Intel, Broadcom CEOs flag AI industry challenges at financial event
SiliconANGLE Maria Deutscher
Intel and Broadcom’s CEOs say AI’s biggest bottleneck may be money, not chips. They also warn that demand is still outrunning supply, and that AI investors are acting like it’s a casino.
Based on reporting by SiliconANGLE, Maria Deutscher — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Intel chief Lip-Bu Tan and Broadcom chief Hock Tan used a financial conference in Singapore to put a less glamorous problem at the center of the AI boom: capital. Not compute, not model quality. Money. Or more precisely, the cost of getting enough of it to keep buying the gear and paying for the training bills.
That message carries weight because these are not outside commentators. Intel’s CPUs are in high demand as AI agents run more of their tools, and the company said last month it can satisfy only 50% of orders. Broadcom sits even deeper in the stack, supplying network processors and custom AI accelerators, right where data center bottlenecks show up first.
Hock Tan said the chip shortage should start easing in 2028 or 2029, but he still sees the mismatch between AI demand and supply lasting into 2030. He also pointed to the cost of training large language models, focusing on OpenAI. Broadcom helped design OpenAI’s first custom inference chip earlier this year, but Tan said the company is still losing money because training is so expensive.
That is the awkward reality under all the hype. Enterprise revenue at AI labs still leans heavily on pilot projects, Tan said, and he compared that to kids burning through arcade tokens without learning much. He argued those labs need more customers who can actually get a return from inference spending, not just curious buyers trying out the toy before the next round of funding.
Lip-Bu Tan made a similar point from Intel’s side, saying the front-end problem is capital and the cost of capital, while the longer-term test is ROI. He also said open-source language models from Chinese labs are already “good enough” for many budget-conscious users, and he called open source the best way to drive efficiency. That line sounds especially sharp coming from a chip CEO: the market may love the frontier, but plenty of buyers are shopping at the cheaper end.
My take — AI-written commentary, not fact-checked reporting
This is the part of the AI story that keeps getting politely stepped around: the bill arrives before the victory lap. The open-source angle matters here because “good enough” is what usually crushes premium pricing, and the market still talks as if every model needs to be a shrine. It doesn’t; some of them just need to work and not torch the budget.
Read more about this at: SiliconANGLE
