TECHNOLOGY

Nvidia drives forward accelerated computing advantage

Digital Author August 31, 2024

0 4 3 minutes read

Tierney – stock.adobe.com

Nvidia is experiencing mammoth search recordsdata from for accelerated computing the usage of GPUs to walk compute-intensive datacentre workloads

Cliff Saran

Cliff Saran,
Managing Editor

Revealed: 29 Aug 2024 14: 12

Nvidia has posted a 154% boost in revenue for its datacentre alternate within the second quarter of 2025, when put next with the identical interval a one year within the past, producing $26.3bn.

General, the company posted what it said changed into “narrative quarterly revenue” of $30bn, up 15% from the important thing quarter and up 122% from a one year within the past.

Nvidia founder and CEO Jensen Huang said the outcomes confirmed that datacentre operators were transitioning to accelerated computing. “Nvidia completed narrative revenues as world datacentres are in stout throttle to modernise the total computing stack with accelerated computing and generative AI [artificial intelligence],” he said.

Engaging remarks accompanying the monetary results display conceal that Nvidia’s datacentre numbers were pushed by search recordsdata from for its Hopper graphics processing unit (GPU) computing platform, which is being feeble for coaching and inferencing in tidy language units, recommendation engines and generative AI (GenAI) applications.

The corporate said the sequential boost changed into pushed by client cyber net and endeavor firms. The results also display conceal a roughly even split between the spending by cloud provider companies (45%) and the skills investments being made by client cyber net and endeavor firms (50%). The last boost in Nvidia’s datacentre alternate came from its networking merchandise, which saw a sequential upward thrust in revenue of 16%.

In accordance with a transcript of the earnings name, posted on Looking for Alpha, when Huang changed into requested about how datacentre operators justified the return on investment for GPU-powered servers, in step with Nvidia skills, he described a transition from frequent-motive computing to accelerated computing.

“CPU scaling has been identified to be slowing for some time. It’s slowing to a plod,” he said. “And yet the amount of computing search recordsdata from continues to develop somewhat greatly. You may well maybe presumably even estimate it to be doubling each single one year, so if we don’t hang a fresh methodology, computing inflation may perhaps well maybe be using up the associated price for each company, and it may perhaps perhaps well well maybe be using up the energy consumption of datacentres world wide.”

Over the previous one year, Huang has been pushing the root of accelerated computing, powered by datacentre GPUs. “It’s no longer strange to deem somebody save 90% of their computing price. And the motive within the assist of that’s, actually, you appropriate speeded up an utility 50 cases, [so] that you may well query the computing price to advise no somewhat greatly,” he said.

Attributable to the hang to invent more computing search recordsdata from, Huang predicted that each single datacentre would incorporate GPU skills. “The enviornment builds about $1tn worth of datacentres – $1tn worth of datacentres in just a few years will more than seemingly be all accelerated computing.”

Huang is making a bet that his imaginative and prescient for accelerating workloads with Nvidia GPUs will offer datacentre operators a methodology to boost computational vitality in their datacentres in a sustainable methodology while also using down the worth of computing, averting what he calls “computing inflation”.

He claimed that a liquid-cooled datacentre provides three to 5 cases the AI throughput when put next with the previous: “Liquid cooling is more cost-effective and lets you hang the worthy thing about this ability we name NVLink, which permits us to boost it to 72 Grace Blackwell packages, which has truly 144 GPUs.”