The heads of two chip giants issued warnings on the same day in Singapore: the biggest bottleneck for the AI industry has shifted from "being unable to make chips" to "being unable to borrow money."

On October 12, Broadcom (AVGO) CEO Hock Tan and Intel (INTC) CEO Lip-Bu Tan appeared together at the DBS Asian Future Frontiers Conference in Singapore.

Hock Tan said: "I think the biggest constraint factor is capital." He pointed out that over the past two years the AI industry has encountered supply bottlenecks one after another 鈥?memory shortages, insufficient wafers, power constraints 鈥?but each was overcome. However, the nature of the capital problem is different. He warned that as AI infrastructure construction continues to expand, financing difficulty may exceed any kind of hardware shortage.

Lip-Bu Tan immediately concurred, listing rising capital costs and ROI uncertainty as the most core challenges currently facing the AI industry.

"If there is any challenge ahead, it is capital and the cost of capital; on the other hand, in the long run, people need to see ROI."

The two men's remarks quickly triggered market reaction. Broadcom's stock fell 1.88% that day, while Intel dropped 3.20%.

Musk: "Don't cry in the casino"

Broadcom CEO Hock Tan made another even sharper remark at the conference.

According to the latest Bloomberg report, he said: "Many investors simply do not know what they are investing in." He compared the behavior of some AI investors to betting in a casino.

These remarks quickly went viral on social media. After Musk reposted the related post on the X platform, he left only five words as a response:

"No crying in the casino."

Compute demand remains strong, but the supply-demand gap will persist until 2030

Despite the warnings, both CEOs remain optimistic about near-term demand.

Hock Tan said: "Demand for compute infrastructure exceeds what we can supply." He expects this supply-demand imbalance to continue until 2029 and even extend into 2030.

Lip-Bu Tan said that in the next two to three years, the AI industry is unlikely to see a significant slowdown, with the continued adoption of AI applications and growing compute demand as the main support.

But he also drew a warning line: longer-term growth will depend on whether enterprises can truly implement AI applications, achieve efficiency improvements, and generate quantifiable economic benefits.

Under competitive pressure from Chinese models, ROI pressure intensifies further

Lip-Bu Tan also mentioned another variable: competition from China.

He said that Chinese frontier AI models have lower token costs and are spreading "very quickly." Models such as DeepSeek, Kimi, and GLM are already "good enough" to meet the needs of some use cases.

This directly challenges the industry assumption that "the most expensive systems will inevitably dominate the market," and it also puts greater ROI pressure on investors betting on high-end compute infrastructure.

But Lip-Bu Tan said, "I firmly believe that open source is the best way to drive efficiency, and the market will ultimately decide which models are most suitable for different application scenarios."

Broadcom and Intel are still raising large amounts of capital

Even as the two CEOs issued warnings, both companies themselves are actively seeking capital.

According to earlier reports, Broadcom is seeking to raise more than $50 billion from investors including Apollo Global Management and Blackstone to customize AI chips for OpenAI and support multi-gigawatt-scale compute deployment, with an initial agreement expected to be completed before the end of the year.

Intel, meanwhile, completed an upsized public offering in August this year, issuing 210.5 million shares at $95 per share and raising about $20 billion, higher than the initially planned $15 billion.