Nvidia Weighs Memory Cut for AI Chips as AMD Secures Supply
Yahoo Finance ·
NVIDIA Corporation (NASDAQ: NVDA) is considering a surprising shift for its upcoming Rubin Ultra GPU, testing versions with reduced high-bandwidth memory ranging from 192GB to 256GB, well below Jensen Huang's initial 1-terabyte promise. Amid a severe industry-wide memory shortage, this reduction aims to protect production volume and lower costs to broaden its customer base. Rubin Ultra is scheduled to ship in late 2027, and Nvidia recently formed a $500 billion partnership with SK Hynix's parent company to secure stable supply. Meanwhile, rival Advanced Micro Devices, Inc. (NASDAQ: AMD) claims it has already secured the necessary HBM from Samsung Electronics, SK Hynix, and Micron Technology for its upcoming Helios system. Set to ship to Microsoft, Meta, OpenAI, and Oracle, AMD's data center revenue jumped 57% year-over-year in Q1 2026, becoming its largest segment. Despite AMD's confidence, Helios carries execution risks and higher estimated system costs of $5 million to $5.5 million compared to Nvidia's Vera Rubin at $3.5 million to $4 million. Furthermore, Nvidia commands over 95% of the data center GPU market, with 275 hedge funds holding its stock in Q1 2026, easily outpacing AMD's 134 holders.
AI 시장 분석
NVIDIA (NVDA) is reportedly considering reducing the High Bandwidth Memory (HBM) capacity of its next-generation Rubin Ultra GPU from the initial 1TB to around 192GB-256GB. Meanwhile, AMD announced that it has secured sufficient memory for its Helios system, attracting major clients like Microsoft and Meta. Amid memory supply shortages, the strategic differences between the two semiconductor giants are expected to significantly impact future market share.
상승 영향
- Semiconductors — AMD is accelerating data center revenue growth by stably securing HBM through partnerships with major memory suppliers such as Samsung Electronics, SK Hynix, and Micron.
하락 영향
- Semiconductors — Due to a severe industry-wide high bandwidth memory shortage, NVIDIA is reportedly considering launching its next-generation Rubin Ultra GPU with significantly reduced memory capacity, increasing cost burdens for clients.
DYAX 전담 분석
Nvidia's consideration of memory reduction reflects HBM shortages and cost burdens, which are key components of AI chips, raising concerns over increased customer costs and performance constraints. Conversely, while AMD claims to have secured supply through partnerships with major memory suppliers, Helios's high system price and a weaker software ecosystem compared to CUDA act as risks.
In the bullish scenario, Nvidia maintains market share through the shipment of large volumes of entry-level chips while AMD expands supply to big tech companies. In the bearish scenario, intensifying memory shortages could subject both companies to cost pressures and shipment disruptions. Key indicators to watch are HBM supply price trends and the actual adoption scale of major clients' chips.
AI가 생성한 분석으로 투자 자문이 아닙니다.
DYAX Investor Sentiment
Bullish (Long) 38% · Bearish (Short) 62%
429 participants
Related News
- Two Years of Friday Paydays: The 44% Fund Sustains NAV and Outperforms Half of Wall Street
- Consumer Subscription Stocks Q2 Results: Benchmarking Netflix
- JEPI Manages $46 Billion with 5.8% YTD Gain as Four Similar Strategy Funds Surge by Double Digits
- Stock Market Today, Aug. 21: Tesla Gains on Nevada Robotaxi Permit Approval
- Chime Report Highlights K-Shaped Wealth Divide Among Millennials
- Hims & Hers Faces Visa Sanctions Amid Surge in Customer Disputes
