The insatiable demand for Nvidia Corp.’s NCDA cutting-edge GB200 and upcoming GB300 systems is driving an unprecedented acceleration in production, according to Clark Tang, a Partner at Altimeter Capital, who recently spent a week engaging with the chipmaker’s supply chain partners in Taiwan.
What Happened: Tang’s insights, shared via an X post, paint a picture of intense activity and significant breakthroughs in manufacturing, despite the immense complexity of the new hardware. His observations suggest a surge in demand from leading AI labs and cloud service providers (CSPs), intensified by the recent advancements in reasoning models.
This escalating interest echoes recent commentary from tech giants like Microsoft Corp. MSFT, Amazon.com Inc.‘s AMZN AWS, and OpenAI, alluding to their growing need for advanced AI infrastructure.
A critical hurdle for Jensen Huag‘s company has been the production ramp-up of its highly intricate GB NVL72 systems.
Unlike previous generations, each Blackwell GPU within these systems must seamlessly communicate with 71 other GPUs simultaneously through the NVLink Switch and Backplane, presenting a manufacturing challenge “an order of magnitude harder” than the earlier 8-GPU HGX boxes.
Despite initial “pessimistic reads from Taiwan supply chain checks,” Tang says, “Taiwan engineers are working incredibly hard.” However, their relentless efforts are yielding positive results, with “shortening of cycle times for GB200, which will increase finished product shipment rates.”
A significant development highlighted by Tang is the shift in responsibility for testing and integration.
Why It Matters: CEO Huang and various Original Design Manufacturers (ODMs) have emphasized that CSPs are increasingly relying on ODMs to handle more of the traditional testing and “bring up” processes, which were previously performed by the customers for Hopper systems.
This means that when a GB200 is shipped, it is “as close to plug and play as shipping a server ever was.”
Hon Hai (Foxconn) and Wistron, two key ODMs, reportedly expressed strong confidence in their production capabilities. They are actively developing their testing protocols and automating production processes.
Tang anticipates a dramatic increase in rack throughput, projecting growth of “over 100% month-over-month for this quarter.”
This accelerated deployment of GB200s for training and inference is expected to “usher in another era of AI models altogether,” as “models and software are like water, and they adapt/expand to fit the new vessel that is deployed.”
Price Action: Nvidia’s shares closed 1.16% lower on Friday. The stock was down 5.08% on a year-to-date basis, and 15.27% higher over a year.
Benzinga Edge Stock Rankings shows that Nvidia had a stronger price trend over the short, medium, and long term. Its momentum ranking was solid, however, its value ranking was poor at the 6.61th percentile. The details of other metrics are available here.
The SPDR S&P 500 ETF Trust SPY and Invesco QQQ Trust ETF QQQ, which track the S&P 500 index and Nasdaq 100 index, respectively, fell on Friday. The SPY was down 0.68% to $579.11, while the QQQ declined 0.93% to $509.24, according to Benzinga Pro data.
Read Next:
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.
Photo courtesy: jamesonwu1972 / Shutterstock.com
Edge Rankings
Price Trend
© 2025 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.
Trade confidently with insights and alerts from analyst ratings, free reports and breaking news that affects the stocks you care about.