EU Antitrust Chief Declares the 'Entire AI Stack' Under Investigation as Brussels Targets Nvidia, Meta, and Google Simultaneously
Teresa Ribera signals the EU will examine every layer of AI infrastructure for competition distortions, from chips to cloud to models.
Overview
The European Commission is no longer content to pursue individual antitrust cases against technology giants one company at a time. On March 12, EU competition chief Teresa Ribera declared at Berlin’s International Conference on Competition that her regulators are examining the “entire AI stack” for signs of market distortion, according to PYMNTS, citing Bloomberg reporting. The announcement marks a strategic escalation: rather than treating chips, cloud infrastructure, foundation models, and consumer applications as separate markets, Brussels now views them as interconnected layers where dominance at one level can entrench power across the rest.
The declaration arrives alongside active investigations into Nvidia’s GPU sales practices, Meta’s restrictions on rival AI chatbots within WhatsApp, and Google’s use of publisher content for AI training. Taken together, they represent the most comprehensive antitrust campaign the EU has ever mounted against the artificial intelligence industry.
What We Know
The Full-Stack Approach
Ribera’s remarks at the Berlin conference signaled a departure from the EU’s traditional case-by-case enforcement model. “We are looking at the entire AI stack,” she said, with a focus “not just on final applications but also on underlying models that power them, the data the models are trained on and the cloud infrastructure and energy sources at their foundation,” according to PYMNTS. She warned that large technology firms risk using their positions to “entrench corporate power” across emerging AI segments.
The Commission is also weighing whether to expand the Digital Markets Act to cover cloud computing services, which would subject major cloud providers to the same structural obligations that already apply to gatekeepers in search, app distribution, and messaging.
Nvidia’s GPU Dominance
Nvidia holds an estimated 84 percent of the market for advanced graphics processing units, the chips that underpin virtually all frontier AI training and inference. EU investigators have distributed questionnaires to Nvidia’s competitors and customers to determine whether the company ties GPU purchases to networking equipment, effectively bundling products in a way that forecloses rivals. Nvidia denies wrongdoing, stating that its “products compete on merit and support customer choice.” France’s antitrust authority is running a parallel investigation, with potential charges pending. If violations are established, the company faces fines of up to 10 percent of its annual global turnover.
Meta’s WhatsApp AI Reversal
In October 2025, Meta announced a policy prohibiting third-party AI chatbot providers from accessing WhatsApp’s Business API, a move that would have blocked services including ChatGPT and Claude from reaching WhatsApp’s more than two billion users. The European Commission opened a formal investigation in December 2025, citing concerns that the ban constituted an abuse of dominant position under Article 102 TFEU, given that WhatsApp holds over 90 percent market share in several European countries while Meta simultaneously promotes its own Meta AI chatbot on the platform.
Under regulatory pressure, Meta reversed course on March 6, agreeing to a 12-month trial period during which rival AI providers can operate on WhatsApp in Europe, according to TechCrunch. The concession comes with per-message fees ranging from EUR 0.0490 to EUR 0.1323 depending on the country. The Commission has stated it is analyzing whether this pricing model constitutes a continuation of anticompetitive conduct by financial rather than technical means.
Google’s AI and Publisher Content
A separate investigation examines whether Google unfairly uses publisher content for its AI Overviews feature without proper consent or compensation. Research cited by the Commission indicates that AI-generated summaries have caused up to 80 percent fewer clickthroughs to original news sources, according to TechPolicy.Press. The probe also extends to Google’s exclusive use of YouTube content for AI model training, raising questions about whether vertical integration in data supply creates unfair advantages.
What We Don’t Know
Ribera said a decision on whether further regulatory intervention is needed would come “soon,” but provided no specific timeline. It remains unclear whether the Commission will open a single overarching investigation into AI market structure or continue pursuing individual cases that share a common analytical framework.
The scope of any DMA expansion to cloud computing is also undefined. Amazon Web Services, Microsoft Azure, and Google Cloud would be obvious candidates for gatekeeper designation, but the Commission has not indicated which providers it considers systemically important in the AI context.
Whether Meta’s per-message fee model will satisfy regulators is an open question. Consumer groups and rival AI providers may argue that charging up to EUR 0.1323 per message creates an economic barrier functionally equivalent to an outright ban, particularly for startups without the capital to absorb messaging costs at scale.
Analysis
The EU’s full-stack approach reflects a recognition that AI market power does not operate in silos. A company that controls GPU supply can influence which cloud providers thrive; a cloud provider that controls infrastructure can shape which foundation models gain traction; a platform that controls distribution can determine which AI applications reach consumers. By examining these layers simultaneously, the Commission is attempting to prevent the kind of vertical integration that took decades to unwind in telecommunications and that it is still struggling to address in digital advertising, where it fined Google EUR 2.95 billion without achieving structural separation.
The strategy also positions the EU as the de facto global regulator of AI competition at a moment when the United States has moved in the opposite direction. The Trump administration has signaled that federal antitrust enforcement will focus on perceived ideological bias in technology platforms rather than market structure, and has directed the Commerce Department to identify state AI laws that “merit referral” to a litigation task force.
For technology companies, the practical consequence is a regulatory environment where compliance costs are rising across multiple jurisdictions and multiple layers of the technology stack simultaneously. Nvidia must defend its bundling practices; Meta must price WhatsApp access in a way that satisfies antitrust scrutiny without conceding its platform advantage; Google must navigate both advertising remedies and emerging AI content disputes. The companies that manage these overlapping pressures most effectively will not necessarily be the ones with the best technology, but the ones with the most adaptable business models.