MIT Researchers Build Rapid AI Power Estimator

Data centers are becoming energy hogs. According to the Lawrence Berkeley National Laboratory, facilities will consume up to 12 percent of total U.S. electricity by 2028. That number alone should make any infrastructure planner pause. Now MIT and the MIT-IBM Watson AI Lab have released a tool that could help operators make smarter decisions before they flip the switch.

The researchers developed a rapid prediction tool called EnergAIzer. It tells data center operators how much power will be consumed by running a particular AI workload on a certain processor or AI accelerator chip. The method produces reliable power estimates in a few seconds. Traditional modeling techniques can take hours or even days to yield results (a problem that has plagued users for years, frankly).

Inside a data center, thousands of powerful graphics processing units (GPUs) perform operations to train and deploy AI models. The power consumption of a particular GPU will vary based on its configuration and the workload it is handling. Many traditional methods used to predict energy consumption involve breaking a workload into individual steps and emulating how each module inside the GPU is being utilized one step at a time. But AI workloads like model training and data preprocessing are extremely large and can take hours or even days to simulate in this manner.

Lead author Kyungmi Lee, an MIT postdoc, explained the core insight. AI workloads often have many repeatable patterns. Algorithm developers write programs to run as efficiently as possible on a GPU. They use well-structured optimizations to distribute the work across parallel processing cores and move chunks of data around in the most efficient manner. These optimizations create a regular structure, and that is what the researchers are trying to leverage.

The lightweight estimation model captures the power usage pattern of a GPU from those optimizations. But while the estimation was fast, the researchers found that it didn't take all energy costs into account. Every time a GPU runs a program, there is a fixed energy cost required for setting up and configuring that program. Then each time the GPU runs an operation on a chunk of data, an additional energy cost must be paid.

Due to fluctuations in the hardware or conflicts in accessing or moving data, a GPU might not be able to use all available bandwidth, slowing operations down and drawing more energy over time. To include these additional costs and variances, the researchers gathered real measurements from GPUs to generate correction terms they applied to their estimation model. This way, they can get a fast estimation that is also very accurate.

When the researchers tested EnergAIzer using real AI workload information from actual GPUs, it could estimate the power consumption with only about 8 percent error. That is comparable to traditional methods that can take hours to produce results. The user can provide their workload information, like the AI model they want to run and the number and length of user inputs to process, and EnergAIzer will output an energy consumption estimation in a matter of seconds.

She is joined on the paper by Zhiye Song, an electrical engineering and computer science graduate student; Eun Kyung Lee and Xin Zhang, research managers at IBM Research and the MIT-IBM Watson AI Lab; Tamar Eilam, IBM Fellow and chief scientist of sustainable computing at IBM Research; and senior author Anantha P. Chandrakasan, MIT provost and Vannevar Bush Professor of Electrical Engineering and Computer Science. The research is being presented at the IEEE International Symposium on Performance Analysis of Systems and Software.

Data center operators could use these estimates to effectively allocate limited resources across multiple AI models and processors, improving energy efficiency. In addition, this tool could allow algorithm developers and model providers to assess potential energy consumption of a new model before they deploy it. The user can also change the GPU configuration or adjust the operating speed to see how such design choices impact the overall power consumption.

Their method could also be used to predict the power consumption of future GPUs and emerging device configurations, as long as the hardware doesn't change drastically in a short amount of time. In the future, the researchers want to test EnergAIzer on the newest GPU configurations and scale the model up so it can be applied to many GPUs that are collaborating to run a workload.

Think about the physical reality of this. A data center operator sits at a terminal. They type in parameters. Within seconds, they see a number. No waiting. No running a full emulation that would heat up the room and drain the battery backup systems. They can iterate. They can compare. They can make decisions before committing to a deployment that might cost thousands in electricity.

But here's the thing. A faster estimation tool doesn't automatically mean lower consumption. It just means people know what they're spending before they spend it. Whether that knowledge translates into actual behavioral change remains the real question. The tool is available. The data centers are still hungry. The rest is up to the operators who hold the keys.

For more details on the research, see the MIT News announcement. Context on the broader AI energy landscape can be found in MIT Technology Review's analysis of the industry's carbon footprint.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn