The next phase of the artificial intelligence (AI) industry's development will focus on improving the efficiency of technology applications, according to Chris Kelly, former privacy chief at Facebook. On Monday, Kelly noted that leading global AI companies are racing to build infrastructure to support AI computing demands, but the industry must optimize these energy-intensive construction models.
"The human brain operates on just 20 watts—we reason and think without needing gigawatt-scale power plants," Kelly said. "I believe exploring efficiency potential will become a core focus for top AI firms." He added that companies capable of reducing data center costs will ultimately emerge as winners in the AI sector.
According to S&P Global, large-scale cloud service providers have spurred a global data center construction boom in 2025, with infrastructure-related transactions exceeding $61 billion. OpenAI alone has committed over $1.4 trillion in AI investments in the coming years, including major partnerships with NVIDIA (NVDA.US), Oracle (ORCL.US), and CoreWeave (CRWV.US).
However, this rapid expansion has raised concerns about strained power grids' ability to meet the electricity demands of these computing infrastructures. In September, NVIDIA and OpenAI announced a joint project requiring at least 10 gigawatts of power—equivalent to the annual electricity consumption of 8 million U.S. households. Data from the New York Independent System Operator shows this matches New York City's peak summer demand in 2024.
Concerns over costs intensified after DeepSeek launched a free, open-source large language model in December 2024, developed for under $6 million—far below U.S. competitors' costs. Kelly expects "multiple Chinese companies to rise," especially after the recent U.S. approval of NVIDIA's H200 chip exports to China. He added that open-source models, particularly from China, will provide "basic-level computing power" and access to generative and agentic AI capabilities.
Comments