Context: In October 2023, Ofcom (Office of communications; the UK’s telecommunications regulator) made a market investigation reference to another UK government agency, the Competition & Markets Authority (CMA), relating to the cloud services market (Ofcom website). Ofcom concluded its own investigation by asking a sister agency to pick up the thread and take it from there, with a competition or antitrust focus.
What’s new: On May 23, 2024, the CMA updated the webpage related to the cloud services market investigation (which investigation resulted from the Ofcom referral and has a statutory deadline in April 2025). In addition to a general progress update (PDF), there is a qualitative customer research report (PDF). Simultaneously, the CMA released three working papers: on the competitive landscape (PDF), on egress fees (PDF) and on committed spend agreements (PDF).
Direct impact: Taken together, the documents reflect the absence of evidence, if not even in some respects the evidence of absence, of serious competition issues in the cloud services market. But it is too early to tell whether the CMA will nevertheless seek to intervene. The picture will become far clearer in a few months when the CMA releases its provisional decision (slated for September or October 2024).
Wider ramifications: In the particular context of merger reviews, the question of whether the UK is “open for business” has come up before (May 21, 2024 ai fray article).
ai fray has carefully studied the slew of documents. There are some references to AI, such as to the supply of specialized chips and some large cloud companies’ development of their own AI chips. AI is also seen as an increasingly important application of cloud services. But there is no AI-specific issue. The CMA focuses on application-agnostic aspects of the cloud business. Two of them involve discounts: committed spend agreements and software licensing. There is also a general concern over potential customer lock-in for technical reasons.
If the CMA’s objective is to spur competition and innovation in AI systems and services, intervention in the cloud service market is not a suitable lever, at least not unless and until there is evidence of competition or innovation in AI services and technologies being hindered by insufficient access to cloud services. The CMA recently also published a working paper on AI-related partnerships, and it actually shows that AI startups have major opportunities and can choose among various potential “Big Tech” partners (April 14, 2024 ai fray article).
“Progressives” (with respect to competition enforcement) have a point when they say that regulators often did too little, too late. But at the moment there is a risk of solutions being developed that are in search of actual problems. Partnerships between large technology companies and AI startups appear, at least so far, to just grow the market and enable more effective competition. In the cloud services market, there are a few very large players (“hyperscalers”), but there is also a competitive tier 2 (with companies like IBM and Oracle), and in any event, prices are going down (despite inflation affecting the economy at large). When customer lock-ins are exploited, the opposite is typically the case: prices then tend to go up.
With the new Digital Markets Competition and Consumers (DMCC) Act of 2024 having been adopted by the British legislature (Parliament website), the CMA now has a more powerful tool to tackle some pressing issues, such as the abuse of gatekeeper power over closed ecosystems. It is at least unclear why potential intervention in the cloud services market would be a priority in the current environment, on the current facts.
The CMA’s qualitative market research report interestingly shows that cloud service customers do have choice, but they have different priorities and motivations. For instance, the interviews that the market research firm commissioned by the CMA conducted with cloud service customers showed that existing software licenses are just one of various factors that influence purchasing decisions, and if those licenses were to be considered a lock-in factor, one would have to consider other factors that dissuade customers from switching from, for instance, Amazon Web Services (AWS).
IBM’s offerings appeal to companies in some strongly regulated industries. Google uses the data stored in its cloud to optimize the same customer’s online advertising campaigns. There is no single winning formula: each of the major cloud providers has a unique strategy at the moment.
An open-minded market investigation would likely lead to the conclusion that “hands off” is the best policy for the time being, unless and until some parties’ conduct or the overall circumstances change so that intervention would be warranted. Sadly, there is an indication of the investigation being driven by a preconceived notion. Instead of the standard way to measure market share, the CMA also takes into account what capacity the different companies could make available. In other words, unsold inventory would inflate market share, an approach that no one would consider in traditional industries. Similarly, the CMA also places particular emphasis on the share of new customers that each market participant can sign up. All of that raises the question of whether the CMA has to resort to such novel criteria because it absolutely wants to impose remedies on a functioning market. In competition enforcement, gerrymandering is normally associated with market definition. Here, there appears to be some gerrymandering with respect to how each company’s market share is calculated.
In a few months’ time, the provisional decision will indicate whether the CMA considers regulation a means to an end or pursues regulation even where it may unnecessarily interfere with a functioning market and go against the UK’s economic policy interests. It’s hard to see how any of UK-based customers would benefit from “remedies” that prevent cloud service providers from offering certain discounts, for instance.