[REVISION HISTORY]
Evolution of AI infrastructure, power, and agent security
13 recorded revisions of this situation.
What changed
2026-10-07 08:51 UTC → 2026-10-09 03:27 UTC ·
added
removed
The technology sector is seeing significant developments in AI-driven infrastructure, power management, and agent security. NVIDIA has released its Alpamayo 2 Super autonomous driving reasoning model and proposed a transition to 800VDC power architectures to manage the increasing power density of GPU racks in data centers. To support these requirements, Renesas Electronics has expanded its gallium nitride (GaN) semiconductor portfolio. Following its acquisition of Transphorm, the company is sampling a 650V GaN device for 800V HVDC architectures in megawatt-scale AI racks, featuring a dual-side cooled package that reduces implementation area by 57% compared to traditional models. Mass production is slated for mid-2027. Renesas has also introduced a low-voltage launched its first family of 100V enhancement-mode (E-mode) GaN discrete power transistor series aimed at improving efficiency in humanoid robots and AI data centers. transistors. This new 100V lineup, consisting of four specific items, models, is designed for high-efficiency applications including AI data centers, humanoid robotics, and factory automation. The technology aims to reduce switching losses by 40% to 70% and potentially double system-level power density while maintaining compatibility with standard silicon density. To facilitate industry adoption of 800V DC distribution and 48V bus architectures, these transistors feature silicon-compatible footprints, allowing designers to migrate from conventional MOSFET packages. designs with minimal changes to existing PCB layouts. In a separate collaboration, Microchip Technology and Navitas Semiconductor have launched an 800V DC-to-6V DC reference design for AI data center rack power. Aligned with the Open Compute Project (OCP) 800V DC standard, this platform integrates Microchip’s digital power control and hardware-based security with Navitas’ GaNFast GaN devices. The design aims for a power density of 2,100 W/in³ and up to 96 percent peak efficiency, utilizing a 1 MHz switching frequency to allow close integration with GPU boards. This solution addresses the increasing power demands of high-power GPU clusters by combining two conventional conversion stages into a single converter. The companies plan to showcase the design at the 2026 OCP Global Summit. efficiency. In computing architecture, Arm-based rack-scale servers have surpassed x86 in the accelerated computing market.
Versions
- 2026-10-09 03:27 UTC Evolution of AI infrastructure, power, and agent security
- 2026-10-07 08:51 UTC Evolution of AI infrastructure, power, and agent security
- 2026-10-07 06:49 UTC Evolution of AI infrastructure, power, and agent security
- 2026-10-06 11:11 UTC Evolution of AI infrastructure, power, and agent security
- 2026-10-06 04:02 UTC Evolution of AI infrastructure, power, and agent security
- 2026-10-02 06:40 UTC Evolution of AI infrastructure, power, and agent security
- 2026-10-02 04:05 UTC Evolution of AI infrastructure, power, and agent security
- 2026-10-02 02:21 UTC Evolution of AI infrastructure, power, and agent security
- 2026-09-30 23:06 UTC Evolution of AI infrastructure, power, and agent security
- 2026-09-30 18:32 UTC Evolution of AI infrastructure, power, and agent security
- 2026-09-30 08:23 UTC Evolution of AI infrastructure and data center power
- 2026-09-30 05:53 UTC Evolution of AI infrastructure and data center power
- 2026-09-25 00:58 UTC Evolution of AI infrastructure and data center power
- 2026-09-17 05:27 UTC AI infrastructure and data center power evolution
Select a version to see what changed compared to the one before it.