Key facts
- Qualcomm is launching new AI accelerator chips, AI200 and AI250, for data centers.
- The chips are designed to compete with NVIDIA and AMD in the AI infrastructure market.
- Qualcomm's strategy focuses on inference efficiency and cost/power advantages.
- Saudi Arabia's Humain has committed to deploying Qualcomm's inferencing systems in its data centers.
- The move represents a significant diversification for Qualcomm beyond its core smartphone chip business.
Qualcomm is making a significant push into the lucrative AI data center market with a new lineup of accelerator chips designed to challenge established players like NVIDIA and AMD. The company announced its AI200 and AI250 chips, slated for release in 2026 and 2027 respectively, which are engineered for inference efficiency and aim to offer cost and power advantages. This move marks Qualcomm's return to the data center processor space, leveraging its acquisition of Nuvia and its expertise in neural processing units from smartphone chips.
NVIDIA currently dominates the data center GPU market, essential for training large language models, while AMD is a key competitor. Cloud providers like Amazon, Microsoft, and Google are also developing their own custom silicon. Qualcomm's strategy focuses on inference, the process of running AI models, and claims its rack-scale designs can reduce total cost of ownership for cloud providers through optimized memory and power efficiency. The company has already secured a commitment from Saudi Arabia's AI firm Humain to deploy its inferencing systems, signaling global ambitions.
This diversification is crucial for Qualcomm as it seeks to reduce its reliance on the smartphone market, where some manufacturers are developing their own processors. The data center market is experiencing explosive growth, driven by AI infrastructure demand, with McKinsey estimating nearly $6.7 trillion in capital spending through 2030. Qualcomm's entry is expected to intensify competition among hyperscalers and AI labs.