Building a Better Grid

AI for DC

Dec 09, 2025

Hi

Welcome (back) to The Prompt. AI isn’t well understood, but we learn a lot in our work that can help. In this newsletter, we share some of these learnings with you. If you find them helpful, make sure you’re signed up for the next issue.

[News] Building a better grid

We just introduced our Grid Flexibility Initiative – an approach to building AI data centers as assets capable of reducing their energy consumption on short notice when there is a strain on the grid, such as during times of extreme heat or cold. This process frees up energy resources for other users on the grid, including consumers.

It’s important to acknowledge where the American public currently stands on data centers. Opposition is growing, and people are more likely to believe that they increase electricity costs than that they create good-paying jobs. At the same time, we’re at a defining moment for America’s electric grid and industrial future. AI has transformed electricity from a commodity input into a strategic national asset. In the global race to add more power to sustain the advancement of AI, the momentum is with China. In 2024, it added new electricity capacity amounting to about one third of the entire existing US grid. It has 34 nuclear reactors under construction; the US has none.

This gap isn’t just about energy stats – it reflects a growing gap in the physical infrastructure on which America’s future economic growth, technological leadership, and strategic power depend.

That’s where our Grid Flexible data centers come in. These won’t be yesterday’s data centers, and here’s how they work. Today’s grid was built to meet the highest or “peak” demand, which only happens a few days a year. Building to constantly meet a level of demand that’s rarely reached means infrastructure building costs that are passed on to consumers. AI data centers that can reduce their consumption during peak periods means less power infrastructure needs to be built and consumers will be saved from related rate increases. As a recent financial analysis has shown, by powering down when not needed, Grid Flexibility can unlock substantially more capacity for consumer use.

Facts, unfortunately, get overshadowed by increasingly popular but inaccurate narratives that AI data centers always cause power price increases and hog water. A recent analysis from the Lawrence Berkeley National Laboratory found that the biggest drivers of rising electricity costs were inflation and natural-disaster mitigation. Another study showed that data centers consumed just 0.2% of the nation’s fresh water in 2023 – significantly below what agriculture, mining, and golf courses used – John McCarrick, OpenAI’s Head of Global Energy Policy

[Insight] How AI is creating economic value

OpenAI’s new report, The State of Enterprise AI, highlights how AI is creating economic value by supporting workers’ everyday productivity via ChatGPT Enterprise and helping businesses develop custom applications via the API.

Most economically valuable work happens inside firms, which makes enterprise adoption an important gauge of how AI is affecting productivity, growth, and competitiveness. The report draws on anonymized aggregated usage data from our more than 1 million business customers, as well as a survey of approximately 9,000 enterprise workers. Its key findings:

Enterprise AI usage is both speeding up and taking deeper hold. Enterprise message volume has grown eightfold year over year, and API reasoning tokens per organization are up 320x. These are not marginal changes. They suggest that AI is shifting from an experimental tool to a core input in how firms generate output—particularly in knowledge-intensive tasks where small efficiency gains compound at scale.

The productivity returns are already visible in day-to-day work. Enterprise users report saving 40-60 minutes per day when using ChatGPT. Time savings alone, however, understate the economic significance. ChatGPT is also changing the types of tasks people complete. Roughly 75% of enterprise users say ChatGPT enables them to perform new or more technically complex tasks than they previously could. AI is not only accelerating existing work, but expanding the effective skill set of workers inside firms.

Adoption patterns suggest that AI is increasingly globalized with business customer growth in Australia, Brazil, the Netherlands, and France now exceeding 143% year-over-year. AI is also growing rapidly across industries, with the median sector experiencing 6x customer growth in the past 12 months. While early uptake was concentrated in technology and professional services, some of the fastest recent growth is now occurring in manufacturing and healthcare, industries that started from a smaller AI base but where productivity gains can translate into substantial downstream effects.

There’s a widening gap between “frontier” and median firms and workers. Frontier workers are sending 6x more messages, and frontier firms are sending 2x as many messages per seat as median firms. The report also highlights several case study examples of leading AI firms whose deployments are contributing to improved business outcomes. Across technology, retail, finance, health and more, these customers report that AI is associated with revenue growth, improvements in customer experience, automation of manual processes, and faster product development.

You can read the full enterprise report here. – OpenAI Economic Research

[About] OpenAI Forum

Explore Forum programming by and for our community of 60,000 AI experts and enthusiasts from across tech, science, medicine, education, government, and other fields.

6:00 PM – 7:00 PM ET on Dec 15

[About] OpenAI Academy

The Academy is OpenAI’s free online and in-person AI literacy trainings for beginners through experts.

OpenAI has called for a nationwide AI education strategy – rooted in local communities in partnership with American companies – to help our current workforce and students become AI-ready, bolster the economy, and secure America’s continued leadership on innovation.