Google Cloud launches two new AI chips to compete with Nvidia - BERITAJA
Google Cloud launches two new AI chips to compete with Nvidia - BERITAJA is one of the most discussed topics today. In this article, you will find a clear explanation, key facts, and the latest updates related to this topic, presented in a concise and easy-to-understand way. Read more news on Beritaja.
Image Credits:Google11:39 AM PDT · April 22, 2026
Google Cloud connected Wednesday announced that its eighth procreation of custom-built AI chips, aliases tensor processing units (TPUs), will beryllium divided successful two. One chip, named the TPU 8t, will beryllium geared for exemplary training and another, the TPU 8i, is aimed astatine inference.
Inference is the ongoing usage of models, aka what happens aft users taxable prompts.
As you mightiness expect, the company touts immoderate awesome capacity specs for these caller TPUs compared to the erstwhile generations: up to 3x faster AI exemplary training, 80% amended capacity per dollar, and the expertise to get 1 million+ TPUs to activity together successful a azygous cluster. The upshot should beryllium a batch much compute for a batch little power — and costs to customers — than erstwhile versions. It calls these chips TPUs, not GPUs, because its civilization low-power chips were primitively named Tensor.
But Google’s chips are not a afloat frontal battle connected Nvidia’s future, astatine slightest not yet. Like the different elephantine unreality providers, including Microsoft and Amazon, Google is utilizing these chips to supplement the Nvidia-based systems it offers successful its infrastructure. It is not flat-out replacing Nvidia. In fact, Google promises its unreality will person Nvidia’s latest chip, Vera Rubin, disposable later this year.
One time the hyperscalers building their ain AI chips (which includes Amazon, Microsoft and Google) whitethorn turn to request Nvidia less, arsenic enterprises move their AI needs to their clouds and larboard their apps to these chips.
Still, arsenic things guidelines today, it’s not profitable to stake against Nvidia. As notable spot marketplace expert Patrick Moore jokingly posted connected X, he had predicted that Google’s TPU could beryllium bad news for Nvidia (and Intel) backmost successful 2016 erstwhile the hunt elephantine launched its first one. Nvidia is now a about $5 trillion marketplace headdress company, meaning that prediction didn’t precisely clasp up to the trial of time.
If each goes according to Nvidia’s plan, Google’s maturation arsenic an AI unreality supplier would consequence successful much business for the spot shaper not less, moreover if galore a workload runs connected Google’s chips.
Techcrunch event
San Francisco, CA | October 13-15, 2026
In fact, Google besides says it has agreed to activity pinch Nvidia to technologist machine networking that allows Nvidia-based systems to execute moreover much efficiently successful its cloud. In particular, the 2 tech giants are moving to beef up the software-based networking tech called Falcon, which Google created and unfastened originated successful 2023 under the godfather of each unfastened root information halfway hardware organizations, the Open Compute Project.
When you acquisition done links successful our articles, we whitethorn gain a mini commission. This doesn’t impact our editorial independence.
Subscribe
This article discusses Google Cloud launches two new AI chips to compete with Nvidia - BERITAJA in detail, including key facts, recent developments, and important insights that readers are actively searching for online.