Context: The UK Competition & Markets Authority (CMA) points to the importance of cloud computing resources as an input to AI services when arguing that the cloud services market required regulatory attention, though even Google, which hopes to benefit from the investigation, does not see AI as a major driver of demand in that field (July 8, 2024 ai fray article).
What’s new: Last week, the CMA updated the cloud market investigation webpage with additional stakeholder submissions. At that stage, responses to a second round of working papers, including one on software licensing, were due. Google made its own filing, as did the Google-controlled CCIA and a CCIA-supported organization. Google would like the CMA to act as a price regulator for Microsoft software licenses. But a Microsoft filing disputes that there is a competition issue, given that whatever Google and Amazon pay to Microsoft in license fees is a small amount compared to their cloud revenues.
Direct impact: The next step is for the CMA to draft a decision, which it will then discuss with the parties concerned. At this stage of proceeding, the absence of evidence of a competition problem is unlikely to be cured. The question is then whether the CMA will recognize, consistent with a qualitative survey that was conducted on its behalf, that software licenses are only one of various factors influencing cloud service customers’ decisions.
Wider ramifications: The global technology industry and also other segments of the economy are watching the regulatory environment in the UK. It has been only about a year since a major public debate started over whether the UK was open or closed for business. A new government has recently been sworn in, and while it may be ideologically more favorable to aggressive regulation, it needs a prosperous economy and foreign direct investment.
Earlier this month, Microsoft settled a pending EU complaint by CISPE, an Amazon-funded lobbying group, over software licensing terms (July 11, 2024 CISPE press release). According to Bloomberg, Google made an attempt to dissuade CISPE’s members from resolving the dispute and offered a $470 million package.
Regulators don’t formally need complainants to launch investigations and impose remedies, but it generally doesn’t look good if there are no complainants left or if there is only one (Google) that is a frequent target of antitrust action, even in the cloud services context.
Google spent money on (at least) three submissions to the CMA on July 24, but the argument comes down to saying that Microsoft has market power in certain segments of the software market (in some of which that is actually debatable), allegedly uses that market power to move on-premises enterprise software installations to its Azure service, and that this is evidenced by an allegedly very high share of new UK cloud customer sign-ups on Azure. That last point is particularly interesting, given that Google always argues that its own offerings win on the merits, regardless of market share.
In some other ways, Microsoft’s submission on software licensing (PDF) is also very interesting. Its writing style is actually surprising: while submitted by a conservative law firm that would normally produce rather dry documents, it almost reads like a transcript of someone delivering oral argument in a U.S. jury trial. Just a couple of examples:
“[I]t is also clear that there is a spectrum and no bright line or threshold at which point one can say: ah-ha, when Microsoft IP-based services amount to this percentage of total costs, it becomes a separate downstream market.”
“Cash is the ultimate way to compensate for a disadvantage (such as not having your own relevant IP), but Amazon and Google have ample margins (cash) to fund discounts that benefit customers. Is that effective competition in a well-functioning market? Yes it is.”
Google or anyone else who would like to complain, or the CMA if it wants to take action even in the absence of a complainant, faces considerable hurdles.
First, only because Microsoft does license its software (Windows Server, Office etc.) to cloud rivals doesn’t necessarily mean it has to. Is there a duty to deal? If so, the hurdle under the law is high, especially when intellectual property rights are involved. And if Microsoft had to do so, what would that mean for Amazon’s non-public Linux fork that is made available to its cloud customers, but not to other cloud providers? What about data that Google offers only to its own cloud customers?
Second, even if one simply skipped that first step of the analysis, Amazon and Google would have to be strongly disadvantaged because of Microsoft’s licensing terms. The legal standard is not that they’d have to be forced out of the market, but what doesn’t count is that they’d like to save money on a rather limited cost of doing business.
The Microsoft filing explains that if one car maker also made speedometers and sold them to other car makers, the relevant market in which those car makers have to compete is the market for cars, not the market for speedometers. Similarly, Amazon and Google offer cloud services with all that it entails, as opposed to just being software resellers.
The public redacted version of the filing hides how much Amazon and Google pay for Microsoft software licenses. But what’s clear is that it’s a small number compared to the $80 billion that Amazon generates in AWS revenues every year and Google Cloud Platform’s $26 billion.
Above those numbers (on page 5 of the document), Microsoft’s lawyers (though the author may have been an economist in this case) say this:
“Why then all the fuss? While the input cost numerator is such a small percentage of their costs and revenues, because the cloud denominator is such a huge market, the absolute numbers are worth lobbying regulators.”
They are telling the CMA not to allow itself to be used as a price reduction tool by the likes of Google.
Cloud customers will always on total costs. There are many levels, not just software licensing terms, that allow Amazon and Google to compete.
The CMA now has to choose between relying on only words or drawing the appropriate conclusions from numbers.