# Qualcomm's AI250 Attacks the AI Inference Memory Bottleneck | Durga Malladi Interview

Source: https://www.youtube.com/watch?v=acuSJeDRpIc
Recap page: https://rapidrecap.app/video/acuSJeDRpIc
Generated: 2025-11-11T15:10:16.641+00:00

---
## Quick Overview

Qualcomm is revamping its data center lineup with the new AI200 and AI250 solutions, which are designed to address AI inference memory bottlenecks by leveraging near-memory computing architecture that promises a 10x increase in effective memory bandwidth while maintaining low power consumption, targeting Tier-1 Cloud Service Providers (CSPs) and Hyperscalers with a flexible, modular toolkit approach.

**Key Points:**
- Qualcomm introduced the AI200 and AI250 chips, cards, and racks to target data center AI inference use cases, aiming to solve memory bottlenecks.
- The new architecture incorporates near-memory computing, projecting a 10x increase in effective memory bandwidth compared to previous generations.
- The AI200 is presented as a rack-level AI inference solution, directly liquid-cooled, and commercially available in 3Q26.
- The AI250 adds near-memory computing architecture and supports disaggregated inference, with sampling expected in 2027.
- Qualcomm emphasizes flexibility, allowing customers to mix and match components or use the full stack, supporting various frameworks and operating systems like Linux distributions.
- The company is maintaining an annual cadence for new data center solutions, following the Qualcomm 4000 (2025) and Qualcomm 4000 with innovative memory architecture (2027) roadmap.
- Durga Malladi notes that the modular toolkit strategy allows customers to cherry-pick components or use the entire stack, supporting both off-the-shelf models and custom development.

![Screenshot at 00:18: Durga Malladi introduces the new AI200 & AI250 Chips, Cards & Racks which target data center use cases, signaling the core focus of the interview on new hardware solutions.](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-00-18.png)

**Context:** Durga Malladi, SVP & GM of Edge & Data Center Technologies at Qualcomm, joins Sam Charrington of TWiML to discuss Qualcomm's revamped data center product lineup, focusing specifically on new AI-focused solutions designed for inference workloads in cloud and hyperscale environments. The discussion centers around how these new offerings, particularly the AI200 and AI250, address critical memory bandwidth constraints prevalent in modern AI processing.

## Detailed Analysis

Durga Malladi details Qualcomm's strategic move to revamp its data center offerings, specifically targeting AI inference workloads with the AI200 and AI250 product families. The key innovation is the introduction of near-memory computing architecture, which aims to deliver a 10x increase in effective memory bandwidth while maintaining low power consumption, overcoming a major bottleneck in current AI processing. The AI200 is positioned as a rack-level, directly liquid-cooled solution, slated for commercial availability in 3Q26. The AI250, expected to sample in 2027, further enhances this with near-memory computing and support for disaggregated inference. Malladi highlights that their Hexagon NPU IP, designed from the ground up for generative AI workloads, is the foundational unit. The go-to-market strategy involves flexibility, allowing CSPs and Hyperscalers to integrate specific Qualcomm components or use the entire stack, supporting major frameworks and Linux variants. This flexibility complements existing investments by enabling customers to mix and match hardware for optimal performance and TCO. The roadmap indicates an annual cadence of new product updates.

### New Data Center Lineup

- Introduction of AI200 and AI250 chips, cards, and racks targeting AI inference bottlenecks
- AI200 is a rack-level, direct liquid-cooled solution available in 3Q26
- AI250 adds near-memory computing and supports disaggregated inference, sampling in 2027

### Architectural Innovation

- Near-memory computing architecture projects a 10x increase in effective memory bandwidth with low power consumption
- Based on the Hexagon NPU, designed specifically for generative AI workloads
- The architecture is highly scalable from chip to rack level.

### Go-To-Market Strategy

- Targeting Tier-1 CSPs and Hyperscalers globally
- Focus on flexibility, allowing customers to use off-the-shelf models or develop custom solutions
- Ecosystem support includes various frameworks (PyTorch, ONNX, etc.) and Linux distributions.

### Roadmap Cadence

- Qualcomm commits to an annual cadence for data center solutions
- Roadmap shows Qualcomm 4000 (2025), Qualcomm 4000 with near-memory architecture (2027), and Next-gen Qualcomm SoC (2028) releases.

![Screenshot at 00:01: Sam Charrington introduces Durga Malladi, SVP & GM, Edge & Data Center at Qualcomm, for a briefing on the revamped data center lineup.](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-00-01.png)
![Screenshot at 00:18: The title card announces the focus: New AI200 & AI250 Chips, Cards & Racks Target Data Center Use Cases.](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-00-18.png)
![Screenshot at 00:34: The Qualcomm AI200 Rack specifications are displayed, highlighting features like direct liquid cooling and commercial availability in 3Q26.](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-00-34.png)
![Screenshot at 00:46: A diagram illustrates the internal architecture of the Qualcomm AI200, featuring the Hexagon NPU alongside the Oryon CPU and Adreno GPU.](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-00-46.png)
![Screenshot at 01:23: The screen shifts to the Qualcomm AI250 Rack specifications, noting its near-memory computing architecture and flexible solution for disaggregated inference.](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-01-23.png)
![Screenshot at 03:36: A timeline graphic shows the 'Evolution of Qualcomm Data Center Solutions' with an annual cadence roadmap from 2025 to 2028+.](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-03-36.png)
![Screenshot at 04:55: A diagram of the Qualcomm AI stack illustrates the modular nature of the toolkit, supporting various frameworks and operating systems.](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-04-55.png)
![Screenshot at 06:20: A diagram detailing the Qualcomm SoC architecture, showing the Orion CPU, Adreno GPU, and Hexagon NPU.](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-06-20.png)
![Screenshot at 14:02: A slide titled 'Qualcomm Data Center Solutions Roadmap' displays an annual cadence timeline featuring Qualcomm 4000 \(2025\), the innovative memory architecture \(2027\), and the next-gen Qualcomm SoC \(2028\).](https://ss.rapidrecap.app/screens/acuSJeDRpIc/00-14-02.png)