Brazil opens another front: Amazon's investment in Anthropic comes under antitrust scrutiny

By MLex · July 8, 2026. Brazil's competition regulator (Cade) has begun examining whether the relationship between Amazon and Anthropic —which combines an equity investment, commitments to consume AWS cloud infrastructure and hardware agreements— constitutes, in practice, something more than the position of a minority investor. The information available is still preliminary: this is an inquiry, not a formal charge or a sanction, and the material contains no details on timelines, exact scope or possible remedies.

The fact itself, however, fits a pattern we already know well. Regulators in the United Kingdom, the European Union and the United States have asked similar questions about Microsoft-OpenAI and about Amazon-Anthropic itself: when does a strategic investment plus an infrastructure contract stop being a simple commercial relationship and start functioning as effective control, thereby bypassing the traditional merger filter? It is a legitimate question. The major cloud providers not only put money into generative AI labs: they also set which chips they use, how much compute capacity they have guaranteed and, ultimately, how much room to maneuver they have left against a client that is at once their main shareholder and their technological landlord.

That it is Brazil making a move now is significant in itself. Emerging markets have long watched how these cases are resolved in Washington and Brussels before acting, and an inquiry of Cade's own indicates that concern over the concentration of power in AI infrastructure is no longer exclusive to regulators of the global north. For Anthropic, which needs Amazon's compute muscle to train and serve its models at scale, any scrutiny touching that relationship is delicate: the company depends on that capital and that infrastructure continuing to flow without conditions that would complicate its editorial and technical independence from its investors.

Our reading is that these kinds of cases are the real regulatory battleground of the AI era, more so than the abstract discussions about superintelligence. As we had been noting when analyzing the rapprochement between Google and Microsoft against OpenAI and Anthropic, the underlying dispute is no longer only about which model is smarter, but about who controls the 'plumbing' —cloud, chips, capital— on which everyone depends. The more regulators, in more jurisdictions, begin to ask these questions, the more likely we are to see contractual adjustments (governance clauses, limits on compute exclusivity, transparency about decision-making) that will not halt the AI race, but will force it to operate with more checks and balances. In the long run, that kind of oversight is not an obstacle to the abundance that AI promises, but a condition for that abundance not to be captured by a handful of players that simultaneously control capital, the cloud and the models.