Analysts sounded the alarm: the AI bubble is expanding, and financing risks are becoming the biggest concern

The Zhitong Finance App learned that Tech Contrarians analyst Sara Awad said that the artificial intelligence (AI) investment boom is showing signs of a widening bubble, and increasingly innovative financing structures have raised questions in the market: how much capital can we continue to invest until there are no more clear returns?

Awad said bubble risk is more prominent in the AI capital expenditure field than in the broader semiconductor sector. She pointed out that there is a growing disconnect between the money invested in AI infrastructure and the return on those investments.

“As far as the bubble problem is concerned, I think the bubble does exist and is still expanding,” Awad said. She believes that the greater risk is focused on capital expenditure in the AI sector and how companies can finance these construction projects.

Awad pointed out that capital flows in the AI ecosystem are increasingly characterized by cycles. As an example, she said that Nvidia (NVDA.US) had considered providing financial support of about 250 billion US dollars for OpenAI and negotiated up to 350 billion US dollars to equip data centers with chips. In addition, Nvidia also invested in so-called “new cloud” companies including Nebius and CoreWeave.

She said there are concerns that chipmakers may increasingly finance customers who buy their products. “It looks like Nvidia is increasingly funding its customers,” Awad said, and believes this structure raises the question: how much additional capital can be invested in AI infrastructure construction without a substantial return on investment?

Awad also pointed out that the financial situation of big tech companies is also under pressure. Google and Amazon's free cash flow has turned negative, and Meta's discussions about possible leasing excess computing power capacity have raised questions about whether the company is spending too aggressively on AI infrastructure.

Another new trend that has attracted Awad's attention is that Nvidia is cooperating with six major Wall Street banks to set up independent computing power financing platforms, and plans to gradually mobilize more than 500 billion US dollars of third-party capital for AI infrastructure construction. The partnership aims to provide funding to Nvidia customers at preferential interest rates, and Nvidia CEO Hwang In-hoon once described chips as an investable asset class.

Awad said that the planned AI infrastructure spending is huge, making the issue of financing increasingly important. She pointed out that AI capital expenditure has now exceeded $740 billion, and at the same time, the cost of using AI models is declining.

She also mentioned that increasing competition from Chinese AI models has become another pressure point. Jefferies's report shows that the average price of AI inference services has dropped from $2.04 to $1.45 per million tokens in late May to mid-July to $1.16 to $1.18 in August. Price declines suggest that the AI industry is becoming more cost sensitive, even as infrastructure spending remains high.

Thus, according to Awad, the potential burst of the bubble is not so much driven by a single event as by the sustainability of financing behind AI infrastructure construction. She said, “I think the key to the bursting of the bubble is how we start raising capital for AI infrastructure, from free cash flow to credit... and this has brought us into a more risky field.”

In other words, the real risk is not whether AI technology itself has a future, but rather where the money that supports this big gamble comes from and whether it can be paid back. AI is certainly a revolutionary technological direction, but when capital games replace technology itself as the main line of the story, the pricing logic of the market needs to be re-examined. And that day may not be far off.