Industry leader predicts strategic shift from raw computational power toward AI efficiency and algorithmic optimization.

The artificial intelligence sector faces mounting challenges from rising energy consumption and infrastructure costs as massive buildouts continue to support advanced workloads. In a CNBC interview, former Facebook Chief Privacy Officer Chris Kelly noted that human cognition functions efficiently on just 20 watts of power, highlighting the potential for AI systems that could operate without dependence on gigawatt-scale facilities. He stressed that breakthroughs in lowering data center expenses will be critical in determining which companies emerge as industry leaders.

This push for efficiency comes amid substantial investments in supporting infrastructure. The data center market has seen over $61 billion in dealmaking in 2025, driven by hyperscalers pursuing extensive global expansions, according to S&P Global. OpenAI has committed over $1.4 trillion in AI-related obligations over the coming years, involving key partnerships with GPU provider Nvidia (NVDA) and infrastructure firms Oracle (ORCL) and CoreWeave (CRWV).

One notable initiative between Nvidia and OpenAI involves plans for at least 10 gigawatts of data center capacity, equivalent to the annual electricity use of 8 million U.S. households or comparable to New York City's peak summer demand in 2024, as reported by the New York Independent System Operator. These developments have intensified scrutiny over power availability, given pressures on existing electrical grids.

At the same time, innovations in cost reduction are emerging, exemplified by DeepSeek's release of a free, open-source large language model in December 2024, reportedly developed for under $6 million—far below expenditures by leading U.S. developers. Kelly anticipates increased prominence for Chinese AI entities, particularly following President Donald Trump's approval of Nvidia's H200 chip sales to the country. Open-source approaches, especially from China, are expected to democratize access to foundational computing resources for generative and agentic AI applications. As the industry progresses, prioritizing operational efficiencies and sustainable power management will be critical to maintaining momentum in AI advancement.