CoreWeave, which previously triggered a crisis of confidence in the AI infrastructure industry due to data center delays, has announced a key breakthrough. On Monday, citing internal company information, it was reported that CoreWeave has successfully achieved the important milestone of its first chip deliveries for the data center it is building for its client OpenAI in Denton, Texas. According to CoreWeave executives, the company rapidly scaled from "a few racks delivered" in mid-November last year to over 16,000 GPUs by the end of December. Company leadership noted that the peak number of GPUs brought online in a single day exceeded 2,000 units. This progress signals that CoreWeave is recovering from the revenue hit it suffered in last year's fourth quarter due to delays from a data center supplier. Following the news, CoreWeave's stock surged over 12%, bringing its year-to-date gain to over 20%.

CoreWeave's Chief Operating Officer Sachin Jain, Chief Technology Officer Peter Salanki, and Senior Vice President of Engineering Chen Goldberg stated in a joint message to employees:

We completed this milestone under an intensely focused spotlight. The past few months have tested us in ways few companies experience. We faced setbacks from partners and factors beyond our control.

The delays with the Denton project have become a classic case study of the increasingly prominent "blame game" phenomenon within the data center construction sector. As construction and deployment delays become more common in the AI infrastructure industry, the involved parties have begun shifting responsibility onto each other. The market panic triggered by this incident not only caused CoreWeave's stock to plummet over 60% but also impacted Broadcom and Oracle, with both companies' stocks crashing over 17% within three trading days.

Heavy Rain Delays Data Center Delivery

This large AI data center cluster in Denton, Texas, planned to install approximately 260 megawatts of computing capacity for lease to OpenAI, but was delayed by about 60 days last year due to summer heavy rain and high winds, which prevented the contractor from pouring concrete as scheduled. The delay directly impacted CoreWeave's business model—the company relies on high-interest debt to purchase thousands of advanced Nvidia AI chips, installs them in server racks within data centers leased from third-party landlords, and then subleases chip access to AI companies. During an earnings call on November 10 last year, contradictory statements from CoreWeave's management exacerbated investor panic. CEO Michael Intrator initially stated that "just one data center" was experiencing issues, but was promptly corrected on the spot by CFO Nitin Agrawal, who clarified that it was actually "one data center supplier" facing delays, implying a broader scope for the problem. The following day, CoreWeave's stock price fell from $105.61 to $88.39, a 16.3% drop, and continued to decline through December. Despite achieving delivery progress, CoreWeave still faces severe financial challenges. DA Davidson analyst Gil Luria previously stated that CoreWeave has "by far the ugliest balance sheet in the tech industry." The company's recent quarterly revenue doubled year-over-year to nearly $1.4 billion, but it remains unprofitable, posting a loss of $110 million in its most recent quarter.