In Singapore, the leaders of Intel and Broadcom have raised concerns that capital access may pose a significant challenge to the ongoing expansion of the artificial intelligence (AI) sector, even as the demand for computing infrastructure continues to surpass available supply.
During the DBS Asia Future Forward Conference, Broadcom's CEO Hock Tan emphasized that securing financing might become a more pressing limit on AI development than the actual shortages of chips or power supply. He stated, “What I think might be the biggest constraint is capital,” highlighting the urgent need for infrastructure to support increasingly advanced AI models.
Echoing Tan's sentiments, Intel's CEO Lip-Bu Tan noted that the rising cost of capital poses a challenge, as there are lingering doubts about whether heavy investments in AI will yield sufficient returns. "If there's any challenge in the front, it's capital and the cost of capital," he remarked, adding that stakeholders are eager to see a return on investment (ROI) over the long term.
Their comments underscore a growing concern for the AI industry, which has invested heavily in data centers, advanced chip production, and power infrastructure in anticipation of escalating demand for computing power.
As interest rates rise, financing for tech firms heavily investing in AI infrastructure may become more complicated. Higher bond yields could drive up the costs associated with building data centers and acquiring chips, potentially placing financial strain on projects that require years to realize returns.
Hock Tan mentioned that while the industry has faced numerous supply chain bottlenecks over recent years, he expects these issues to start alleviating as production capabilities improve. “In the last two years, we have seen a series of bottlenecks that keep coming out, but we overcome it," he explained. He further indicated that wafer shortages might be resolved by 2028 or 2029, and that power supply could also improve with the introduction of new infrastructure.
Despite these financial challenges, both executives expressed optimism about the near-term demand for AI computing infrastructure. Hock Tan noted, “There is more demand for compute infrastructure than we can supply,” suggesting this imbalance might continue until at least 2029, possibly stretching into 2030.
Lip-Bu Tan from Intel also stated that the next two to three years are unlikely to see a significant downturn in growth, driven by the increasing adoption of AI technologies and the consequent demand for computing resources.
However, the executives cautioned that sustaining this growth in the long run will require rapid business adoption of AI and a demonstration of valuable economic benefits. Broadcom’s CEO warned that the need for computing infrastructure may diminish if expectations regarding the adoption of advanced AI models are not met or if their performance improvements start to plateau. Lip-Bu Tan further highlighted that businesses must showcase tangible results from AI investments, such as enhanced efficiency and novel business models.
