On 27 April 2026, Microsoft and OpenAI announced the next phase of their partnership. The headline framing was liberalisation: Microsoft's licence to OpenAI's models and products now runs through 2032 but is non-exclusive; OpenAI is free to ship on any cloud, with Azure merely "primary" rather than exclusive; revenue-share payments to Microsoft are capped and end in 2030; and the deal is no longer tethered to vague artificial-general-intelligence milestones. Sam Altman summarised it plainly: "Microsoft will remain our primary cloud partner, but we are now able to make our products and services available across all clouds."
That is a genuinely pro-competitive set of changes. But it lands directly inside an open question that Germany's Bundeskartellamt left on the table more than two years ago — and the regulator has every formal tool it needs to pull the thread.
What the Bundeskartellamt already decided
On 15 November 2023, the Bundeskartellamt closed a merger-control look at the Microsoft–OpenAI relationship without intervening — but its reasoning was anything but a clean bill of health. The authority concluded that Microsoft "gained material competitive influence on OpenAI as early as 2019, or at the latest when the partnership was deepened in 2021." It declined to require a merger notification only because OpenAI's German activities were too small at the time to trigger the domestic turnover thresholds.
Crucially, President Andreas Mundt added a standing caveat: "If Microsoft were to increase its influence on OpenAI in the future, we would have to re-examine whether the companies are subject to notification under merger control." That sentence is the hinge on which the 2026 analysis turns.
Less than a year later, on 27 September 2024, the Bundeskartellamt formally determined that Microsoft is an undertaking of "paramount significance for competition across markets" under Section 19a of the German Competition Act (GWB). The designation — published 30 September 2024 and valid for five years — lets the regulator pre-emptively prohibit anti-competitive conduct across Microsoft's entire ecosystem, from Windows and Microsoft 365 to Azure, without first proving market dominance the slow way. It is the German analogue to the EU's Digital Markets Act, but broader in reach.
Why the restructuring re-opens the file
The 2026 deal was pitched as Microsoft loosening its grip. On exclusivity and cloud lock-in, it genuinely does. But the two features the Bundeskartellamt cared about most are untouched or arguably reinforced. Microsoft retains its equity stake — roughly 27% of OpenAI Group PBC on an as-converted basis as of late 2025 — and it retains a long-dated IP licence through 2032. A capped-but-continuing revenue share through 2030 keeps the two firms financially intertwined for years.
In antitrust terms, "material competitive influence" does not require exclusivity. It can flow from a large minority shareholding, privileged access to a partner's core technology, and deep commercial entanglement — precisely the bundle the 2026 agreement preserves. A regulator could plausibly read the restructuring not as a reduction in influence but as its institutionalisation through 2032, which is exactly the contingency Mundt flagged in 2023.
Steelmanning the case for a fresh probe
The strongest argument for the Bundeskartellamt acting is straightforward and worth stating fairly: foundation-model markets are concentrating fast, the compute and capital required to compete are enormous, and a designated gatekeeper holding a quarter of the leading lab plus a licence to its models for the better part of a decade is exactly the structural entanglement Section 19a was written to police. If German enforcers wait for harm to materialise in a market moving this quickly, the remedy may arrive too late to matter. Pre-emptive scrutiny is the entire point of the 19a regime.
Where proportionality should win
That case is real — but on these facts it points toward monitoring, not intervention. The April 2026 changes move every visible lever in the pro-competitive direction: they dismantle cloud exclusivity, free OpenAI to deploy on AWS and Google Cloud, cap the money flowing to Microsoft, and sever the AGI trigger that made the old contract so opaque. OpenAI's parallel build-out — multi-billion-dollar commitments with Oracle, SoftBank and Amazon — is evidence that the lock-in the regulator feared is actively dissolving, not deepening.
A minority equity stake and a non-exclusive licence are ordinary features of technology partnerships, not antitrust violations in themselves. Section 19a's power to prohibit conduct "across markets" is potent precisely because it can be deployed on suspicion of structure rather than proof of harm; that makes restraint a feature, not a weakness, of good enforcement. The right posture is the one the Bundeskartellamt has, in fact, signalled — keep the matter under the 19a microscope, demand transparency on how the licence and stake are exercised, and act only if Microsoft uses them to foreclose rivals (for example, by degrading OpenAI's portability to the competing clouds it just won the right to use).
The broader test
The Microsoft–OpenAI file is becoming a template for how AI partnerships get regulated in Europe — the European Commission and the UK's CMA have circled the same relationship. Germany's instrument is the sharpest of the three, and how it reads a deal that is simultaneously more open (on cloud and exclusivity) and more durable (on equity and IP) will signal whether 19a is a scalpel or a hammer. Evidence-based, proportionate enforcement would treat the 2026 restructuring as what it largely is: movement in the right direction that earns continued watching, not punishment. The burden should sit with demonstrating foreclosure — not with assuming it from a cap table.