Nvidia and Meta Reach Massive Cooperation Deal: The two parties announced a multi-year strategic agreement. Meta will deploy millions of Nvidia Blackwell and next-generation Rubin architecture GPUs and will become the first hyperscaler to independently deploy Nvidia Grace CPUs on a large scale, aiming to optimize its next-generation large language models.
Nvidia May Tier Memory for Rubin GPUs: To balance performance and supply, Nvidia is considering a "dual-sourcing strategy" for its next-generation Rubin AI GPUs. This would involve using high-speed 11.7Gbps HBM4 memory in core products, while secondary products would feature a 10Gbps version, addressing potential production capacity issues for high-speed memory.
Jensen Huang Teases "Unseen" New Chips: Nvidia CEO Jensen Huang revealed that the company will launch several "world-unseen" new chips at the GTC 2026 conference, which opens on March 15th. Speculation suggests these could be derivatives of the Rubin series or the next-generation Feynman series.
SoftBank and AMD Jointly Verify Next-Gen AI GPUs: SoftBank and AMD announced a joint verification project aimed at using AMD Instinct GPUs to advance next-generation AI infrastructure. The partners will develop a "coordinator" function to enable flexible partitioning and allocation of GPU resources based on AI model sizes, improving operational efficiency.
Intel Accelerates GPU Push and Warns of Chip Shortage: Intel CEO Lip-Bu Tan stated the company is accelerating its efforts in the GPU field and has appointed a new chief architect. He also warned that due to the data center construction boom, the ongoing memory chip shortage might not ease until 2028.
Altman Questions Musk's Space Computing Plan: OpenAI CEO Sam Altman publicly questioned Elon Musk's concept of space-based computing power, arguing that deploying data centers in space faces practical challenges like maintaining equipment with high failure rates. He also refuted Musk's vision of "1,000 GPUs per person," calling the idea of 8 trillion GPUs for a global population of 8 billion "absurd and infeasible."
Meta-Nvidia Deal Shakes Up Chip Market Landscape: Following the announcement of the cooperation agreement, market reaction was swift. Nvidia and Meta's stock prices rose, while shares of competitors AMD, Intel, and Arista Networks fell in pre-market trading. Analysts suggest Meta's shift towards Arm architecture CPUs presents a serious challenge to the traditional x86 camp.
Nvidia Earnings Report Becomes Market Focus: As the "ultimate bellwether" for AI infrastructure, Nvidia is scheduled to release its Q4 FY2026 earnings after the U.S. market closes on February 25th. Market consensus expects revenue around $65.6 billion, with investors highly focused on any comments regarding Blackwell chip production capacity and the Rubin roadmap.
CoreWeave and Dell Earnings to Reveal Compute Demand: CoreWeave, a specialized cloud service provider, and Dell Technologies, a leader in AI-optimized servers, will report earnings on February 26th. CoreWeave's order backlog of $56 billion and whether Dell's $18.4 billion in AI server orders translate into healthy gross margins will be key indicators of actual enterprise demand for computing power.
Nvidia Tech to Permeate Meta App Privacy Protection: According to the cooperation agreement, Meta will not only adopt Nvidia's Spectrum-X Ethernet networking platform but also plans to apply Nvidia's confidential computing technology within its apps like WhatsApp. This aims to enhance AI performance while meeting global data security and user privacy standards.