VENDORiQ: Azure Cobalt and Accelerated VMs Highlights Cloud AI Economics

Why it’s Important

The race is on to provide high-performance and lower-cost Cloud compute for AI applications. Like its rivals, Microsoft is investing in custom silicon. The Azure Cobalt VM promises greater efficiency and optimised performance, particularly in Microsoft applications, which will help constrain the operational costs even as the use of AI and demanding applications continue to rise. Incorporating AMD’s MI300x AI accelerator, specifically optimised for GPT-4, means more efficient processing of AI-driven tasks and enhanced AI inferencing and training capabilities.  

It is the delta between the cost of provisioning and running such demanding workloads and the fees the market will accept that Microsoft will attempt to grow as quickly as possible, in order to reap the rewards of its heavy investments in OpenAI.

Who is Impacted

• AI/ML engineers: Enhanced capabilities for AI inferencing and training.
• Cloud architects: Increased options for optimising workloads in the cloud.
• IT operations managers: Improved efficiency and performance of critical applications like Microsoft Teams.

What’s Next?

• Assess the potential for integrating Azure Cobalt VMs into current workloads.
• Evaluate the benefits of AMD AI Accelerator VMs for AI-driven projects, particularly those using GPT-4.
• While the new VMs offer excellent performance for their cost, alternatives on other Cloud platforms are worth exploring. However, running a multi-Cloud environment adds additional cost and complexity.

Related IBRS Advisory

• Microsoft Azure Desktop Services – Getting the Best Bang for Your Buck
• VENDORiQ: Enterprises Can Now Build Their Own Copilot Using Azure AI Studio
• VENDORiQ: Microsoft Adds ChatGPT to Azure OpenAI Service