[Weekpick Cloud] CoreWeave Shares and Bonds Plunge as Microsoft Withdraws Data Center Contract Over Power Delays
IT DAILY ·
✦ AI Summary
CoreWeave’s stock price and bond prices plunged amid debt burdens and concerns over a neocloud-driven credit crisis.
Microsoft withdrew a large-scale AI data center contract in the UK, citing delays in grid connection.
Ahead of a CapEx investment of USD 60–64 billion, TSMC is reviewing the pace of customers’ data center construction and power supply conditions.
[Weekpick Cloud] This week’s major cloud news brought together the key issues currently shaping the cloud industry. At CoreWeave, one of the leading AI cloud companies in the United States, the stock price plunged against the backdrop of a heavy debt burden, and the company’s bonds also fell sharply. In the process, massive debt risks tied to AI infrastructure expansion came into focus, while concerns over a credit crisis stemming from neoclouds intensified, fueling worries about CoreWeave’s AI infrastructure borrowing turning bad.
Market anxiety surrounding CoreWeave led to a simultaneous plunge in both its stock price and bond prices. The debt burden was cited as a key factor behind the drop, and as fears of a credit crisis stemming from neoclouds came into play, caution over CoreWeave’s financial stability deepened.
Microsoft officially withdrew from a large-scale AI data center contract in the UK. The cancellation came in the wake of delays in grid connection.
Taiwan’s TSMC has begun on-site due diligence ahead of a large-scale investment of more than USD 60 billion. Before investing, TSMC has started checking the pace at which customers are building data centers in the United States, and it has also begun reviewing power supply conditions.
CoreWeave has relied on borrowing to secure high-performance GPUs and build data centers. CoreWeave’s debt stands at about USD 50 billion, or about KRW 70 trillion. However, the price of its bonds due in 2032 fell to around USD 77 per USD 100 of face value, pushing the bonds into distressed territory. At the same time, CoreWeave CDS surged, and the company’s stock on the New York Stock Exchange also plunged more than 15% in a single day.
Reflecting this market reaction, CoreWeave’s probability of default within five years was estimated at about 63%. CoreWeave’s backlog exceeds more than USD 100 billion, but this year’s expected net loss is about USD 2.1 billion. As a result, concerns have spread that until orders are converted into cash flow, the company could face the burden of high-interest expenses and costs for replacing chips with new ones.
Meanwhile, MS terminated a contract to supply computing resources for an AI data center in Laindon, Essex, UK. The project is led by the UK neocloud company Nscale. This business is one of Nscale’s core projects.
Nscale had planned to introduce 23,040 GB300 GPUs from NVIDIA. Nscale had also planned to supply AI computing to MS Azure starting in the first quarter of 2027, with a supply capacity of up to 90 MW.
In the UK, delays in expanding the power grid have made it difficult to process demand for power connections from data centers in time. As the surge in data centers deepens bottlenecks in the UK power infrastructure, applications for grid connections in the UK rose from 41 GW in 2024 to 125 GW in 2025.
Behind this situation was a notice from UK Power Networks. UK Power Networks said the expansion of transmission and distribution networks was being delayed, and projected that sufficient power supply would be available in the early to mid-2030s.
As a result, Nscale was hit by delays in grid connection as the reason for the contract withdrawal and faced the negative fallout of losing a key customer. MS and Anthropic have accounted for about 85% of Nscale’s total contracts, and Nscale is preparing an IPO with a funding target of up to USD 3 billion.
Meanwhile, TSMC dispatched due diligence staff to sites where data center construction is underway in the United States. The purpose was to check power supply conditions before additional capacity expansion.
TSMC plans to spend USD 60–64 billion in CapEx this year and has also begun preparing for its next-generation 1.4-nanometer (A14) process. However, TSMC believes that for customers’ semiconductor orders to be converted into revenue, securing power and infrastructure for data centers must come first.
TSMC CEO C.C. Wei said that while AI demand remains solid, capacity planning reflects not only demand but also a comprehensive review of factors including CSP CapEx, the pace of data center construction, grid supply, and constraints in securing HBM.
TSMC posted quarterly revenue of USD 46.72 billion in the third quarter, up 41.2% from a year earlier, continuing its strong third-quarter performance. However, the company said it plans to take a conservative approach in reviewing business plans toward 2027 in preparation for potential risks stemming from infrastructure bottlenecks.
Source: IT DAILY · Kwon Young-seok
Original: https://www.itdaily.kr/news/articleView.html?idxno=242291
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.