Slashdot: Microsoft’s CTO Hopes to Swap Most AMD and NVIDIA GPUs for In-House Chips

Oct 4, 2025

—

Source URL: https://hardware.slashdot.org/story/25/10/04/2142243/microsofts-cto-hopes-to-swap-most-amd-and-nvidia-gpus-for-in-house-chips?utm_source=rss1.0mainlinkanon&utm_medium=feed
Source: Slashdot
Title: Microsoft’s CTO Hopes to Swap Most AMD and NVIDIA GPUs for In-House Chips

Feedly Summary:

AI Summary and Description: Yes

Summary: Microsoft is transitioning its AI workloads from traditional GPUs to its proprietary accelerators to enhance cost efficiency in its datacenters. This move exemplifies a trend towards customized hardware solutions and signifies Microsoft’s commitment to optimizing its infrastructure for performance and security.

Detailed Description: The provided content highlights significant shifts in Microsoft’s approach to hardware, particularly in relation to AI and infrastructure security. The key points include:

– **Shift from GPUs to Homegrown Accelerators**: Microsoft aims to reduce reliance on Nvidia and AMD GPUs by transitioning to its own AI accelerators. This strategy is driven by the need for better performance per dollar, especially for a hyperscale cloud provider.

– **Focus on Cost Efficiency**: Kevin Scott, Microsoft’s CTO, emphasized that price-performance metrics are crucial for large-scale cloud operations. Microsoft is willing to explore various options to satisfy demand while focusing on their proprietary technology.

– **Development of Custom Hardware**: The introduction of the second-generation Maia accelerator suggests that Microsoft is serious about developing competitive solutions that might outperform existing GPU offerings in various performance aspects (compute, memory, interconnect).

– **Broader Hardware Initiatives**: Other than AI accelerators, Microsoft is also developing its own CPU, called Cobalt, and various platform security silicon that enhance cryptographic functions and secure data exchanges. This indicates an integrated approach to not only accelerate performance but also bolster security across its datacenters.

– **Implications for Security and Compliance**: The move towards custom silicon can lead to enhanced security postures, as Microsoft develops hardware that incorporates specific security features aimed at protecting infrastructure and data integrity.

This shift towards proprietary hardware is significant for AI security, infrastructure security, and the overarching strategy of managing performance and costs effectively within cloud environments.

1 10 2 24 3 4 5 a accelerator accelerators ads AGI AI AI accelerator AI accelerators AI security AI workloads All alt AMD and app Arch art as at ated by C centers chip chips CI CIA Cloud cloud environment cloud environments Cloud Operations Cloud Provider co commit commitment competitive compliance compute content cost cost efficiency Costs CoT CPU cross Crypto cryptographic cryptographic functions custom custom hardware custom silicon customized D data data exchange data exchanges data integrity datacenter datacenters de demand development DoT drive driven e effective efficiency end enhanced security environment environments exp feature features for function g Gen generation GPU GPUs graph gs H hardware hardware solutions high Highlight http HTTPS Hyper hyperscal hyperscale hyperscale cloud ICO implications implications for security in infrastructure infrastructure security initiatives integrity inter io Iron k Key l large Lead led Li Link load M Maia man memory metrics Micro Microsoft N no non Nvidia NVIDIA GPUs o of off on only ons operation operations OPM ops opt options ory oS oss other out over per performance performance metrics phi platform point post price pro proprietary proprietary technology ps R rate RCE re red Ro RoT row s Scale sec secure security security and compliance Security Feature security features security posture security postures shift Sig Silicon size solutions source specific Strategy T tech technology ted the to Tor TP transition UN US use V Ware Wi workload workloads x z