coinmaster60freespin| The world's largest cloud computing company has suspended purchasing Nvidia's "super chips" because the upgrade is too fast

Financial Associated Press, May 22 (editor Shi Zhengcheng)-Huang Renxun, who always instills in the market that "the more you buy, the more you save", Huang Renxun is facing a challenge.Coinmaster60freespinCustomers now know that buying the latest generation of chips can "save more"Coinmaster60freespinStart to stop placing orders for the most powerful numeric product still on sale.

According to the latest news on Tuesday, Amazon Cloud Services (AWS), the world's largest cloud computing platform, has suspended the purchase of Nvidia's most powerful computing chip, the Grace Hopper super chip, waiting for the latest model to go on sale later this year.

What does Amazon think?

The idea of Amazon cloud services is not hard to understand: in a few months, Nvidia's next-generation data center, GPU, will be on the market, doubling its performance and dramatically reducing its energy consumption.

At the GTC conference in March this year, Nvidia just unveiled a new generation of Blackwell architecture GPU. The H100 that investors are more familiar with is Hopper architecture GPU, which just started shipping to customers for about a year at that time.

In Huang Renxun's own words, if you want to train a 1Coinmaster60freespinThe .8 trillion-parameter GPT model requires 8000 Hopper GPU, consumes 15 megawatts of electricity and runs for 90 days in a row. But if you use GB200 Blackwell GPU, you only need 2000, and running for 90 days consumes only 1/4 of the electricity. In addition to training large models, the cost of generating Token will also be significantly reduced.

The bigger difference is at the system performance level. Huang Renxun said that at 1.Coinmaster60freespinAt the level of GPT-MoE reasoning performance with .8 trillion parameters, the performance of Blackwell clusters can be as much as 30 times that of Hopper clusters.

Amazon Cloud Service told the media that the company had "completely transferred" its previous order for Grace Hopper super chip and replaced it with Grace Blackwell, which will be listed next. AWS stressed that the move was "reasonable" given the short interval between the two generations of chips.

Nvidia declined to comment on the report, saying it needed to comply with the silence period before the results were released on Wednesday.

Will it make a difference?

For Wednesday's after-hours results, the market expects revenue to continue to grow to $24.59 billion. For comparison, Nvidia had revenue of $7.192 billion in the same period of the previous fiscal year, compared with $22.1 billion in the previous quarter.

What can really affect the share price is Nvidia's outlook for when this doubling of growth will continue. Now a fresh uncertainty has been added-will there be a period of standstill in revenue growth this year as more and more customers wait for Blackwell chips to hit the market?

Blackwell chip capacity will continue to climb for the rest of the year, and the mainstream expectation in the market is that Nvidia will start delivering next-generation servers in the fourth quarter of this year. Citigroup analysts pointed out that as the waiting time for Nvidia to deliver products has shortened since last year, there may be a "potential gap" in the demand for AI chips.

Perhaps coincidentally, Nvidia's share price was created after the release of the Blackwell chip, followed by a concussion that lasted for three months until recently.

Of course, from another point of view, Nvidia is far from the point where it needs to consider "chip unsalability". Analysts at Morgan Stanley say that even as the supply of Nvidia's previous generation of chips becomes more plentiful, new cloud, corporate and sovereign customers can still be seen buying all available Hopper chips on the market.