Qualcomm unveils AI chips to challenge NVIDIA's dominance
Qualcomm just announced two new AI accelerator chips—the AI200 and AI250—aimed at making data centers faster and more efficient for artificial intelligence tasks.
The AI200 lands in 2026, followed by the AI250 in 2027.
Both are built for handling big language models and complex, multimodal AI, with Qualcomm promising better memory capacity and lower costs.
Chips support up to 768GB of memory per card
Qualcomm's AI cards support 768GB of memory per card.
Both chips are designed for racks with liquid cooling and can handle serious workloads, drawing up to 160kW of power.
Qualcomm is teaming up with Saudi Arabia's Humain to roll out up to 200 megawatts of Qualcomm-powered AI infrastructure.
Qualcomm's focus on cost efficiency and energy savings could help
NVIDIA still rules the AI chip scene, holding over 90% of the market, and AMD is making moves too.
Still, Qualcomm's focus on cost efficiency, energy savings, and flexible deployment could catch the eye of companies looking to do more with less—especially as demand for smarter, greener data centers keeps growing.