AI Investment Deals Create Hidden Anticompetitive Ties

The artificial intelligence market appears fiercely competitive on the surface, with OpenAI, Google, Anthropic, Meta, and others racing to deploy frontier models. But beneath this activity lies a web of investment arrangements that may undermine genuine competition, according to legal analysis published by ProMarket.

The Federal Trade Commission has documented how major cloud providers are structuring investments in AI labs to secure exclusive commitments. Microsoft's funding of OpenAI and Amazon's investment in Anthropic both included requirements that the AI companies spend heavily on the investors' own cloud platforms—Azure and AWS, respectively. OpenAI even threatened antitrust action over what it viewed as overly restrictive cloud spending requirements before later signing a separate cloud deal with Amazon.

These arrangements effectively foreclose smaller cloud providers like CoreWeave from competing for AI lab business, regardless of their service quality or pricing. The investors' massive balance sheets let them bundle capital with infrastructure access in ways smaller competitors cannot match.

Why it matters

As AI development becomes more capital-intensive, these bundled deals may reshape market structure without traditional mergers or explicit tying agreements. Applications built on OpenAI or Anthropic models may align with their chosen cloud infrastructure, creating cascading lock-in effects throughout the AI stack. The result could be that all AI markets become effectively tied to a handful of dominant cloud platforms, raising barriers to entry across multiple layers of the technology ecosystem.

The Nvidia conflict

Nvidia presents a particularly complex case. The company dominates the supply of GPUs essential for training large language models while simultaneously holding equity stakes in OpenAI, xAI, and CoreWeave. This dual role creates potential conflicts of interest in a supply-constrained market where allocation decisions carry significant competitive consequences.

While no public evidence suggests Nvidia has favored portfolio companies in chip allocation, the legal duty to maximize shareholder value would make such conduct not just incentivized but arguably required. Under current market conditions, where demand for high-end training GPUs substantially exceeds supply, prioritization decisions directly affect which companies can train better-performing models and achieve commercial adoption.

The arrangement also threatens nonprofit and open-source AI development. Academic institutions and organizations like the Allen Institute for AI depend on access to cutting-edge hardware. If Nvidia prioritizes well-funded portfolio companies, it could concentrate AI capabilities within Big Tech.

The SpaceX-Anthropic arrangement

Anthropic's $45 billion compute deal with SpaceX raises different concerns. Elon Musk controls both SpaceX and xAI, which competes directly with Anthropic in developing frontier AI models. The arrangement gives SpaceX access to information about Anthropic's training schedules, model architecture choices, and research priorities.

Even if SpaceX never exploits this informational advantage, the potential for leverage may distort Anthropic's behavior. The company becomes dependent on infrastructure controlled by a competitor's owner. Alternatively, generating substantial revenue from Anthropic as a customer may incentivize SpaceX to avoid head-to-head competition that would benefit consumers but damage its own revenue stream.

Regulatory implications

These cases illustrate how proclaimed market competitiveness can mask arrangements that jeopardize genuine dynamism. Competition in AI extends beyond model capabilities to access to compute, capital, and talent. Deals that tie AI companies to their suppliers and customers along the supply chain distort this multi-layered competition.

Antitrust analysis must continue monitoring how these arrangements between direct competitors and supply chain partners risk foreclosing access to essential inputs and distorting competitive behavior, according to the analysis first reported by ProMarket and written by Shishene Jing, who previously worked on the FTC's AI industry investigation.