Raspberry Pi Goes All In on AI With $70 Hailo Kit (2024)

Raspberry Pi Ltd is the latest company to jump on the artificial intelligence trend with a new kit built with AI chipmaker Hailo. The $70 bundle brings AI to the Raspberry Pi 5 and provides a surprising amount of performance in a compact and low-power package.

The core elements of the Raspberry Pi AI Kit are the Raspberry Pi M.2 HAT+ and a Hailo-8L AI acceleration module that is set on an M.2 board. The kit also includes a pre-fitted thermal pad to aid in cooling the Hailo device, a 16mm stacking general-purpose input/output (GPIO) header, and the necessary mounting hardware to connect the Raspberry Pi M.2 HAT+ to a Raspberry Pi 5 (which must be purchased separately).

Hailo-8L PCIe device (Credit: Raspberry Pi Ltd)

“We are thrilled to support Raspberry Pi and empower its vibrant community of professional engineers and creative makers with cutting-edge AI capabilities,” says Hailo CEO and co-founder Orr Danon. “Our partnership with the world’s leading single-board computer provider will inspire a new era of computing, enhanced by our high-performance AI processing capacity.”

Raspberry Pi's HAT (which stands for hardware attached on top) module utilizes the Pi 5’s lone PCIe 2.0 connection to connect the two devices, but the GPIO pins remain open for use with other devices. Since a PCIe connection is required to support the device, only the Raspberry Pi 5 will be compatible with the AI kit at launch.

The Hailo-8L can process 13 trillion operations per second (TOPS), placing it ahead of the AI hardware in AMD’s first-generation XDNA Ryzen 7040-series devices and Intel’s Meteor Lake processors. AMD’s 7040-series chips top out at 10 TOPS, and Intel’s Meteor Lake processors hit their max at 11 TOPs, making the Hailo-8L measurably faster than both.

Though this level of AI performance is considerably slower than AMD’s upcoming AI 300 chips, which hit 50 TOPS, and Intel’s Lunar Lake chips, which can achieve 120 TOPS, it’s plenty for DIY computing. These are entry-level devices, and the cost of ownership is only a fraction of what these newer AI chips will cost, especially as they will only be available in full-fledged laptops.

“Our collaboration with Hailo enables Raspberry Pi's industrial customers to integrate AI into high-performance solutions that are extremely cost-effective and power-efficient. For enthusiasts, the AI Kit provides an accessible way to enhance their creative projects with AI,” says Eben Upton, Raspberry Pi CEO. “Hailo’s combination of high compute power and low power consumption make it an incredibly attractive AI solution for professionals and enthusiasts alike.”

Potential Use Cases for Pi With AI

Both Raspberry Pi and Hailo see this level of performance as more than sufficient for the device's targeted use cases. In particular, both companies called out using AI hardware with a camera to perform various tasks, such as object identification and image recognition.

However, the AI hardware can flex to accommodate numerous other purposes. In a briefing call, Danon said the company had many AI models that could be used for various applications and suggested the AI hardware could power AI features in games, too.

We asked if Microsoft's Copilot software would be supported by the Hailo-8L running on a Raspberry Pi 5 system and were told that some of those functions would probably work. Given how new Copilot is and that the Raspberry Pi AI Kit isn't even out yet, it's unlikely that either has been optimized to work together—but that doesn't mean they won't. We were also left with the impression that they likely could be optimized to work better together if someone was so inclined to do so, but as Windows 11 isn't a key target for the Raspberry Pi 5 or its AI hardware, this has not been done as of yet.

Raspberry Pi AI Kit assembled (Credit: Raspberry Pi Foundation)

The full extent of what can be done with these AI accelerators will likely only be known in time. Raspberry Pi devices have always focused on software and hardware development, which doesn’t change with the new AI hardware. This Hailo accelerator simply adds to it, allowing coding for AI using the Raspberry Pi and developing your own uses for the hardware.

For those of us who aren’t programmers at heart or trying to learn, a Raspberry Pi representative said the company will provide many demos and software utilities to support the new AI functionality. Likely, some of this will focus on writing AI code, but other utilities are meant to be easy to set up and run quickly. Setting this device up and running AI applications in a matter of minutes was apparently a key goal of its development.

We don’t have an exact release date for the Raspberry Pi AI Kit, but it will likely land soon. All of the parts are already sold separately, which should reduce the time required to bundle them into a kit, which is expected to cost $70.