Rephrase and rearrange the whole content into a news article. I want you to respond only in language English. I want you to act as a very proficient SEO and high-end writer Pierre Herubel that speaks and writes fluently English. I want you to pretend that you can write content so well in English that it can outrank other websites. Make sure there is zero plagiarism.:
- Nvidia is in talks with tech giants like OpenAI and Google to build custom AI chips.
- That could pose a threat to custom chip competitors like Broadcom, Reuters reported.
As the AI sector booms, Nvidia has increasingly dominated the niche artificial intelligence chip market. That’s driven the chipmaker’s share price up more than 200% in the last year as companies pay up for its specialized semiconductors.
Now Nvidia, which had a market cap of nearly $1.8 trillion on February 9, is making moves to establish a foothold in the custom AI chip market — and it wants to work with the most powerful tech companies in the world to do just that.
It’s held talks with leaders from Meta, Microsoft, Google, and OpenAI to build custom chips for data centers, two sources told Reuters. The discussions come as Nvidia plans to build a new business division dedicated to designing such chips — and AI processors — for cloud computing companies, multiple sources familiar with Nvidia’s plans told Reuters.
Microsoft, Google, and OpenAI didn’t respond to a request for comment from Business Insider. Meta and Nvidia declined to comment.
Nvidia is knuckling down on creating custom chips in part because enterprises are already looking for semiconductors that address their specific needs — and going to competitors like Broadcom and Marvell Technology for them.
Instead of trying to get their hands on Nvidia’s expensive, highly coveted H100 and A100 graphic processing units, or GPUs, tech companies can work with Nvidia rivals to build chips that can lower their energy consumption, reduce costs, and save time on production.
“If you’re really trying to optimize on things like power, or optimize on cost for your application, you can’t afford to go drop an H100 or A100 in there,” Greg Reichow, general partner at venture capital firm Eclipse Ventures, told Reuters. “You want to have the exact right mixture of compute and just the kind of compute that you need.”
The custom chip market was worth $30 billion in 2023, one analyst told Reuters. And it’s expected to explode another $10 billion this year and double that in 2025, another said to Reuters.
Any move Nvidia into the custom chip market bodes ill for other manufacturers.
“With Broadcom’s custom silicon business touching $10 billion, and Marvell’s around $2 billion, this is a real threat” to them, Dylan Patel, founder of semiconductor research group SemiAnalysis, told Reuters. “It’s a real big negative — there’s more competition entering the fray.”
Nvidia isn’t just trying to work with major tech companies. According to Reuters, the chip giant has also been in talks with businesses in the telecommunications, automotive, and video game industries.
Nvidia’s success so far has partially been due to a limited supply of its chips, which companies have been scrambling to buy so they can build and train their own AI models. Companies like Meta have been stockpiling Nvidia’s GPUs to get ahead of competitors in the AI space.
To increase global supply, Sam Altman, CEO of OpenAI, announced this week that he aims to raise up to $7 trillion to address the shortage. The face behind ChatGPT is talking to potential investors such as the United Arab Emirates’ government to reach that goal.
Nvidia, too, has plans to address the supply-demand mismatch. Last October, the company said it would build so-called AI factories. CEO Jensen Huang said these will be able to support high-powered training efforts that he believes will make the AI behind cutting-edge machines like autonomous cars smarter. The company is also turning to its own AI to produce its own AI chips faster.
Do you work at Nvidia and have a story to share? Contact BI reporter Aaron Mok at [email protected] through a non-work email or encrypted messaging app Signal at 718-710-8200 using a non-work phone.