Access Intel Loihi Hardware

Intel Loihi and Loihi 2 chips are not currently available as Intel products. They can only be obtained for your research or evaluation programs once you .

Intel Labs supports members of the INRC by providing access to pre-production Loihi and Loihi 2 hardware.

Kapoho Point includes up to 8 Loihi 2 chips. Pictured here is a Kapoho Point (bottom board) attached to a host and FPGA (top two boards).
Kapoho Bay includes two Loihi chips in a self-contained USB form factor with integrated heatsink.

 

Neuromorphic Research Cloud

Most members will have access to Loihi hardware via the Neuromorphic Research Cloud, sometimes referred to as vLab, which is a shared pool of virtual machines (VMs) and Loihi systems that can be accessed from an SSH terminal anywhere in the world. These VMs and attached Loihi systems can be used to build, test, and evaluate (benchmark) a wide range of neuromorphic algorithms and applications.

The majority of Intel Labs' own research is conducted on the Neuromorphic Research Cloud, and once you properly configure the VM, it offers an easy and productive development experience.

For INRC members, more information on cloud setup and usage is available in

Request Loihi 2 Hardware

Projects that need or clearly benefit from on-site hardware can request a Loihi 2 system. These systems can be loaned for free for up to 1 year to academic community members or purchased by commercial or government organizations with longer term needs.

As noted above, Loihi hardware is currently not available to any organizations outside the INRC. Intel Labs uses the following checklist to evaluate hardware requests.

Has the research group…

Submitted a project plan or description of the intended evaluation clearly describing the need or benefit of the requested system? See the RFP on for more details. If you are an INRC member already but did not include your hardware plans in the project proposal, contact Intel Labs to update your plan.

Built and tested a neuromorphic model (spiking neural network)? For neuromorphic projects, most of the algorithm design and implementation should be done on your local machine using standard hardware. It is faster and easier to train and test locally, before moving your model to a neuromorphic device. Note that building related models, such as deep neural networks or simple recurrent networks, will not directly prepare your team to use Loihi hardware.

Implemented the model in Lava? As of 2022, INRC members should use the open-source Lava framework to develop and evaluate spiking neural networks for Loihi 2. Older NxSDK models should be easily re-implemented in Lava, as the core features are similar, but Lava offers a vastly improved ability to compose and re-use model code for neuromorphic applications.

Evaluated the model or algorithm on the Neuromorphic Research Cloud or in simulation? Most performance testing and benchmarking can be conducted under the controlled conditions of the research cloud, where Intel Labs handles the management of lots of Loihi 2 systems so you can run experiments quickly.

Groups who complete the above list clearly demonstrate the ability to immediately make use of Loihi 2 hardware, and thus will receive priority access. Once you’re ready, check out how to

Questions or feedback? Please comment below or email the INRC team.