Unlock Editor’s Digest for free
Roula Khalaf, editor of the FT, picks her favorite stories in this weekly newsletter.
Amazon’s cloud computing division has halted orders for Nvidia’s most advanced “superchip” to wait for a more powerful new model, as investors worry about a drop in demand between product cycles at the $2.3 trillion chipmaker.
The Silicon Valley-based chipmaker introduced a new generation of processors called Blackwell in March, barely a year after its predecessor, Hopper, began shipping to customers. Nvidia CEO Jensen Huang said the new products are 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 has “completely transitioned” its previous orders for Nvidia’s Grace Hopper superchip, which launched in August, and replaced it with its successor, Grace Blackwell have.
The company said the move “made sense” because “the window of opportunity between Grace Hopper and Grace Blackwell was small.”
Nvidia declined to comment ahead of its quarterly earnings report on Wednesday, citing quiet periods.
Shares of Nvidia fell about 1 percent in early trading on Tuesday.
Analysts expect the chipmaker to report that revenue tripled last quarter, fueled by rapid spending by big tech companies on artificial intelligence, but some investors are beginning to question how long the company can sustain its exceptional growth trajectory.
While Big Tech companies have pledged to continue investing tens of billions of dollars this year to build data center infrastructure for AI, “there are concerns.” [on Wall Street] about a pause ahead of Blackwell,” Morgan Stanley analysts wrote in a note to clients this week.
Production of the new Blackwell chips will ramp up later 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 long waits for Nvidia’s chips ended last year.
Chips based on Nvidia’s Hopper architecture, such as the sought-after H100 graphics processors, entered series production in September 2022. The Grace Hopper superchip, also known as GH200, contains multiple H100 GPUs as well as high-speed memory, connectivity and a central processing unit.
Its successor, the GB200, is the first product to take advantage of Blackwell.
Neither Amazon nor Nvidia would confirm the value of the order. Analysts at HSBC have estimated that a GB200 chip containing two B100 chips will cost up to $70,000, with the price of an entire server using the new technology reaching up to $3 million .
AWS continues to offer other Nvidia chips, including H100s, to its cloud customers. However, as one of Nvidia’s largest customers, AWS’s move is likely to worry investors who have already worried that tech companies would delay their purchases while they wait for Blackwell’s release.
For much of last year, demand for Nvidia’s H100 chips far outstripped supply as the launch of OpenAI’s groundbreaking ChatGPT sparked a wave of investment in AI infrastructure by 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 eased.
Although Nvidia stock has nearly doubled in value since the start of the year on investor confidence about demand for AI chips, shares have struggled to gain momentum since the chipmaker’s GPU technology conference in March, where the Blackwell was unveiled there has been a constant increase.
Nvidia has historically struggled to smoothly manage supply and demand between product upgrades. The recent surge in demand for chips suitable for video games and cryptocurrency mining during the Covid-19 pandemic was soon followed by a flood of GPUs.
Analysts at Morgan Stanley said they are confident that even as Nvidia’s last-generation chip supply is now more readily available, “in the meantime, new cloud, enterprise and government customers are taking up the full Hopper offering available.”