Alts News

Nvidia Chases $30 Billion Custom Chip Market With New Unit

Reuters is one of the world’s premier news agencies, offering text, video, images and data services to clients worldwide.

Nvidia plans to establish a custom chip design unit to cater to cloud computing companies and others seeking custom AI processors, according to nine sources familiar with its plan. This could help it maintain its dominant market position for data center equipment as customers explore alternatives.

Table of Contents

Toggle

AI

Artificial intelligence software that produces text passages, draws pictures and performs other tasks that resemble those done by humans has sparked an industry gold rush among tech firms. Some are opting to build their own custom chips rather than buying commercial AI processors like Nvidia’s $8,000 Lamborghini chip which now accounts for most of the high-end market.

Nvidia is taking steps to combat this trend with the launch of a unit dedicated to designing custom AI chips for customers such as Amazon, Meta, Microsoft, Google and OpenAI. Sources familiar with this matter report that talks have begun between Nvidia and these entities regarding producing custom chips.

Dina McKinney, formerly an Advanced Micro Devices and Marvell executive who now leads Nvidia’s custom chip division, reports her team is designing silicon for cloud computing, 5G wireless, video games and automotive use cases. Nvidia already holds leadership in data centers via their DGX servers that run thousands of inference tasks concurrently with training; additionally, their DRIVE auto-computing system runs inference tasks at vehicle edges.

Telecom

Nvidia’s graphic processing units were originally developed for gaming but have proved themselves better for computing-intensive tasks than standard Intel central processors in data centers, propelling Nvidia to become one of the highest market values among Amazon, Alphabet/Google parent Alphabet and Microsoft combined.

Now, Nvidia is making waves in telecom, where operators are using AI-on-5G services from operators like AT&T to accelerate problem-solving with less time compared to Intel processors, according to 650 Group analyst Wekel. AT&T uses Nvidia’s BlueField 2 A100 chip as part of its 5G Radio Access Network (RAN) and edge AI deployments; AT&T reports problems are being resolved within 10 seconds instead of 1,000 with Intel chips, as per analyst Wekel.

However, they may be slowing as other stocks related to AI gain the limelight and earnings are threatened by market turmoil. Nvidia stock fell on Wednesday but still increased by over 40% since 2022 – reflecting how AI-centric growth continues boosting value long after bubble burst.

Auto

Nvidia’s automotive efforts account for only a relatively minor portion of overall revenue; however, they have seen steady expansion as carmakers demand the technology for features like driver assistance and infotainment systems.

They debuted their latest car processor, Thor, at CES this week. This 2,000 teraflop chip aims to consolidate car AI cockpit functions onto one single system-on-chip and will be featured in vehicles produced by Chinese automaker Zeekr and others from 2024 onward.

Hopper graphics processing units will power the artificial intelligence software essential for self-driving vehicles, with them hoping to sell directly to automakers who will integrate it into their designs and send it off for manufacture and packaging at contract manufacturers.

Their announcement echoed similar moves by Intel and AMD to supply custom chips to automotive markets, and analyst Weckel of 650 Group predicts an expected annual increase of about 20% in automotive custom chip demand.

Gaming

Nvidia, best known for producing GPUs used in video games, is diversifying into other areas of the chip market and targeting its 20% annual growth to capture share in a $30 billion custom AI chip market.

They recently held an event at CES to unveil their latest graphics processing units, which allow gamers and designers to utilize AI on their personal computers without being dependent on remote services over the internet. Nvidia’s chips offer higher performance at lower costs.

Nvidia showcased several generative AI models that will bring realistic characters to games and apps, as well as its TensorRT GPU accelerators to enhance speech recognition performance.

Urumqi complex relied heavily on American chips from Nvidia and Intel, who sold their products directly to Sugon, the Chinese company operating the facility. Nvidia and Intel contend they do not have control over where their products end up; critics contend they should monitor sales more closely to identify potential repressive uses for them.


Exit mobile version