Nvidia's Growing Demand for Hopper and Blackwell GPUs
NVIDIA (NVDA) is experiencing a strong demand for its Hopper and Blackwell GPUs, a trend that analysts at Morgan Stanley have noted as well. Their recent reports highlight the significant momentum that Nvidia's latest chip technologies are gaining.
Mass Production of Blackwell Chips
As per Morgan Stanley, Nvidia's Blackwell chips are now moving into mass production, driven by considerable demand from key clients. This ramp-up in production capabilities positions Nvidia favorably for future growth in a highly competitive semiconductor market.
Oracle’s AI Supercluster Project
In a noteworthy announcement, Oracle has revealed its plans to build a Zettascale AI supercluster, powered by a remarkable 131,000 Nvidia Blackwell GPUs. This project is expected to deliver an astonishing 2.4 ZettaFLOPS of AI performance, representing a substantial advancement in AI infrastructure capabilities.
Effects on Nvidia’s Supply Chain
The demand from Oracle for an increased GPU supply bodes well not just for Nvidia but also for its suppliers throughout the semiconductor industry. Morgan Stanley mentions that during a recent expo, Oracle showed interest in boosting its GPU orders, which has resulted in an optimistic outlook for the broader Asia AI semiconductor supply chain.
Enhanced Chip Production Capability
Looking forward, Morgan Stanley predicts that TSMC’s Chip-on-Wafer-on-Substrate (CoWoS) capacity will rise to between 80,000 and 90,000 wafers per month by 2025, an increase from the previously expected 70,000. This growth is projected to aid Nvidia in producing and delivering more advanced AI chips.
Ongoing Demand for Hopper GPUs
Nvidia's Hopper GPUs are also seeing a notable rise in demand as the market evolves with AI advancements. Smaller cloud service providers and various government AI projects are increasingly opting for Hopper H200 chips, securing a steady growth in sales and market infiltration for Nvidia.
Revenue Prospects from Blackwell Chips
Additionally, Nvidia expects to manufacture 450,000 Blackwell units by the fourth quarter of 2024, which could lead to potential revenue of more than $10 billion. This forecast underscores Nvidia's strong positioning within the AI semiconductor market.
Technical Challenges and Developments
Morgan Stanley recognizes that Nvidia is currently working through some technical issues related to its GB200 server racks. However, these challenges are viewed as standard in the debugging process for new product releases. Nvidia’s partner, Hon Hai, is on track to begin shipping the GB200 server racks by late Q4 2024, aligning with Nvidia’s goals to enhance its AI supply chain.
Positive Outlook for Nvidia’s AI Supply Chain
Overall, Morgan Stanley holds a positive view of Nvidia’s AI supply chain, highlighting its advantages in AI semiconductors over other segments like memory, PCs, and cloud services. As AI technology continues to grow, Nvidia is well-prepared to meet the rising demands across various sectors, ensuring its ongoing importance in the tech industry.
Frequently Asked Questions
What are the key chips driving Nvidia's growth?
The key chips driving Nvidia's growth are the Hopper and Blackwell GPUs, which are seeing increased demand across multiple sectors.
How is Oracle contributing to Nvidia’s market potential?
Oracle is contributing by planning to build a Zettascale AI supercluster powered by Nvidia GPUs, creating significant demand for their products.
What is the expected production increase for TSMC's capacity?
TSMC's CoWoS capacity is expected to increase to 80,000-90,000 wafers per month by 2025, supporting Nvidia’s growth.
What challenges is Nvidia currently facing?
Nvidia is currently addressing technical challenges with its GB200 server racks, which are typical during new product launches.
How much potential revenue can Blackwell chips generate?
Blackwell chips are projected to generate potential revenues exceeding $10 billion with an expected production of 450,000 units in Q4 of 2024.
