Nasdaq Composite falls 2.4%, Nvidia slides on AI financing concerns
Key Facts
The Nasdaq Composite fell 2.4% during trading on Thursday, October 22, 2026, according to an Associated Press update at 12:56 p.m. U.S. Eastern Time, as the global selloff in artificial intelligence stocks continued. Nvidia shares fell 4.3% during the session, extending pressure on the sector amid concerns about financing its expansion. The losses followed broader declines in chip stocks, with Reuters reporting on October 16 that the Philadelphia Semiconductor Index had lost about 12% during the month through that report. The agency said investors were monitoring developments at CoreWeave, the AI cloud company with about $50 billion in debt. The focus on the company's position came amid concerns surrounding financing for the sector's expansion, whose arrangements include commitments and guarantees Nvidia provides to selected partners.
On capital raising, Nvidia announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent financing platforms for AI infrastructure. The platforms aim to mobilize more than $500 billion in third-party capital over time, according to the company's announcement. The announcement places the partner financial institutions within arrangements dedicated to financing computing infrastructure, with the stated target dependent on raising funds from outside Nvidia. The company tied implementation of the partnerships to the completion of agreements, saying, "These partnerships remain subject to the signing of definitive agreements." Establishing the platforms therefore depends on completing those agreements, while the capital-raising target extends over time.
Existing arrangements include commitments under AI cloud agreements totaling $36 billion as of July 26, 2026, according to Nvidia's quarterly filing. The company said these agreements typically run for 6 years, tying it to the use of computing capacity over an extended period. The commitments decline as customers or Nvidia use that capacity, according to the mechanism described in the filing. Alongside the cloud agreements, the company's maximum aggregate exposure under land, power and building guarantees provided to selected AI cloud partners was $3.5 billion. These guarantees relate to lease obligations in the event of default, meaning Nvidia's relationships with its partners include commitments covering infrastructure sites, power and the buildings used.
During its earnings call in August, Nvidia outlined another model based on guaranteeing a minimum level of revenue for specialized cloud providers. In return for that guarantee, the company receives a share of the cloud provider's revenue above the agreed threshold. Explaining what it receives in return, it said, "We share in a portion of specialized cloud providers' revenue," linking its returns to revenue generated above that level. It added that independent capital assesses each deal on its merits, as part of the arrangements it presented to investors. It defined its role in the revenue-sharing model by saying, "We do not provide loans," while its announced arrangement is based on guarantees and revenue sharing.
Alongside its infrastructure commitments, Nvidia continued allocating capital to shareholders, with Reuters reporting on September 28 that it had increased its share buyback authorization by a record $150 billion. The agency said in the same report that the company ended the quarter that closed in July with cash and cash equivalents of $22.44 billion. In Europe, Huang said during Nvidia's conference in Berlin, according to the Associated Press on October 21, that the company had invested $10 billion in European startups and partners over the previous year. The company's conference presentation covered aspects of how computing infrastructure is deployed, alongside funds directed to partners. Huang presented a ready-made modular design for computing centers, saying it cuts the time needed to build an AI center at a prepared site "from months to days." He tied that reduction to the availability of a prepared site, as part of his presentation on accelerating the construction of computing centers.
Upcoming dates include Nvidia's call to discuss results for the third quarter of its fiscal 2027, which the company scheduled for November 17. The earnings date follows disclosures about cloud commitments and guarantees, and an explanation of the revenue-sharing model during the August call. In Tesla's computing expansion plans, the company said in its investor update that it aims to reach 400 megawatts of superintelligence training computing capacity by the end of 2026. Tesla earmarked this capacity to support the development of vehicle autonomy software and humanoid robots. It said in the update, "We plan to reach 400 megawatts of training computing capacity by the end of the year."