Nvidia AI demand remains strong as custom-chip push tests selective market

  • Nvidia’s strong quarter reinforced persistent AI demand, while its NVHBM architecture expands its participation in custom accelerator infrastructure.
  • Nvidia fell 1.35% to $217.80 in Tuesday’s premarket session; Amazon’s Annapurna Labs plans to use NVHBM with its Trainium4 AI chip.
  • Marvell’s post-earnings decline, Alphabet’s 15% retreat from its May high and potential hyperscaler spending constraints highlighted rising investor selectivity.

Nvidia’s strong Wednesday quarter reinforced resilient artificial-intelligence demand, but its shares fell 1.35% to $217.80 in Tuesday’s premarket session as investors reduced exposure to large-cap technology stocks. The company’s NVHBM memory architecture expands its role in custom AI chips, potentially allowing it to sell networking, rack architecture, software and memory subsystems to accelerator developers without designing their accelerators. Amazon.com’s Annapurna Labs plans to use NVHBM with its Trainium4 chip, while MediaTek has adopted NVIDIA’s NVLink Fusion platform. The move could intensify competition with Broadcom and Marvell, although Counterpoint Research analyst Neil Shah questioned NVIDIA’s comparisons because they used not-yet-shipping HBM4E and lacked a full testing baseline. Elsewhere, Marvell’s post-earnings plunge, Alphabet’s decline and a cooling momentum trade showed investors becoming more selective across AI-linked stocks, while software shares rallied as markets focused on resilience and accelerating growth. Hyperscaler capital expenditure could exceed $1 trillion in 2027, but labor shortages, permitting delays, NIMBY opposition and component shortages could prevent the spending from being completed.

The information on this website is generated using AI and we cannot guarantee its accuracy. Please use it as reference information only.