Investing in new AI infrastructure in 2022? My colleague Mike Gualtieri and I published a new Forrester Wave™ report: The Forrester Wave™: AI Infrastructure, Q4 2021. This is a brand-new topic for Forrester Wave evaluations. Its purpose is to dig into the massive explosion of AI that is quickly transforming enterprises. We looked at the top 13 vendors in this space: Amazon Web Services (AWS), Dell, Exxact, Google, Hewlett Packard Enterprise (HPE), IBM, Inspur Information, Lambda, Microsoft, NVIDIA, Run:AI, Spell, and Supermicro. We evaluated them across 25 different criteria. Although the report stands alone with its results and key takeaways, there are a few items we wanted to point out:

  • AI is not a single workload. There’s a lot of AI-washing when it comes to proving differentiation of various technology offerings. As such, it can be easily mistaken as one workload solution or one technology when it is not. In fact, AI acts as three different workloads: data management, training, and inferencing. We evaluated each of the vendor offerings based on these three pillars — some performing better than others depending on their origins.
  • Every vendor supports NVIDIA GPUs. There’s a resurgence of hardware innovation in AI. NVIDIA has established itself as a market leader for its deep learning training and in its graphics processing units (GPUs). In fact, every vendor in this evaluation supports NVIDIA’s GPUs, but don’t let the homogeneity in GPU support fool you. It’s still early days, and there’s been a ton of innovation coming from startups that are focused on the AI hardware accelerator market along with the cloud providers themselves.
  • Most solutions come with a software management layer. At first, it may seem like a “nice to have,” but gaining visibility into the various components of an AI technology stack is very difficult. The management layer that’s provided with many of these solutions enables easier use and better visibility across all components of the system. Features offered are similar to a traditional cloud management solution. Usual features include role-based access (RBAC), resource scheduling, and dashboard visibility.

There was some uncertainty on how to execute a Wave on a completely new topic, but the team came together and was able to produce an amazing evaluation on this burgeoning market. Also, keep in mind that the devil is in the details. The upper-right corner may not necessarily be the best choice for your organization. You can easily download the spreadsheet and customize the results for your priorities. Forrester clients can read the full report on our website.

If you’d like to discuss these findings or the AI infrastructure landscape at large, please schedule an inquiry with us (inquiry@forrester.com). In the meantime, happy reading.