AI Bubble Is Real, Hong Kong Investors Say — But Great Companies Come Out of It

Only a quarter of business leaders have pushed even 40% of their AI pilots into production — and that gap between hype and utility, investors said at the Nov. 4 Cyberport Venture and Capital Forum in Hong Kong, is what really defines today's AI "bubble."

The word was not a whispered fear at the forum. It was a baseline for discussion, as venture capitalists moved past the marketing noise of .ai domain names and onto harder questions: monetization, scaling bottlenecks, and a geopolitical race in which China and the United States are locked in a contest for dominance.

Are investors living through a bubble?

For Ganesh Rengaswamy, founding managing partner at Quona Capital, the atmosphere feels hauntingly familiar — yet fundamentally different from the dot-com era.

"Yes, of course we are in a bubble, but great transformation also comes out of the bubble," Rengaswamy said.

Modern capital deployment features "a couple of more zeros," he noted, but AI's foundational stack — compute, cloud, models and tooling — remains in flux. The early internet stabilized its protocols and browsers quickly. AI is still searching for its definitive architecture.

"Human beings don't like that change; we don't like this rapid transformation. How do we break through that and find real value? It's tough, and I think we're all going through that experience," Rengaswamy added.

Christine Keung, an advisor at Capital6 and former Bridgewater associate, said it remains "unclear if the public markets can withstand" the massive private valuations assigned to hot companies like Anthropic. She described a "Cambrian explosion" of small businesses empowered by AI, set against a shrinking pool of "venture-backable" companies capable of returning the massive funds institutional investors now manage.

"Our funds are getting bigger, which means the hurdle in which we need to return those funds increases substantially. It is only a bubble if the markets can't absorb it… right now the bubble actually hasn't burst yet," Keung said.

Jeremy Dalton, an advisor to FOV Ventures, issued a blunt warning to founders: "Stop buying .ai domain names." He cited a recent Deloitte report showing only a quarter of business leaders have successfully moved 40% of their AI pilots into production — proof, in his view, that the hype-utility disconnect drives the current turbulence.

Michael Phillips, executive managing director at Alumni Ventures, echoed the need for resilience, emphasizing that "investing through the bubbles" is the only way to catch the next generation of breakout leaders. The focus, he and others agreed, has shifted to companies that leverage existing AI layers to create tangible profit and efficiency — not those simply burning capital on research.

What does the China-U.S. race change?

For the first time in modern tech history, the U.S. faces an unchallenged rival in China — and not just in large language models. Panelists framed the real battle as one for "physical AI," the integration of intelligence into manufacturing and robotics.

"China has always had an advantage with physical AI," Keung said, pointing to the country's lead in manufacturing.

China is also leading in open-source models, which panelists argued may democratize AI more than the proprietary ecosystems of Silicon Valley giants. The "DeepSeek moment" — a reference to recent Chinese AI breakthroughs — served as a reality check that the monopoly on high-end innovation has ended.

"China and the U.S. are much more polarized. The world is not able to follow a linear or user-driven technology evolution in this transformation," Rengaswamy said.

Keung saw the open-source rise as a positive force: "It opened up the market. It allowed for a view towards a path of democratization. My hope is that it will lead us on a positive path towards getting AI into the hands of every individual globally."

She compared the lack of international alignment on AI safety to earlier debates over human rights during peak globalization, and grounded today's labor disruption in the Industrial Revolution. "What's happening right now in our labor economies isn't new," she said. "It just feels scary because we're living in it."

How do the investors themselves use AI?

Less dramatic replacement, more daily pragmatism. Dalton said he holds "six to eight hour conversations with AI regularly… to code solutions. To do that would have cost hundreds of thousands of dollars without it." AI has also enabled 24-hour research cycles.

But skepticism has crept in. Rengaswamy said AI cut his team's "time to do work" by two-thirds — while tripling his own "time to review" as he filters AI-generated sludge.

"Every single deck, every single memo I look at, it takes me probably five minutes to know if it's done on AI or not. And that just turns me off," Rengaswamy said.

The bet these investors are making: the bubble hasn't burst, and the winners of the next cycle will be the ones monetizing AI's existing layers — before the architecture underneath them settles.