The Zhitong Finance App learned that some media revealed that Microsoft (MSFT.US), the global leader in cloud computing, plans to launch a new customized Cobalt 100 server CPU to Microsoft Azure customers at the Microsoft Build Conference next week. As an open preview version, this means that Microsoft will fully open the call rights for the new self-developed server CPU, the Cobalt 100, to users of the Azure cloud platform at that time.

In an analyst briefing prior to the Build conference, Scott Guthrie, executive vice president of Microsoft's cloud computing and artificial intelligence business, directly compared Cobalt to Amazon AWS's self-developed server CPU Graviton series chips, which have been providing services to developers for several years. Guthrie said that Microsoft's chip will improve the performance of other ARM-based server chips on the market by about 40%. According to information, technology companies cooperating with AWS, such as Adobe and Snowflake, have begun using self-developed chips such as AWS.

Microsoft first released this self-developed server CPU chip called “Cobalt 100” in November last year. This new 64-bit chip is based on Arm architecture, has 128 cores, and is built using TSMC's latest 5nm process. It integrates up to 105 billion transistors and supports 12 DDR5 channels. Microsoft emphasized that compared with all current ARM architecture server CPUs on the market (currently mainly the AWS Graviton series), the performance is improved by at least 40%.

Therefore, it is basically foreseeable that in the future, the Blue Factory (Intel) and Red Factory (AMD) server CPU procurement scale will drop significantly. This also means that Microsoft may increasingly rely on self-developed CPU supply systems rather than external procurement in the field of server CPUs. This is similar to the strategy of Amazon's AWS cloud computing platform. AWS has launched the AWS Graviton series of self-developed server CPU chips, gradually replacing Intel and AMD server CPUs.

In addition to the Cobalt chip, Microsoft will also offer AMD's Mi300x AI accelerator to Azure customers next week. Although AMD (AMD.US) is the world's leading GPU manufacturer, the development and expansion of AI chips has always lagged behind Nvidia (NVDA.US), but as large cloud computing service providers begin to find alternatives to Nvidia's expensive and undersupplied AI chips, and AMD has begun to make some progress in AI chips by providing better software and hardware coordination system support, the AMD MI300X has now also become a popular basic hardware in the AI field. Guthrie described it as “the most cost-effective AI GPU currently available in Azure OpenAI cloud service products.”

Microsoft will also reduce the price of accessing and running large language models at next week's Build conference and preview a new “real-time intelligence system” that will allow real-time data streaming into Microsoft's data analysis system Fabric. The system will provide native Kafka integration and support AWS Kinesis and Google Cloud's Pub/Sub data flow systems.

Microsoft will also announce a partnership with Snowflake. Fabric will now support Snowflake's Iceberg format, which will enable “seamless interoperability with Snowflake cloud data warehouses and enable any data in Snowflake to be displayed in Fabric and vice versa.”

Microsoft and Amazon AWS have chosen self-developed CPUs one after another, and ARM racks are the biggest winners

ARM can be described as one of the most important chip companies. It sells licenses for the core instruction set architecture of almost all smartphone mobile chips in the world. ARM architectures are mainly used in the smartphone field, but now, ARM architectures are also appearing more and more frequently in the field of computers and AI server chips.

Nvidia's self-developed Grace CPU is based on the ARM architecture. Amazon's self-developed data center Graviton server processor also uses ARM architecture. Microsoft's latest self-developed customized AI chip Azure Cobalt 100, a server CPU, is also built on the ARM architecture and is specifically designed to run cloud computing workloads on Microsoft Azure cloud servers. As a result, taking advantage of this global AI investment boom, ARM's stock price rose as high as 66% in the first quarter.

Microsoft and Amazon AWS have chosen to embrace the self-developed server CPU model, mainly to gain greater advantages in the overall performance, operating costs, flexibility and sustainability of cloud servers, and to provide more optimized cloud computing services and AI solutions through deep integration of hardware and various proprietary application software.

For two of the world's top cloud computing giants, Microsoft and Amazon, self-developed server CPUs allow them to optimize for specific cloud computing needs, thereby achieving higher performance and efficiency in cloud and AI workloads. On the cost side, through self-developed chips, Microsoft and Amazon can reduce their dependence on external suppliers, optimize the cost structure, and maintain stable supply when the hardware market price or supply volume fluctuates greatly. Self-developed CPU chips also enable Microsoft and AWS to quickly respond to technological changes and market demand, deeply integrate their various proprietary application software (such as Azure OpenAI and Azure Machine Learning), and maintain overall cloud computing platform performance ahead of other vendors through customized design and rapid iteration.