Nvidia (NVDA) closed out a volatile trading week with a remarkable gain of over 10%, signaling renewed investor confidence in the artificial intelligence infrastructure narrative. The rally lifted the broader semiconductor complex, with the Philadelphia Semiconductor Index (^SOX) advancing more than 8% across the five-day period. This recovery comes after the prior week’s sharp selloff driven by concerns over whether massive AI capital expenditures by hyperscalers are justified by near-term fundamentals.
SpaceX Partnership Validates Nvidia’s Next-Gen Architecture
A significant catalyst emerged from SpaceX’s first public earnings call, where CEO Elon Musk announced the rocket and satellite company will deploy data centers both terrestrially and in orbit exclusively using Nvidia’s upcoming Vera Rubin architecture. Musk specifically highlighted the NVL72 rack-scale system as “the best AI computer,” underscoring Nvidia’s dominant position in high-performance accelerated computing. This endorsement from a cutting-edge aerospace firm reinforces the structural demand for Nvidia’s full-stack solutions beyond traditional cloud providers.
Rubin Ultra Memory Configuration Under Review
Despite the positive momentum, Nvidia faces supply chain constraints. A report from The Information revealed the company is evaluating reduced high-bandwidth memory (HBM) configurations for its flagship Rubin Ultra chip, which was originally designed with 1TB (1,000GB) of HBM. Due to global memory shortages, Nvidia is testing variants with 192GB and 256GB of HBM. For context, the current-generation Rubin chip utilizes 288GB. Any memory reduction could impact training throughput for large language models, though it may improve yield and cost efficiency.
AMD Diverges on Earnings Reaction
Rival Advanced Micro Devices (AMD) delivered better-than-expected Q2 results and raised Q3 guidance, yet shares initially dropped roughly 8% as investors demanded stronger AI accelerator revenue visibility. AMD still finished the week nearly 2% higher. Notably, AMD has outperformed Nvidia over the trailing 12 months, surging approximately 180% versus Nvidia’s 28%, driven by EPYC CPU market share gains and anticipation for its Helios rack-scale platform competing directly with NVL72.
Market Implications
The semiconductor sector’s rebound reflects a shift from valuation anxiety to execution focus. With major tech capex cycles intact and new enterprise customers like SpaceX committing to Nvidia’s roadmap, the AI infrastructure build-out appears to be entering a more durable phase. However, HBM supply constraints remain a key risk variable for 2025 product ramps.
FAQ
- Why did Nvidia stock rise over 10% this week? The rally was driven by easing AI spending fears, a major endorsement from SpaceX adopting Nvidia’s Vera Rubin architecture for orbital and ground data centers, and broad semiconductor sector rotation.
- What is the Rubin Ultra HBM issue? Nvidia is considering reducing high-bandwidth memory on its next-gen Rubin Ultra chip from 1TB to as low as 192GB due to global HBM supply shortages, which could affect AI training performance but improve manufacturing yields.
- How does AMD compare to Nvidia recently? AMD posted strong Q2 earnings but saw a sharp intraday selloff on guidance expectations; however, AMD shares have outperformed Nvidia over the past year (+180% vs +28%) due to CPU strength and upcoming Helios platform competition.
