Nvidia Extends AI Reign To 2028: JPMorgan Says It's Still '1-2 Steps Ahead'

Core Insights - Nvidia is redefining the AI landscape with an aggressive product roadmap extending to 2028, emphasizing its leadership in AI acceleration [1] - The upcoming Blackwell Ultra (B300) is set for a second-half 2025 launch, promising a 50% increase in compute performance compared to current Blackwell chips [2] - Nvidia's innovations in AI models, such as DeepSeek, are expected to significantly increase compute complexity, requiring 20 times more tokens and 150 times more compute power than traditional models [3] Product Developments - The Rubin GPU architecture, expected by late 2026, will offer 3.3 times the performance of Blackwell Ultra, utilizing next-gen HBM4 DRAM with 288GB capacity [4] - By 2027, the Rubin Ultra NVL 576 will introduce a 144-GPU chipset based on a 4-chiplet architecture, delivering a remarkable 14 times boost over Blackwell Ultra NVL72 [4] Market Strategy - Nvidia is exploring co-packaged optics (CPO) for future networking applications, although adoption may face challenges related to technical reliability and vendor lock-in [5] - The company is also expanding into humanoid robotics, leveraging its full-stack development platform to potentially transform automation in the coming years [5] Analyst Outlook - JPMorgan maintains a bullish outlook on Nvidia, asserting that the company is consistently ahead of its competitors, with a price target of $170 indicating significant upside potential [6]