Qualcomm Launches New AI Chips, Competing with Nvidia
On Monday, Qualcomm announced the launch of two new artificial intelligence chips, entering a competitive arena with Nvidia, led by billionaire Jensen Huang.
This marks a significant change for Qualcomm, a company that has primarily focused on semiconductors for mobile phones and wireless communication instead of the demanding data centers that are fueling the AI boom.
Consequently, Qualcomm’s stock saw an 11% increase on Monday, ending the day at $187.68.
CEO Cristiano Amon, who was born in Brazil, indicated that the new chips will be available in liquid-cooled server racks. This design is aimed at helping clients, such as cloud service providers, reduce operational costs.
Nvidia and AMD also offer graphics processing units (GPUs) in full rack configurations. Systems like OpenAI’s ChatGPT rely on these powerful setups.
However, Qualcomm is up against some serious challenges. Nvidia dominates the data center market, holding around 90% of it, along with a market capitalization of $4.6 trillion.
Nvidia’s chips were instrumental in training OpenAI’s ChatGPT, widely regarded as a leading AI chatbot.
Interestingly, OpenAI has started to diversify its supply chain, announcing earlier this month plans to source chips from AMD and even consider acquiring a stake in the company.
The race in AI technology shows no signs of slowing down. McKinsey estimates that by 2030, around $6.7 trillion will be invested in AI infrastructure.
Qualcomm’s AI200 and AI250 chips are scheduled for release in 2026 and 2027, respectively, as stated by the company.
The new rack is designed to require 160 kilowatts, similar to what Nvidia offers.
Qualcomm’s chip-based accelerator card boasts 768 gigabytes of memory, which surpasses what is available from Nvidia and AMD.
As for pricing, Qualcomm hasn’t disclosed any specific figures for its AI chips, cards, or racks.
Durga Maradi, the company’s senior vice president and general manager of technology planning, mentioned, “Qualcomm AI200 and AI250 are designed for smooth deployment and rapid innovation.”
“Our rich software stack and open ecosystem support make it easier than ever for developers and enterprises to integrate, manage, and extend already trained AI models into our optimized AI inference solutions,” Maradi added.
The new AI chip from Qualcomm utilizes the hexagonal neural processing unit (NPU) seen in its smartphone chips.
Qualcomm aims to sell its chips and components separately, allowing customers more flexibility in designing their own racks. Maradi also hinted that Nvidia and AMD could potentially be interested in some of Qualcomm’s data center parts.
“What we’ve tried to do is allow customers to either take everything or say, ‘Let’s mix and match,'” he explained.
