NVIDIA’s GB200 AI servers, initially expected for a quicker release, are now projected to hit mass production between Q2 and Q3 2025 due to supply chain challenges. The servers’ advanced specifications, including fifth-generation NVLink and a high TDP of 140kW, contribute to the delay. This delay underscores the complexities of scaling cutting-edge technology. Cloud service providers are shifting towards liquid cooling to handle the increased thermal demands.
Elon Musk personally intervened to secure a massive $1.08 billion order of NVIDIA’s highly sought-after GB200 AI servers for his xAI startup, ensuring priority access to bolster its Colossus supercomputer and accelerate AI development. This significant investment underscores xAI’s ambitious growth plans and the intense demand for top-tier AI hardware.
Microsoft has placed a significant order for NVIDIA’s new GB200 AI servers, with demand exceeding that of other cloud service providers. The order volume is expected to skyrocket in Q4 2024, pushing mass production and shipments of key components earlier than anticipated. This news highlights the escalating demand for AI computing power and Microsoft’s aggressive push into the field.
Taiwanese tech giant Foxconn is set to construct the world’s largest manufacturing facility for Nvidia’s powerful GB200 chip, aimed at meeting the skyrocketing demand for the company’s Blackwell platform, which is driving the AI revolution. The move highlights Foxconn’s commitment to the AI market and its ability to meet the challenges of producing high-performance AI chips.
NVIDIA’s upcoming GB200 NVL72 AI server, boasting an unprecedented 132kW TDP, is encountering significant development challenges. Analyst Ming-Chi Kuo reports delays in production due to complex thermal management issues and potential instability concerns. Kuo suggests that the server’s massive power consumption may necessitate additional cooling components, increasing complexity and jeopardizing its space-saving advantages.
NVIDIA’s Blackwell AI GPUs, initially facing production delays, are now being shipped to major cloud providers like Microsoft, Oracle, AWS, Meta, and others. The GB200 NVL36 and NVL72 AI servers, originally scheduled for November, will now be delivered in early December, thanks to a concerted effort by NVIDIA, TSMC, and their supply chain partners to resolve manufacturing issues.
NVIDIA’s highly anticipated GB200 AI servers are expected to begin production in September and October. These servers, which boast advanced liquid cooling systems, are crucial for NVIDIA’s AI strategy, but present unique challenges and increasing costs.
