NVIDIA’s Blackwell GPUs, the company’s most advanced AI hardware to date, have become a hot commodity, selling out for the next 12 months. This high demand, confirmed by NVIDIA management and analysts from Morgan Stanley, comes from a diverse range of tech giants including Google, Meta, Microsoft, AWS, and Oracle.
The demand for Blackwell is driven by its groundbreaking performance in next-generation AI training and inference. Blackwell utilizes TSMC’s CoWoS-L packaging technology and next-gen HBM3E memory, making it a powerhouse for even the most demanding AI projects. This is exemplified by OpenAI’s recent acquisition of an NVIDIA DGX B200 AI server, equipped with eight B200 AI GPUs featuring a staggering 1.4TB of high-speed HBM3 memory.
Despite initial delays caused by packaging issues that required a redesign, Blackwell production is now in full swing. However, the intense demand means that new customers ordering Blackwell GPUs will have to wait until late 2024 to receive their orders. This bottleneck, while a challenge for NVIDIA, is expected to translate into a significant market share gain in 2025. Analysts at Morgan Stanley predict that NVIDIA will maintain its dominance despite growing competition from companies like AMD and Intel.
This overwhelming demand for Blackwell underscores the burgeoning growth of the AI hardware market. NVIDIA is proactively addressing the high demand, maximizing production within the limitations of available capacity and memory supply. It’s a testament to the rapidly evolving AI landscape and NVIDIA’s position as a leading player in the field.
