Abstract:
While demand for computing infrastructure continues to outstrip supply, the CEOs of Intel and Broadcom have warned that access to capital may become a major constraint on the development of artificial intelligence. Broadcom CEO Chen Fuyang said that the entire industry is stepping up the construction of infrastructure to support increasingly complex AI large models. Financing issues may be a bigger obstacle to AI expansion than shortages of semiconductors, memory chips or electricity.
“I think the biggest constraint may be capital.” Chen Fuyang said on Monday at the DBS Asia Future Foresight Summit in Singapore.
Intel CEO Chen Liwu also expressed the same concerns about financing risks. He pointed out that the cost of capital continues to rise, and the market also has doubts about whether companies investing heavily in AI can ultimately generate sufficient returns.
"If there is a challenge in the future, it is capital and the cost of capital; from a longer-term perspective, the market wants to see a real return on investment (ROI)." He said.
The speeches of the two executives revealed a major emerging problem in the AI industry. In response to the skyrocketing demand for computing power, the industry has invested massive amounts of money in data centers, advanced chips, and power infrastructure.
The current rise in borrowing costs may cause financing problems for technology companies that have heavily invested in AI infrastructure. Higher bond yields will increase the capital costs of data center construction and chip procurement; such projects often take several years to generate revenue, and the commercial viability of the project may be under pressure.
Chen Fuyang said that in the past two years, the industry has encountered various supply bottlenecks one after another, but with the expansion of production capacity, some constraints are expected to be eased.
"In the past two years, various bottlenecks have emerged one after another, but we have always been able to overcome them one by one. Today's storage shortage may no longer exist in a year."
He predicts that the wafer shortage is expected to ease in 2028-2029; at the same time, as new infrastructure is gradually put into production, the power supply situation will also improve.
Although they are worried about the financing environment, the two executives are still optimistic about the short-term demand for AI computing infrastructure.
Chen Fuyang said: "The market demand for computing power infrastructure has exceeded the industry's supply capacity." He added that this supply and demand imbalance may continue until 2029, or even until 2030.
Intel's Chen Liwu also believes that driven by the popularity of AI applications and demand for computing power, the industry is unlikely to see a significant growth stall in the next two to three years.
But the two executives also mentioned that whether the industry can maintain high growth in a few years depends on how quickly companies implement AI and whether AI technology can create real economic benefits.
Broadcom's Chen Fuyang reminded: If the implementation of advanced AI models does not meet expectations, or the model performance improvement reaches a plateau, the demand for computing infrastructure may cool down.
Chen Liwu emphasized that enterprises increasingly need to prove that AI investment can bring tangible returns through efficiency improvements and new business models.
Comments