Axelera AI launched the Metis M.2 Max, a new addition to its Metis AI processor family, designed to handle compute-intensive inference workloads at the edge. The module delivers PCIe-level performance in the compact M.2 form factor, targeting large language models (LLMs), vision transformer networks, and other advanced AI applications. Shipments begin in Q4 2025 through Axelera’s webstore and channel partners.
The Metis M.2 Max doubles memory bandwidth over the existing Metis M.2, supports up to 16 GB of memory, and incorporates advanced thermal management and enhanced security. Customers can choose between standard operating temperature (-20°C to +70°C) and extended industrial range (-40°C to +85°C). The new card provides a 33% performance uplift for convolutional neural networks and doubles throughput for LLMs and vision-language models while staying within an average 6.5W power envelope.
Axelera designed the M.2 Max as a 2280 M-key module with an optional low-profile heatsink, reducing card height by 27% for tighter deployments. Built-in security features include firmware integrity protection and a Root of Trust for secure boot and upgrades. The module integrates with Axelera’s Voyager SDK, enabling simplified deployment of both proprietary and industry-standard AI models.
- PCIe-class AI acceleration in an M.2 2280 module
- Up to 16 GB memory with double the bandwidth of prior generation
- Performance boost: +33% CNNs, 2x tokens/sec for LLMs and VLMs
- Power efficiency: 6.5W average consumption
- Industrial-grade option: operating from -40°C to +85°C
- Security: Root of Trust, firmware integrity checks, secure boot and upgrades
- 27% slimmer profile with optional heatsink for constrained environments
“We continue to set the price-performance ratio benchmark for the AI accelerator market,” said Fabrizio del Maffeo, CEO of Axelera AI. “Our goal is to make it possible for our customers to deploy transformative edge AI applications at scale.”
🌐 Analysis: Edge AI hardware is becoming more critical as LLMs and transformer models move beyond the data center. Axelera’s move to pack PCIe-level performance into the M.2 format reflects the demand for scalable, power-efficient inference across industrial, retail, and healthcare environments. The company is positioning Metis as an alternative to GPU-centric solutions from Nvidia and AMD, with a focus on cost and power efficiency at the edge. This launch follows Axelera’s strategy of expanding the Metis platform while strengthening security, a trend mirrored by competitors such as Hailo and Mythic in edge inference markets.
🌐 We’re tracking the latest developments in semiconductors. Follow our ongoing coverage at: https://convergedigest.com/category/semiconductors/
