Key Takeaways
- Lotus Microsystems introduces vStrata™, a pioneering platform for AI infrastructure designed to tackle electrical and thermal challenges simultaneously.
- The LSC0580 module, scheduled to ship in Q3 2026, achieves over 50% reduction in power conversion losses, significantly lowering energy costs for data centers.
- vStrata features a low-profile design, enhancing compute density while managing rapid load changes efficiently.
Innovative Solution for AI Infrastructure
Lotus Microsystems has announced the launch of vStrata™, a groundbreaking platform aimed at addressing significant power delivery and thermal management challenges faced by modern AI accelerators. As AI systems demand higher current loads, traditional architectures have struggled to keep pace. vStrata is engineered as a single co-integrated system that resolves both electrical and thermal constraints.
The first engineering samples, identified as LSC0580, are set to ship in Q3 2026. The industry is currently shifting toward Vertical Power Delivery (VPD) to mitigate last-inch distribution losses, yet most existing solutions treat power and thermal issues separately. vStrata breaks from this trend by leveraging proprietary silicon Power Interposer Technology (PIT), which delivers power directly beneath the processor while managing thermal loads concurrently.
Hans Hasselby-Andersen, CEO of Lotus Microsystems, emphasized the importance of aligning power delivery and thermal management to boost compute density and reduce the costs associated with AI infrastructure. The LSC0580 has already proven successful in tape-outs for leading xPU and AI infrastructure partners and is tailored for ultra-high efficiency, aiming to diminish power conversion losses by over 50%. This efficiency translates into significant savings, potentially reaching billions of dollars for the global data center sector.
vStrata’s architecture is built upon four core advantages:
-
Multi-Domain Co-Design: This approach integrates electrical, thermal, and mechanical designs into a unified structure, achieving superior thermal conductivity. It effectively manages heat at the load point, which can decrease operating temperatures by as much as 25°C under optimized conditions.
-
Compute Scaling and Efficiency: With the capacity to cater to kiloampere-class AI power requirements, the LSC0580 achieves top point-of-load efficiency of up to 96%. It minimizes power losses and thermal bottlenecks, thereby reducing overhead costs while enhancing sustained performance levels.
-
Low-Profile Vertical Integration: Utilizing advanced packaging technologies, vStrata supports ultra-thin designs with projections under 1 mm. By minimizing electrical distances between power and processors, the platform significantly enhances dynamic performance, handling load transients surpassing 10 A/ns without needing external capacitors.
-
Designed for Compatibility: vStrata seamlessly integrates with established Tier-1 reference designs and existing power management controllers, thus facilitating easier adoption and improving next-generation performance.
Yasser Nour, Co-founder and CTO of Lotus Microsystems, remarked that the traditional architecture was not equipped to effectively scale AI compute. The bottleneck lies in the system’s responsiveness to rapid and unpredictable load changes. vStrata endeavors to close the gap between power supplies and processors, ensuring timely current delivery while maintaining thermal behavior and transient stability.
Lotus Microsystems, founded in 2020 and based in Copenhagen, Denmark, specializes in developing vertical power delivery and thermal management solutions for next-gen compute systems. The company holds multiple patents focused on its innovative technologies, ensuring its commitment to overcoming the electrical, thermal, and mechanical challenges faced by contemporary data centers.
The content above is a summary. For more details, see the source article.