Microsoft Warning: Cloud services may experience interruptions if sufficient GPUs cannot be obtained.

On the evening of Thursday, July 27th, Microsoft updated a risk factor in its annual report, highlighting the importance of protecting the security of its data center graphics processing units (GPUs). The company has been increasing capital expenditures, including expenses for purchasing GPUs, to meet the growing demand for cloud-based artificial intelligence services.

Microsoft emphasized to investors that GPUs are a key component of its rapidly growing cloud business. In its annual report submitted to regulatory authorities on Thursday evening, Microsoft added content about GPUs to the risk factor of "potential service disruptions if the necessary infrastructure cannot be obtained."

Analysts believe that this statement reflects the increasing hardware requirements for providing cloud-based AI services to small businesses.

In Microsoft's fiscal year 2023 report, released earlier this week covering up to June, it stated:

Our data centers depend on licensed buildable land, predictable energy supply and network supply, and the availability of servers, including graphics processing units (GPUs) and other components.

This is one of the three paragraphs mentioning GPUs in the document submitted by Microsoft to regulatory authorities. It is worth noting that GPUs were not mentioned in last year's report. Other major tech companies such as Google, Apple, and Amazon did not use similar wording in their reports submitted to regulatory authorities.

Microsoft, which supports OpenAI, relies on Microsoft's Azure cloud services for the computation of ChatGPT and various AI models. Microsoft has also started using OpenAI's models to enhance existing products, such as Outlook, Word applications, and the Bing search engine, through generative AI.

These efforts by Microsoft in AI technology, as well as the strong market interest in ChatGPT, have prompted Microsoft to seek more GPUs than originally anticipated.

NVIDIA CEO Jensen Huang stated at the company's GTC developer conference in March:

I'm excited to announce that Microsoft is opening up Azure for private preview of its AI supercomputer powered by the H100.

Microsoft has already begun seeking sufficient computing capacity outside of its own data centers and has signed an agreement with CoreWeave, supported by NVIDIA, to rent GPUs to third-party developers, including Microsoft.

Meanwhile, Microsoft has spent several years developing its own custom AI processors. In addition to Microsoft, companies such as Google, Amazon, Meta, and others are also developing their own AI chips in an effort to reduce dependence on NVIDIA.

Microsoft's Chief Financial Officer, Amy Hood, stated in a conference call with analysts this week that Microsoft expects to continue increasing capital expenditures this quarter to cover the costs of data centers, standard central processing units, network hardware, and GPUs. At the end of January this year, when the AI frenzy was just beginning, Microsoft experienced a network outage that caused a paralysis of Azure services, affecting millions of people worldwide who rely on services like Teams and Outlook.

According to Microsoft's data, the outage of Azure, which has 15 million enterprise customers and over 500 million active users, could have a domino effect, impacting multiple services, as nearly all of the world's largest companies are using this platform.

Now, with Microsoft integrating AI tools into its cloud services, market analysts believe that this means Microsoft requires even greater computing power, while also making its cloud service infrastructure more vulnerable.