Advanced Computing in the Age of AI | Thursday, June 8, 2023

Sony Invests in Raspberry Pi for Enterprise AI Push 

One of the world's most popular computer makers, Raspberry Pi, is receiving a cash infusion from Sony in a move to bring AI computing closer to the data than vice versa.

Raspberry Pi's namesake low-cost computers, which are circuit boards without a chassis, have achieved cult status among makers and DIYers that are building electronics at home. The non-profit has sold 50 million Raspberry Pi computers to date.

Sony's investment will serve multiple purposes, which include providing a boost to Raspberry Pi's ability to scale the design, development, and production of its boards. It also tightens a partnership where Raspberry Pi can bring Sony's sensors and imaging assets to its hardware.

The companies did not disclose the amount Sony was investing in Raspberry Pi. The pair were also tight-lipped if the investment would lead to new hardware. But Raspberry Pi CEO Eben Upton gave some hints.

"I suspect what you will see from us – you're going to see image sensor products with integrated machine learning acceleration," Upton told EnterpriseAI.

Raspberry Pi is expecting shipments during the current quarter to be its largest ever, with Upton pushing his team to ship around 2.3 million units. Upton hopes to ship 10 million units this year after limited chip supplies disrupted shipments over the last two years.

Sony’s cash infusion will help meet that sudden surge in demand, Upton said. In 2021, Raspberry Pi shipped seven million units.

"We went into 2021 with a half million backlog and we left the year with a four and a half million backlog," Upton said, adding that chip constraints limited Raspberry Pi board shipments to only able to ship 5 million units in 2022.

The partnership with Sony will also help Raspberry Pi reach a new class of enterprise customers, Upton said.

"It's nice to actually do something new. Although it's another image sensor, because it has this embedded acceleration it actually feels new, not just another camera," Upton said.

Sony has been a Raspberry Pi partner for a long time. Last month, Raspberry Pi announced a Global Shutter Camera based on Sony's 1.6-megapixel IMX296 sensor. Raspberry Pi has been exclusively partnering on its camera attachments with Sony over the last three-to-four years, Upton said. Sony has played a role in producing Raspberry Pi products since August 2012.

But a deeper integration of Sony's imaging products on Raspberry Pi's boards makes sense as applications related to IoT and AI come closer.

Raspberry Pi’s board computers already gather and process data on the edge. Some of Sony's cameras like the IMX500 or IMX501 are able to run AI through on-board pixel analysis, which helps send only relevant image data back to larger AI models.

A Raspberry Pi and Sony product could be something "that looks a lot like a camera board. We know how to do those, we know how to design them," Upton said, adding that Sony knew how to manufacture those, which makes it an obvious match.

Sony has imaging products with its AI cameras, but needs a supporting computing platform, which is where Raspberry Pi fits in. The partnership provides a full hardware stack for its Aitrios platform, which includes the AI camera, a machine-learning model, and development tools.

Sony's AI cameras can be tacked on Raspberry Pi boards as a form of image accelerator, which sends only information upstream that is relevant to AI models. That lightens the overall computing load on the Raspberry Pi board, which can dedicate more computing resources to other tasks.

"The investment will provide the Aitrios platform and Aitrios compatible hardware integration that allows the large community of Raspberry Pi to have a cost-effective, significantly higher performance visual solution paired to the Raspberry Pi," Sony representatives said in an email.

Raspberry Pi’s user base now has expanded to enterprises doing AI on the edge, and needing intelligence in places like factory floors, retail stores, or any point where data needs to be analyzed.