflow tool, and every company will be powered by AI,” Gao said. She argued that despite the volatility, “first-mover advantage is actually real” in the enterprise sector, citing the enduring dominance of Salesforce and Workday that dates back to the cloud era.
Heartbreak Ahead for Robotics
The conversation turned darker regarding the future of robotics. Jang offered a “spicy” prediction for the sector, warning that many current startups are building on “primitive models” roughly equivalent to the “GPT 3.5 phase” of robotics.
“A whole bunch of robotics startups … are going to have a lot of heartbreak when the models improve and they’ve built for something sort of in the past,” Jang predicted. He added that many consumer robotics companies will likely “fall by the wayside” or shut down because the societal and governmental adoption cycles will be too long for startups to survive.
“We’re all going to be using robots on a daily basis,” Jang said. “Our kids are going to be riding in robots in a daily basis. That area is super exciting. But think about all of the startups building humanoids.” They will have to prove that their humanoids won’t, say, fall down or screw up or be buggy. “It’s going to move around in your office or household or on the street even,” he stressed, noting that every part of the physical world is going to have to prepare for humanoid robots potentially malfunctioning. “Think about that. And that is a deep tech problem.”
Looking toward 2026, Gao offered her own counter-intuitive forecast: despite better models, selling into the enterprise is “going to be even more difficult.” She cited unresolved issues regarding trust and visibility as hurdles that the industry has yet to clear. “People are going to be more focused on trust and visibility, and we haven’t really solved that problem yet.”
