The rise of custom silicon for the cloud data center

Richard Brown
2 min readNov 28, 2023

Microsoft’s recent announcement of its Microsoft Azure Maia and Cobalt processors signals a major shift among the world’s leading cloud service providers towards the in-house development of custom silicon platforms for their data centers.

With these processors being built by TSMC, the move is a big nod to the deepening partnership between global cloud giants and Taiwan’s powerhouse semiconductor scene. Microsoft is not alone in this. AWS and Google Cloud also have custom silicon manufactured in Taiwan.

One of the key benefits of in-house processor development is that it reduces the reliance of the cloud service providers on merchant silicon suppliers such as Nvidia, AMD, and Intel and gives them increased control over their own supply chain. This is crucial in times of global chip shortages or trade disputes, ensuring a more stable and predictable supply of necessary components.

A second critical advantage of this approach is that it allows cloud service providers to reduce the long-term costs of running their data centers. By tailoring processor performance and power usage to match specific AI workloads, providers can boost efficiency and cut energy consumption on a massive scale, resulting in significant operational savings.

The development of proprietary processor technology also promises to equip Cloud Service Providers with a significant long-term competitive advantage because it allows them to tailor their hardware to their specific needs and specialized capabilities that are not available in commercial off-the-shelf processors. Security will no doubt be a major priority in this regard.

Of course, this does not mean goodbye to Nvidia, AMD, and Intel in the cloud data center. They are still very much in the picture, powering complex AI demands at a massive scale with their high-performance platforms. Indeed, Microsoft CEO Satya Nadella was keen to highlight the company’s plans to adopt the Nvidia H100 and AMD Instinct MI300X GPUs for the Azure Cloud during the announcement.

Even though custom processors will not replace standard industry platforms, they are set to play a starring role in the cloud data center of the future. The importance of the relationship between cloud service providers and the Taiwan semiconductor ecosystem will also continue to grow as they work to develop cutting-edge solutions that meet ever increasing AI performance needs.

--

--

Richard Brown

I live in Taiwan and am interested in exploring what ancient Chinese philosophy can tell us about technology and the rise of modern China.