Former Facebook chief privacy officer Chris Kelly said the next stage of artificial intelligence development will center on efficiency, as the industry confronts rising costs and power demands tied to large-scale data center expansion.
Efficiency Emerges As Next Priority
Speaking Tuesday on CNBC’s “Squawk Box,” Chris Kelly said major AI companies will need to rethink how they build and operate infrastructure. He said current approaches rely on power-intensive systems that may not be necessary for advanced reasoning.
Kelly compared human cognition to the scale of modern AI infrastructure, saying the brain operates on roughly 20 watts and does not require gigawatt-scale facilities. He said identifying ways to reduce power and infrastructure requirements will become a key area of focus for large AI developers.
Kelly, who also previously served as general counsel at Facebook, said companies that achieve meaningful reductions in data center costs are likely to gain an advantage as the market matures.
Rapid Growth In Data Center Investment
The push to build AI infrastructure has driven a surge in data center investment. According to S&P Global, the data center market recorded more than $61 billion in infrastructure dealmaking in 2025 as hyperscalers expanded capacity worldwide.
OpenAI alone has committed more than $1.4 trillion to AI-related projects over the coming years. Those commitments include large partnerships with Nvidia, as well as infrastructure providers Oracle and CoreWeave.
Power Constraints And Grid Pressure
The rapid expansion has raised concerns about power availability, particularly as electric grids face existing constraints. In September, Nvidia and OpenAI announced a project involving at least 10 gigawatts of data center capacity.
That amount of power is roughly equivalent to the annual electricity consumption of about 8 million U.S. households. It also aligns closely with New York City’s peak summer electricity demand in 2024, according to the New York Independent System Operator.
Cost Pressure From Lower-Cost Models
Cost concerns intensified after DeepSeek introduced a free, open-source large language model in December 2024. The company said the model was developed for under $6 million, a figure far lower than spending reported by many U.S.-based competitors.
Kelly said he expects increased competition from Chinese AI developers. He pointed to President Donald Trump’s recent decision to approve sales of Nvidia’s H200 chips to China as a factor that could support that development.
Open-Source Access And Compute Availability
Kelly said open-source models, particularly those developed in China, are likely to provide broader access to basic levels of computing power and to generative and agent-based AI tools. He said this could widen participation in AI development while putting additional pressure on established players to improve efficiency.
Featured image credits: Freepik
For more stories like it, click the +Follow button at the top of this page to follow us.
