
The Philadelphia Semiconductor Index fell for 8 consecutive days, and CoreWeave, as a vulnerable link in the AI ecosystem, faced market selling pressure.
AI-generated summary
The financial reports of very large cloud service companies show delays in data center plans and slowdown in capital expenditure growth. As an AI infrastructure provider, CoreWeave's operations are highly dependent on debt financing and revolving arrangements with customers such as Huida.
Philadelphia Semiconductor fell for eight consecutive trading days, setting a record for the longest decline since October 2008. CoreWeave, a new cloud service company (neoclouds) known as "Huida's son", was one of the fuses that triggered this wave of panic selling in artificial intelligence (AI) stocks.
Reuters reported that the stock price of CoreWeave, in which Huida holds 10% of the shares and has a market value of US$39 billion, began to plummet after the third-quarter financial report of the ultra-large cloud service company hinted that data center plans will be delayed and capital expenditure growth will slow down.
Worried investors point to new cloud companies such as CoreWeave, arguing that they are the most vulnerable link in the AI ecosystem because of their massive debt borrowing to finance the chips, land and energy needed for data centers that will only generate revenue in the future.
CoreWeave shares have plunged nearly 25% over the past week, while the yield on its bonds maturing in 2030 jumped 500 basis points to nearly 16%.
Although CoreWeave CEO Mike Intrator told employees that the market can sometimes "misunderstand and react irrationally," the company still has the advantage of strong demand for AI. However, investors are concerned about its $50 billion debt and its deep revolving financing arrangements between Huida and customers such as Meta and OpenAI.
Jame Athey, fixed income portfolio manager at Marlborough, said, "The problem is the huge interconnectedness, not only the circular financing, but also the fact that these assets are all staked in the same place and the same portfolio by the same investors. The leverage effect always turns small things into storms and ultimately into bigger disasters." He was referring to the broader uncertainty of AI.
Companies such as CoreWeave, Nebius, Lambda Labs, Applied Digital, Iren, Crusoe and TeraWulf are building energy-intensive data centers and then leasing them to companies developing and operating AI models. Before the leasing business can be launched, it requires huge upfront investment and almost no income. According to data compiled by LSEG, CoreWeave has never made a profit.
CoreWeave stressed that the amount of unfulfilled deals it had struck was much higher than its debt. Its unfulfilled transactions in the second quarter amounted to US$104.2 billion, and at the beginning of the third quarter, it had secured US$25 billion in new order commitments.
However, Coreweave is burning money too fast. Last year, its annual revenue was US$5.1 billion and its net loss was US$1.2 billion. Although revenue is expected to double this year, capital expenditures are expected to quadruple to between US$35 billion and US$39 billion, indicating that more funds will be burned this year.
Tommaso Leodari, chief investment officer of hedge fund Index & Cie, said that Coreweave's revenue growth relies on debt financing rather than equity financing. Therefore, if there are problems with the use of its assets, such as a slowdown in data center construction, it will cause debt repayment problems and affect investor confidence.

Taiwan stocks plummeted last Thursday, with foreign investors selling more than 302.5 billion yuan in a single day, hitting a record high. However, financial stocks bucked the trend and strengthened. The financial index closed at 3,750.52 points last Friday, a record high, with a cumulative increase of 56.64% this year. Analysts pointed out that financial holdings have stable profits and defensive attributes, making them a safe haven for funds. The stock prices of many financial stocks such as Fubon Gold, Cathay Gold, and CITIC Gold have reached new highs.

The demand for AI servers has stimulated a shortage of PCB upstream materials. Key materials such as fiberglass cloth and copper foil continue to rise in price. PTFE is regarded as a potential alternative material. Taihong, Taijun, and Yaden have entered the PTFE field. Stock prices have risen against the trend recently, and supply constraints are expected to continue until 2027.

U.S. Treasury data showed that China's holdings of U.S. government bonds fell to $590.7 billion in August from $618 billion in July, the lowest since August 2008. Japan's selling of U.S. debt reduced its holdings from $1.1 trillion to $1.07 trillion; the United Kingdom and Canada increased their holdings of U.S. debt, partially offsetting the selling pressure from Asian countries. At the same time, China continues to increase its gold reserves, which increased to 7.747 million troy ounces in September, increasing for 23 consecutive months.

Huang Renxun, CEO of Huida, said at the Berlin event that the AI revolution will continue and has just begun, and will have a positive impact on various industries and countries. It is expected that huge returns on hardware investment will alleviate market concerns about the sustainability of AI spending. Although AI stocks, semiconductor and memory stocks have fallen sharply recently, the Philadelphia Semiconductor Index has recorded its longest losing streak since 2008, and CoreWeave stocks and bonds have both declined.

EU leaders met at a Brussels summit to discuss trade policies to deal with a surge in Chinese exports. A survey by the Forsa Institute shows that more than a quarter of German industrial companies have lost competition to Chinese suppliers, but most companies still regard the Chinese market as an indispensable sales and procurement base.

The 2026 Pearl River International Trade Forum was held at the Canton Fair Hall in Guangzhou. The forum focused on changes in the global economy and trade, and discussed the transformation of Chinese enterprises from "exporting products overseas" to "exporting capabilities overseas" as well as the "native globalization" model in the AI era, aiming to promote high-level opening up and international economic and trade mutual trust.