[Digital Today intern reporter Seung-ah Yoo] Elon Musk (일론 머스크) and Jensen Huang (젠슨 황), Nvidia's chief executive, offered differing outlooks on neo-cloud providers, a key business in the AI infrastructure market.

SCMP reported on Oct. 28 local time that Musk warned of large-scale failures and market reshuffling, while Huang predicted that funding strains would be resolved on the back of strong demand.

Neo-cloud providers are specialist cloud companies that rent out AI computing resources using high-performance graphics processing units. They must first secure expensive chips and data centres, but rental income arrives months to years later, creating heavy funding pressure.

Musk warned that companies without clear revenue sources beyond AI data centres are finding it harder to obtain loans, and that not many will survive among more than 200 neo-cloud providers. Huang said neo-cloud demand is "beyond imagination" and predicted that short-term funding shortages would also be resolved. Nvidia announced in August a plan to work with six financial institutions to support the raising of more than $500 billion in external funding over a long period.

Funding pressure is showing up even at leading companies. Bernstein Research estimated that U.S. neo-cloud firm CoreWeave may need an additional $18 billion to $42 billion to fulfil contracts worth about $129 billion. CoreWeave's second-quarter revenue more than doubled from a year earlier to $2.6 billion, but its net interest expense also rose to $640 million from $267 million. Loan maturities are about 5 years, while customer contracts average 3 years, raising worries that earnings could become uncertain after contracts end.

Financing conditions differ by company. Nebius recently secured a $3.5 billion bridge loan from Citibank and is also considering long-term fundraising using more than $40 billion in contracts with investment-grade customers. Morgan Stanley said that even if funding continues to be supplied, costs and terms will vary depending on borrowers' credit quality and contract conditions.

China's big tech companies are also not free from the investment burden. Alibaba Group, Tencent and Baidu spent a combined 131.9 billion yuan on capital expenditure in the second quarter, and all three reported negative free cash flow. China may have an advantage in power connections and permits. SemiAnalysis said China can build a standard 100-megawatt data centre in about 12 months. Advanced AI chip export controls, however, are an obstacle to securing equipment.

Experts are watching the possibility of overheating in AI infrastructure investment and companies' ability to generate actual profits. Morningstar forecast that heavy capital spending could continue until 2028. 86 Research said the key issue is whether AI spending by Chinese companies and consumers will grow enough to justify massive infrastructure investment. Constraints on capacity expansion in the AI supply chain could curb overinvestment, raising the possibility that the global market could make a soft landing rather than suffer a sharp slump.