“I think we’ve probably got everything wrong. It’s a good way, humble way to realize how little we really know about the technology we are creating.”

By Wolf Richter for WOLF STREET.

Broadcom CEO Hock Tan and Intel CEO Lip-Bu Tan said the quiet part out loud on a panel at the DBS Asia Future Forward Conference – the quiet part that everyone has been talking about but that has been kept assiduously quiet in the AI-financial-hype cycle: The business model of AI labs remains uncertain despite massive investments, how businesses will profit from AI remains uncertain, Return on Investment (ROI) on AI infrastructure remains uncertain, and financing all of this is the biggest constraint. And Hock Tan likened “a lot of investors” in AI to gamblers in a casino.

“A lot of investors have no clue what they are investing in,” Broadcom’s Hock Tan said. They “set very extreme expectations only to find out that it’s almost like they are in a casino.”

In terms of OpenAI, he said that its annualized $50 billion in revenues at the end of September doesn’t show that it has a sustainable, profitable business model as its huge expenses related to the costs of AI and the capital expenditures of the AI buildout are producing big losses.

“So frankly, it doesn’t tell anything at this point except that there is a lot of potential revenue growth out of providing frontier generative AI models,” he said.

And he said that AI model developers, such as OpenAI, need to find customers that can generate returns from putting AI to work, rather than relying on users who are now “wasting money” on tokens as they experiment with AI, according to Bloomberg reporting on the panel discussion. He said:

“When you are a kid you go to a video arcade, you play. Your parents give you $10 of tokens, you finish it in 10 minutes and you haven’t learned anything. That’s what’s happening in enterprises today.”

In terms of the AI infrastructure buildout boom, Hock Tan said: “My presumption is this might continue. This would probably continue till 2029 and probably into 2030. But that depends to a large extent on things we are learning as we go along in this process.”

Over the past two years, the AI industry has struggled with “a series of bottlenecks, that keep coming out, but then we overcome it,” Hock Tan said. “Memory shortage today might not be that way a year from now” as production capacity expands. Wafer shortages could ease by 2028 or 2029, and power availability might improve as new infrastructure comes online, he said, according to CNBC, reporting on the same panel discussion.

But financing the AI infrastructure needed to power the AI models “might be the biggest constraint,” Broadcom’s Hock Tan said.

Intel’s Lip-Bu Tan agreed. “If there’s any challenge in the front, it’s capital and the cost of capital,” he said, according to CNBC.

“And then the other part is in the longer run, people want to see the ROI,” he added.

Treasury yields have risen sharply this year, and the spreads between AI bonds and Treasury yields have widened substantially to where SpaceX’s 10-year senior unsecured investment-grade notes, that had been sold in June with a coupon interest rate of 5.875%, traded on Friday at a yield of 7.20%. And this is going to be a template for future sales of similar AI bonds. So the cost of funding this massive AI infrastructure buildout has increased substantially over the past few months, making the hope for ROI more precarious.

“There is more demand for compute infrastructure than we can supply,” and the imbalance could persist through 2029 and potentially into 2030, Broadcom’s Hock Tan said.

But whether that continues depends on how fast businesses adopt AI, and what kind of ROI businesses see on their AI investments, both CEOs agreed, according to The Straits Times reporting on the same panel discussion.

But if the deployment of advanced AI models disappoints or improvements in their capabilities begin to plateau, demand for computing infrastructure could slow, Broadcom’s Hock Tan added, according to CNBC.

And companies will need to demonstrate meaningful ROI from their AI investments through improved efficiency and new business models, Intel’s Lip-Bu Tan said.

If in the future, the CEOs were asked what they got wrong when looking back, Intel’s Lip-Bu Tan said if he got 70% right, he’d be a “damn good CEO.”

Broadcom’s Hock Tan said: “I think we’ve probably got everything wrong. It’s a good way, humble way to realize how little we really know about the technology we are creating.”

In case you missed it: Despite the AI Debt Pileup, Corporate Bond Market Still in La-La-Land, Investors Chasing Yield, Backed by Sky-High AI Stock Valuations

Enjoy reading WOLF STREET and want to support it? You can donate. I appreciate it immensely. Click on the mug to find out how: