Nvidia faces earnings test amid hyperscaler concerns as AI factory push widens

In May, Nvidia changed its financial reporting to separate hyperscalers from customers it now calls AI clouds, industrial and enterprise, or ACIE. In the fiscal first quarter, hyperscaler sales totaled $37.9 billion while ACIE revenue was nearly $37.5 billion. ACIE grew 31% from the prior period, more than double the 12% growth from hyperscalers, Nvidia said. The company has said there are only five or six hyperscalers, while the rest of the industry represents about 250,000 companies.

Skepticism has crept into the stock, with shares falling for a seventh straight session on Monday, down 7.5% over that stretch. CNBC cited Gene Munster, managing partner at Deepwater Asset Management, saying investors worry about how sustainable Nvidia's run has been and whether hyperscalers can give much more. Amazon and Alphabet turned cash flow negative in the second quarter, Meta's cash generation fell more than 90% from a year earlier, and both Elon Musk's public companies, SpaceX and Tesla, reported negative free cash flow as they pursue AI expansions.

Analysts expect the customer mix to shift. According to StreetAccount estimates cited by CNBC, ACIE revenue is expected to grow 149% from a year earlier to $43 billion in the fiscal second quarter, while hyperscaler revenue is expected to grow 83% to $43.6 billion. Total revenue is expected to nearly double to $92.2 billion, with data center revenue reaching $86.3 billion, or 94% of total. For the full year, analysts see revenue rising 83% to $396 billion. Separately, Nvidia launched a program with six financial firms that could pull together up to $500 billion in financing for GPU purchases, treating chips as an investable asset. Bernstein analyst Stacy Rasgon told CNBC that there have been big headlines but few details on how the program will work.

In a separate development, Cisco Systems Inc. and Nvidia Corp. expanded their Secure AI Factory with Nvidia to full rack-scale systems, according to SiliconANGLE. The expanded portfolio, built with Super Micro Computer Inc., includes liquid-cooled compute and Nvidia Cloud Partner-compliant solutions for hyperscale, neocloud and sovereign clouds. The offering combines Nvidia Spectrum-X-based switches on the back end with Cisco Silicon One-based switches on the front end, unified by Cisco's Nexus One management platform. Cisco said the full rack-scale solution will be orderable in September. "We are at the beginning of one of the largest datacenter buildouts in history," Cisco president and chief product officer Jeetu Patel said. The systems support Nvidia NVL72 racks and a path to the Vera Rubin platform, with liquid cooling needed for racks exceeding 200 kilowatts.

Nvidia also announced the Jetson Orin Nano 2, a robotics computer for edge AI that doubles the compute capacity of the previous generation, according to SiliconANGLE. The module provides 78 trillion operations per second of AI compute, 8 gigabytes of memory and an eight-core CPU, while consuming 40% less power in 15-watt mode. Partners including Cognex, Doosan Bobcat, Matic Robotics and Wing Aviation are working on integrating it into smart robotics. Wing, a drone delivery company, said the processor will give it a path to more responsive and energy-efficient drones. The device will be generally available as a module and developer kit in the first half of 2027.