Nvidia

Apple briefly surpassed Nvidia as the world's most valuable company on July 17, touching about $4.88 trillion as Nvidia shares slipped roughly 3% to a market value near $4.84 trillion during an AI-chip selloff, before Nvidia edged back ahead by the close. It marked the first time in more than a year that Apple held the top spot, which Nvidia had claimed since June 2025. Investors have rewarded Apple's lighter capital-spending approach — its stock is up about 23% in 2026 — even as megacap AI names cooled.

Nvidia on July 16 unveiled Cosmos 3 Edge, a 4-billion-parameter 'world model' for on-device vision reasoning and robot-action generation that runs on Jetson and edge GPUs, letting robots perceive and navigate physical environments in real time. During CEO Jensen Huang's visit to Tokyo, the company said Japanese industrial leaders — including FANUC, Fujitsu, Hitachi, Kawasaki Heavy Industries, Kubota, NEC, SoftBank, Sony and Yaskawa Electric — intend to join a new NVIDIA Cosmos Coalition to build open physical-AI models. Kawasaki said it plans to apply Nvidia's Cosmos, Isaac and Holoscan software to surgical-support, nursing-assistance and hospital-transport robots.

A senior U.S. trade official said on July 14 that only 'very few' of Nvidia's H200 artificial-intelligence chips have so far been shipped to China and Hong Kong, describing 'a very small quantity,' after Washington began issuing export licenses for the older Hopper-generation processor. The Trump administration agreed late last year to approve H200 sales to China in exchange for a 25% cut of the revenue, assessing buyers case by case against national-security conditions. The remarks suggest the resumption of China sales — potentially a meaningful new revenue stream for Nvidia — is still ramping slowly.

Nvidia has more than halved the number of Asian customers cleared to buy its AI chips, introducing a 'white list' of firms that passed tougher compliance checks meant to stop the processors from being diverted to China, the Financial Times reported July 14. The company has stepped up vetting in Singapore, Malaysia and Japan — with staff now visiting customers' data centers, verifying contracts and interviewing end users — and many smaller 'neo-cloud' providers that failed the initial review were dropped and must reapply. The U.S. Commerce Department is backing the effort, part of a Trump administration push to keep advanced Blackwell-class chips from reaching Chinese-linked entities through third countries.

Semiconductor research firm SemiAnalysis reported that Nvidia's Kyber NVL144 rack — a system packing 144 next-generation Rubin Ultra GPUs into a single NVLink domain — has slipped from 2027 to 2028 because its 'orthogonal backplane' printed-circuit midplane proved too defect-prone to manufacture at volume, and that a planned stopgap design was scrapped after customer pushback. Nvidia disputed the report, telling Bloomberg its 'road map is intact' and noting that its standard Rubin systems are already in full production. The two claims can both hold because the reported delay affects only the highest-density Rubin Ultra rack, not the base Rubin platform shipping in 2026.