Unlock the Editor’s Digest for free
Roula Khalaf, editor of the FT, selects her favorite stories in this weekly newsletter.
Amazon’s cloud computing unit has halted orders for Nvidia’s most advanced “superchip” pending a more powerful new model, as investors worry about a dip in demand between the $2.3 trillion chipmaker’s product cycles.
The Silicon Valley-based chipmaker unveiled a new generation of processors called Blackwell in March, just a year after its predecessor Hopper shipped to customers. Nvidia’s CEO Jensen Huang said the new products would be twice as powerful for training large language models, the technology behind OpenAI’s ChatGPT.
Amazon Web Services, the world’s largest cloud computing provider, told the Financial Times that it had “completely converted” its previous orders for Nvidia’s Grace Hopper superchip, which launched in August, and replaced it with its successor Grace Blackwell.
The company said the move “made sense” as “the window between Grace Hopper and Grace Blackwell was small.”
Nvidia declined to comment ahead of Wednesday’s quarterly earnings report, citing quiet period rules.
Nvidia shares fell about 1 percent in early trading on Tuesday.
Analysts expect the chipmaker to report that revenue tripled in the most recent quarter, driven by a frenzy of spending by Big Tech companies on artificial intelligence technology. But some investors are beginning to wonder how long the country can sustain its extraordinary growth.
While Big Tech companies have pledged to continue investing tens of billions of dollars this year in building out data center infrastructure for AI, “there is fear [on Wall Street] on a break for Blackwell,” Morgan Stanley analysts wrote in a note to clients this week.
Production of the new Blackwell chips will increase over the course of this year. Analysts expect them to be delivered in the fourth quarter. Meanwhile, analysts at Citi pointed to a “potential bubble” in demand for AI chips after last year’s long wait times for Nvidia’s chips disappeared.
Chips based on Nvidia’s Hopper architecture, such as the sought-after H100 graphics processing units, entered full production in September 2022. The Grace Hopper superchip, also known as GH200, contains several H100 GPUs in addition to fast memory, connectivity and a central processing unit.
Its successor, the GB200, is the first product to benefit from Blackwell.
Neither Amazon nor Nvidia would confirm the value of the order. Analysts at HSBC estimate that a GB200 chip, which contains two B100 chips, will cost as much as $70,000, with the price of an entire server with the new technology installed rising to $3 million.
AWS continues to offer other Nvidia chips, including H100s, to its cloud customers. But as one of Nvidia’s largest customers, the AWS move is likely to worry investors who are already worried that tech companies will delay their purchases pending Blackwell’s release.
Demand for Nvidia’s H100 chips far exceeded supply for much of last year, as the launch of OpenAI’s groundbreaking ChatGPT sparked a wave of investment in AI infrastructure from cloud and internet companies, startups and enterprise buyers.
But since the beginning of 2024, the long waiting times for the delivery of H100 chips have decreased.
While Nvidia’s shares have nearly doubled in value since the start of the year on investor confidence about demand for AI chips, the shares have struggled to post consistent gains since the Nvidia’s GPU Technology Conference chipmaker in March, where the Blackwell was unveiled.
Nvidia has historically struggled to manage supply and demand smoothly between product upgrades. Recent surges in demand for chips suitable for video games and cryptocurrency mining during the Covid-19 pandemic were followed shortly after by a glut of GPUs.
Analysts at Morgan Stanley said they were confident that even though Nvidia’s range of last-generation chips was now more readily available, “in the meantime, we’re seeing new cloud, enterprise and sovereign customers using the full available Hopper decrease supply”.