Investing in Astrus

AI-powered microchip design

Semiconductors and microchips are the indispensable backbone of modern electronics. They have fueled the digital revolution that transformed our world,  powering computers, smartphones, cars, televisions, and of course, AI.

These tiny marvels of engineering have evolved at an astonishing pace, continually pushing the boundaries of computing power and functionality. While digital chip design has advanced in leaps and bounds, analog design — a critical component of electronics — has remained in the ‘stone age’ of technology.

The creation of an integrated circuit (IC) involves two stages:circuit design and layout. In the circuit design stage, the logical schematics of components such as transistors and their connections are designed for the desired functionality. Layout involves the physical placement of these components and routing of nanoscale wires on the chip to map to the circuit schematics. As chip components shrink in size, layout becomes exponentially complex, especially for analog circuits. In the two key categories of design work — digital and analog — analog layout now represents the most time-intensive and resource-hungry portion of the IC design cycle. While digital layout has benefitted from electronic design automation tools, analog layout remains a highly manual process. 

Teams rely on skilled layout engineers to meticulously place analog components and route connections by hand using complex CAD software. The layout engineer must solve visual puzzles, trying different arrangements to meet signal integrity, power, area, and manufacturability requirements. This becomes an extremely iterative process, with each layout revision simulated by the circuit designer for performance verification and tweaks to the requirements. The back-and-forth cycle continues until specifications are met. A single layout can take between 6-20 hours; with multiple iterations required, development for some ICs can stretch over months. 

In this new era of AI and automation, Astrus unlocks a compelling opportunity to fully automate analog layout.

Astrus

Astrus is building an AI engine to fully automate analog layout and speed up the microchip design process. They are building the underlying infrastructure to support the most advanced AI critical workloads. A circuit designer will first build their schematic in an EDA software like Cadence. Then instead of handing that work off to a layout engineer, Astrus’s AI engine will do the automatic placement and routing.

Alex, the MVP Ventures’ team, and I have spent the last 6 months working closely with Brad Moon and getting to know him. He’s no stranger to the world of semiconductors or startups. He’s founded 2 hardware companies in the past and led product discovery at ApplyBoard. His co-founder Zeyi Wang developed a MuZero variant, the most recent version of AlphaGo, and his work is regarded by experts as the best publicly available version of MuZero implementation. As you can see, they have built a strong team ready to radically improve global computation!

Today, we at MVP Ventures are excited to announce our pre-seed investment in Astrus alongside Khosla Ventures, 1517 Fund, HOF Capital, and Alumni Ventures.