Arm has long operated its business as a behind-the-scenes architect, designing the chips that power nearly every smartphone in the world, and making money from royalties from the chips it designs for its customers.
Now, Arm is looking to change that by introducing its own AI chip, the Arm AGI CPU.
Arm CEO Rene Haas said at an internal conference on Tuesday that this major pivot wasn’t just a shift in internal strategy, but a direct plea from the world’s most powerful AI giant. The company has dropped the names of OpenAI and Meta as key partners for the chip.
“The biggest reason we’re doing this is because our partners have asked us to do it,” Haas said Tuesday.
The AI boom is creating major bottlenecks in data centers due to energy constraints and memory scarcity. Faced with this demand, Arm has moved to introduce AI chips that it claims are more energy efficient. Arm says it sees a $1.5 trillion market opportunity by entering AI chips for cloud, edge, and physical AI.
Arm stock rose more than 18% on Wednesday. Mizuho analysts wrote that they see “strong growth opportunities” for Arm in AI infrastructure and the auto industry. Bank of America research analyst Vivek Arya said in a note to investors that the company’s outlook may be “overly ambitious.”
Meta and OpenAI partner with Arm
Meta is building a large data center to power its apps and modern superintelligence business. Meta’s head of infrastructure, Santosh Janardhan, said on stage Tuesday that the upcoming Hyperion cluster could provide 5 gigawatts of power, enough to power 50 towns the size of Palo Alto.
“Even if you meet the performance, you won’t get the power. You might get the power, but you won’t get the performance,” Janardhan said.
This sparked an engineering project within Meta, with engineers “working around the clock” to port the system to Arm within three months, said Meta engineer Paul Saab.
“I didn’t ask my boss here for permission to buy these machines or even start the project,” Saab said on stage.
Saab says there were significant performance benefits, but there were no ARM chips available for purchase at the time.
OpenAI faced a similar problem. Its computing demands have increased significantly as it trains and runs ChatGPT models, AI coding tool Codex, and more.
“This is one of the things we hear a lot within OpenAI: We need more compute,” Kevin Weil, vice president of OpenAI for Science, said on stage, adding that we need more energy-efficient chips.
Arm said it expects the chip to generate $15 billion in revenue by fiscal year 2031.
The chip market is ‘getting very crowded’
In an analyst note, Arya said Arm faces the risk that the CPU market is “very crowded.” Other competitors such as AMD, Nvidia, and Intel have more CPU products and more established customers. Arya wrote that Arm’s chances for new CPUs could remain “limited,” especially since both Meta and OpenAI also work with AMD and Nvidia.
“Furthermore, the more AI grows, the more ARM’s smartphone/consumer market will come under pressure due to limited memory supplies,” Arya wrote.
That said, growing demand is causing many customers to turn to chip companies other than Nvidia for their computing needs. Both Meta and OpenAI are working with Broadcom to build AI chips.
The rise of AI agents has also increased the demand for inference, or the way AI models draw conclusions and make predictions. Nvidia’s core AI chip, the GPU, dominates AI training, but CPUs like Arm’s AGI CPU can also help with inference. Nvidia also recently entered this market.
Any tips? To contact this reporter via email, please specify the following address: rmchan@businessinsider.com, Or send a signal at rosal.13. Use a personal email address, non-work WiFi network, and non-work device. Here’s a guide to sharing your information securely.