Astera Labs reduced AI connectivity chip simulation times by leveraging NVIDIA B200 GPU-accelerated computing on AWS, according to the company's VP of Hardware Engineering Jitendra Mohan.1 The collaboration between Astera Labs, Synopsys, NVIDIA, and AWS is "transforming our ability to design advanced blocks for AI connectivity silicon," Mohan stated.1
Arteris deployed FlexGen technology to generate optimized chip interconnects "in a fraction of the time with significantly improved power, performance, and area results."2 The tool automates network-on-chip design, addressing bottlenecks in complex AI processor development.
Supply chain constraints are driving architectural changes in chip packaging. Microchip Technology introduced the SAM9X75D5M system-in-package to "protect designers from the volatility and supply constraints that have historically affected the discrete DDR memory market," the company said.3 The automotive-qualified package integrates processor and memory in a single unit.
Tower Semiconductor expanded specialty process offerings for power management and analog chips used in AI systems, focusing on mature nodes less affected by cutting-edge manufacturing restrictions.4 These foundry capabilities support companies developing AI infrastructure outside leading-edge process nodes.
Edge AI hardware is converging with wireless connectivity. "Connectivity without intelligence is becoming commoditized and edge AI without seamless wireless connectivity is incomplete," said Mariusz Malkowski of Trident IoT, announcing a partnership with Syntiant for low-power audio AI sensors.5
The design tool improvements arrive as geopolitical tensions reshape semiconductor supply chains. Companies are investing in tools that accelerate development cycles while navigating component availability issues. GPU-accelerated simulation cuts months from verification timelines, critical as AI chip complexity increases with each generation.
Integrated packaging solutions like Microchip's address memory supply risks by eliminating dependence on discrete component markets. This approach trades some flexibility for supply predictability, a calculation more companies are making as they plan multi-year product roadmaps.