And otherwise, they're going direct?

Yes, otherwise they go direct or through distributors. We also direct them to distribution channels. One thing about the design space is when someone buys a Raspberry Pi or a development vehicle, there's an opportunity to fill your pipeline. Typically, when they buy a prototype, they won't be in production for nine to 12 months. After the initial purchase, they don't buy 1,000 units the next month. They develop the prototype for two to three months, reach a point to secure funding, and then move to a pilot test, maybe buying 50 units three months later. They test for another three to six months, and production might start three months after that. So, it takes nine to 12 months before significant revenue returns, regardless of the design.

Please continue with your story about the embedded market.

While 4G LTE was suitable for IoT devices, 5G is not as appropriate unless the devices require high bandwidth for tasks like video capture. Recently, chips have become less expensive and more powerful, enabling processing at the edge of the device. AI plays a significant role in this, with object recognition and condition monitoring using machine learning and artificial intelligence to make decisions based on collected data.

Please continue with your story about the embedded market.

Instead of sending all data back to the cloud, much of it is now processed at the edge, known as edge computing. Microprocessor companies are working on enhancing edge processing at a reduced cost as chip prices decrease and performance increases. This allows more compute power at the edge, reducing the data sent to the cloud and enabling more devices per data plan to perform edge intelligence.