The Federal Trade Commission has invited public comments on the cloud computing industry and the practices of cloud computing providers. Here’s a brief introduction to the issue and our comment. For our full submission, click here.

Our response outlines that the effects of concentration in the cloud computing market and the development of artificial intelligence are profound and mutually reinforcing. The overwhelming trajectory of AI development over the past decade has trended toward larger and larger scale, creating dependencies on massive amounts of data and the computational power to process this data. These dependencies give significant market power to the small handful of companies which have amassed control over these resources – cloud infrastructure firms that also operate widespread platform ecosystems, and which have first mover advantage in large-scale artificial intelligence. This reinforces the findings of the House Digital Markets Report, which underscored that market leaders in cloud computing benefit from early-mover advantages due to network effects and high switching costs.

This holds true even as new start-ups begin to build commercial AI products: to enter the field, small companies must secure compute credits or make other contractual arrangements with Big Tech firms. These companies face significant barriers to entry if they were to seek to build these resources from scratch, due significant start-up costs, lack of interoperability at key points in the compute stack and bottlenecks in the supply chain for key components of compute infrastructure. Talent requirements also grow as compute costs grow, because very specialized knowledge is needed to make the most of scarce hardware, and much of this knowledge is tacit. Talent is thus a significant barrier to entry related to high compute costs.

These effects work in combination to more deeply entrench the infrastructural and economic power of the few firms that retain control over the key components to building AI, with detrimental effects on competition in the AI industry. This also contributes to consumer injury in many forms, including harms to privacy and security, encouraging the spread of false and misleading information, perpetuating patterns of inequality and discrimination, harmful effects on workers, and environmental harms.

Drawing on forthcoming work led by AI Now Research Fellow Jai Vipra, this response traces through the effects of concentration in cloud computing on artificial intelligence across three levels: the infrastructural dominance of cloud infrastructure providers, competition concerns that surface across layers of the technical stack, and harms related to the significant data and platform advantages of these companies. We conclude with a list of potential avenues for further consideration by the Federal Trade Commission and points of intervention:

Suggested points for further consideration:

  1. To address the most pressing concerns about current and future anti-competitive behavior in the AI market, the Federal Trade Commission should explore structural interventions targeted to the largest cloud infrastructure companies and their commercial AI model offerings. 
  2. The Federal Trade Commission should also enforce relevant antitrust laws against likely anti-competitive behavior, including restrictive licensing regimes and tying/bundling practices at different layers of the tech stack.
  3. The FTC should clarify the legality of data use for model training, including whether the use of data by cloud providers to train their AI models violates consumer control over personal data and whether such data use should be considered in the FTC’s commercial surveillance Advanced Notice of Proposed Rulemaking.

For our full submission, click here.