Singapore – CEOs of Intel and Broadcom warn that as the demand for computing infrastructure continues to exceed supply, accessing capital may become a major constraint in the AI boom.
Broadcom CEO Hock Tan ( Hock Tan ) stated that as the industry competes to build the infrastructure required to support increasingly complex models, financing may become a more significant obstacle to AI expansion than shortages in semiconductors, memory chips, or power.
"I think the biggest constraint might be capital," Hawk Tan said on the DBS Asia Future Forward Conference held in Singapore on Monday.
Intel CEO Chen Liwu ( Lip - Bu Tan ) also expressed concerns about financing, and pointed out the issue of whether companies that invest heavily in AI will ultimately be able to obtain sufficient returns, as well as the rising cost of capital.
"If there are any challenges ahead, they are capital and the cost of capital. In the longer term, what people want to see is the return on investment ( ROI )," he said.
The remarks of the two individuals highlight a new challenge faced by the AI industry. This industry has already invested substantial funds in data centers, advanced chips, and power infrastructure to meet the surge in demand for computing power.
As these warnings emerge, rising borrowing costs are threatening the difficulty for tech companies to finance investments in AI infrastructure. Higher bond yields may increase the financing costs for data center construction and chip procurement, and could put pressure on the economic viability of projects that may take several years to generate returns.
Hock Tan stated that over the past two years, the industry has repeatedly encountered various supply bottlenecks, but with the expansion of production capacity, some of these constraints are expected to gradually ease.
"In the past two years, we've seen a series of bottlenecks emerge, but we've overcome them each time," he said. "Today's storage shortage may not be the case next year."
He also stated that the wafer shortage may ease by 2028 or 2029, and with the commissioning of new infrastructure, power supply could also improve.
Despite financing concerns, the two executives remain optimistic about the short-term demand for computing infrastructure for AI.
"The demand for computing infrastructure is greater than what we are able to supply," said Hawk Tan, adding that this imbalance may continue until 2029 and possibly even into 2030.
Chen Liwu also stated that in the next two to three years, there is unlikely to be a significant slowdown in this industry, and he mentioned that the adoption of AI applications is increasing, as is the demand for computing power.
However, the two executives stated that whether growth can be maintained in the coming years will depend on the speed at which companies adopt AI and whether this technology can bring meaningful economic benefits.
Hock Tan warned that if the adoption of the advanced AI model falls short of expectations, or if its capacity improvement begins to slow down, the demand for computing infrastructure may also slow.
At the same time, Chen Liwu emphasized that the company will increasingly need to prove the tangible returns brought by the AI investment by improving efficiency and adopting new business models.
