Who’s winning the enterprise cloud race, why are they winning, and what might others do to change their own fortunes? Enterprises have been undergoing a kind of gestalt moment with the cloud, so it’s actually a good time to present them with other options, as well as to recognize that picking the right approach has been determining the fate of cloud providers for some time and will continue to do so.
The top-line question, the one about who’s winning in the cloud space, is hard to answer because it’s hard to define what “winning” means. The top cloud provider, according to Wall Street, is Amazon. The top cloud provider, according to enterprises, is Microsoft. The most strategic cloud provider is IBM. The most AI-centric is Microsoft. The most technically sophisticated is Google. The one to watch is Oracle. How different can you get?
My Wall Street friends tell me that Amazon’s cloud supremacy is based on the fact that AWS hosts more of the online services, both startups and mature public companies, than any other cloud. Online services, including and perhaps especially those you can characterize as related to social media, are the perfect cloud services. They are highly bursty in nature, making it expensive to build out to the capacity needed to cover peaks while still being at least somewhat economical in the valleys. They exchange information that has fewer governance rules than enterprise applications would. The business case for the cloud in this space is obvious, and what’s valuable above all is scope and stability of operation, and cost. Amazon has met those requirements from the first.
For enterprises, this has created a bit of mystery. Why, they wonder, is Amazon so great? Most bought into the view that “everything’s moving to the cloud”, which implied that somehow the business case for the cloud was universal. This is what resulted in a flood of disillusionment and stories of repatriation. Microsoft, who wasn’t selling to those online companies but to enterprises, had a better understanding of enterprise cloud needs, which is why enterprises have rated them at the top for five years running.
Microsoft’s challenge, according to enterprises, has become its lack of strategic influence among the larger enterprises, whose potential budget for the cloud is the highest. Strategic influence is important in influencing the launching of new projects, which obviously are key to maximizing and optimizing cloud utilization. This has helped companies like IBM and Oracle, who both have greater influence, gain cloud market opportunity.
OK, what I’m seeing is that Microsoft is likely to continue its expansion in enterprise market share, providing that current drivers remain dominant. If new projects come into the picture, either because market conditions drive buyers’ interest in them or players with strategic influence do the same, then things could change in favor of IBM and Oracle. New things, new differentiators, new outcomes.
AI is complicating the cloud picture too, and here we see a mixture of complex forces playing on all the vendors and cloud providers. Enterprises are most interested in the use of AI as an adjunct to current applications, especially ones related to business analytics. This has favored IBM and Oracle, who understand that space best. On the other hand, enterprise interest in AI overall could be classified as interest in the agent form of AI, which utilizes more specialized and limited machine learning or deep learning tools and is less reliant on large language models. IBM is reported to be “unbiased” with regard to cloud hosting and AI, Oracle somewhat favoring their own AI cloud hosting, in implementing agent strategies. None of the major cloud providers seem to be positioning for the broad interest in AI agents at this point, according to enterprises.
Right now, cloud AI is almost entirely generative AI of the type used in conjunction with search engines, chatbots, or email/document copilot elements. Self-hosted AI is used in the latter two missions, primarily to address data governance requirements. It would appear, based on enterprise comments, that if AI agents were offered as a service, either an SaaS-like component for general consumption or in the form of a platform (PaaS) on which enterprises could host their own agents, a business case could be made for cloud hosting. That doesn’t guarantee it would be successful, only that enterprises could consider the economics of a cloud AI agent in the same way as they increasingly expect to consider the economics of cloud hosting overall.
Microsoft is not seen as a leader in this, nor is Amazon. Google might have its own best shot at gaining market share here, since they are thought to have superior technical knowledge of the cloud and cloud-to-premises relationships. However, Google doesn’t have any significant strategic influence with enterprises; IBM, Oracle, HPE, and Dell all score higher and Amazon and Microsoft around the same as Google. Google also appears to be too smart to engage effectively with enterprises whose access to cloud-skilled personnel is limited.
It appears that there is an opportunity for cloud providers to exploit with respect to AI agent hosting, but that they’re not yet exploiting it. That may sound surprising, but the problem seems to be one of “AI literacy”. Enterprises in general have not been aware of the agent AI model for much more than a couple of months, and so have not been exploring its potential. Stories on agent AI have, IMHO, been short on facts and long on hype, and so have not supported effective planning. IBM and Oracle, according to enterprises, are now presenting at least a realistic view, but even for them there’s still the inertia of enterprise strategic planning to deal with. It will likely take until 2H25 for many agent-related projects to get started, and enterprises will likely first explore the way agent AI is used before they worry about where it’s hosted.
Cloud providers, given their relatively low level of enterprise strategic influence, would likely be best served by exploring options for the general-consumption AI agent-as-a-service concept. There are plenty of areas in finance and business analytics that could benefit from an expert agent’s help, and analytic missions generally can tolerate latency better than missions in process control, for example.
