AI power prediction; large-area FPCBs; graphene vibrations.
Researchers from MIT and the MIT-IBM Watson AI Lab developed a prediction tool that can quickly tell data center operators how much power will be consumed by running a particular AI workload on a certain processor or AI accelerator chip. It can be applied to a wide range of hardware configurations.
The lightweight estimation model captures the power usage pattern of a GPU based on the optimizations that software developers use. To improve accuracy, it includes allowances for additional costs and variances such as program startup and bandwidth bottlenecks based on real measurements from GPUs. When the researchers tested the approach using real AI workload information from actual GPUs, it could estimate the power consumption with only about 8% error.
“As an operator, if I want to compare different algorithms or configurations to find the most energy-efficient manner to proceed, if a single emulation is going to take days, that is going to become very impractical,” said Kyungmi Lee, an MIT postdoc, in a press release. “To really make an impact on sustainability, we need a tool that can provide a fast energy estimation solution across the stack, for hardware designers, data center operators, and algorithm developers, so they can all be more aware of power consumption. With this tool, we’ve taken one step toward that goal.” [1]
Researchers from the Korea Institute of Machinery and Materials proposed a roll-to-roll direct lamination process for large-area and long-length flexible printed circuit boards.
The team quantitatively analyzed how semi-cured adhesive films fill the gaps between circuit patterns under various process conditions, including speed and pressure, and identified conditions that enable stable filling even in continuous manufacturing environments. In particular, the team characterized the “filling behavior” of adhesive materials based on process variables to provide a basis for data-driven process optimization.
One key potential application is manufacturing flexible sensing cables for automotive and mobility. [2]
Researchers from the University of Birmingham demonstrated a new technique for creating nanosheets of graphene and other 2D materials such as hexagonal boron nitride and the semiconductors molybdenum disulfide and tungsten disulfide that runs at room temperature with increased production rates compared to current methods.
The vibrational exfoliation method causes graphite particles to fold at the edges, before splitting into thinner layered materials, peeling off the parent particle, and finally undergoing high strain rates in the liquid phase to form atomically thin sheets of graphene. Instead of toxic solvents, the method uses just water and tannic acid. Spectroscopic analyses showed the vibrational approach does not introduce defects into the graphene nanosheets.
“By creating alternate, more sustainable synthetic routes for these exciting materials, we have an opportunity to lower the barrier for industrial translation,” said Jason Stafford, an associate professor in the Department of Mechanical Engineering at the University of Birmingham, in a press release. “This will help facilitate future electronic devices, composites, and catalysts, while also avoiding unintended environmental consequences as production is scaled up.” [3]
[1] K. Lee, Z. Song, E. K. Lee, et al. EnergAIzer: Fast and Accurate GPU Power Estimation Framework for AI Workloads. IEEE International Symposium on Performance Analysis of Systems and Software. https://arxiv.org/pdf/2604.20105
[2] C. Lee, E. Gwak, D. Choi, et al. Scalable Roll-to-Roll Approach for Encapsulating Long-Length Flexible Printed Circuit Boards. ACS Applied Materials & Interfaces 2025 17 (36), 51131-51143 https://doi.org/10.1021/acsami.5c09839
[3] A. Rabani, F. A. Khaleel, F. S. Al-Gburi, et al. Vibrational Exfoliation of 2D Materials. Small (2026): e11652. https://doi.org/10.1002/smll.202511652
 |
 |
 |
 |
 |  |
 |  |
 |  |
Leave a Reply