코그니툼(Cognitum) by ChipStart, TekStart 사업부

Background Problem: Sending vast amounts of data generated from edge devices such as cameras and IoT sensors to the cloud for processing causes bandwidth bottlenecks, latency, and the risk of personal information exposure. For example, in the case of security cameras that require real-time video analysis, immediate response may be difficult due to network delays.

Limitations of Existing Methods: Previously, edge AI processing relied on high-performance cloud servers or required the installation of power-consuming processors in devices. This resulted in limitations such as lack of real-time capability, network dependence, and increased battery consumption. Due to the lack of dedicated AI chips for edge devices, manufacturers had to compromise on performance or power efficiency by using general-purpose chips.

Cognitum adopts a dedicated semiconductor architecture that simultaneously achieves ultra-low power consumption and high AI computing performance, enabling inference within the device without cloud dependence compared to competing solutions. This minimizes latency and enables real-time decision-making without external RF modules or expensive equipment, and enhances privacy protection.

For example, it consumes significantly less power than other companies' chips when performing the same tasks, making it ideal for battery-powered edge devices. The innovative design quality and impact have been proven by winning the CES Innovation Award.

Entities that will actually pay: Device manufacturers and solution providers that require edge AI functionality are the main customers. It is expected that CCTV/smart cameras, AR/VR wearables, and industrial automation equipment will adopt this chip in their designs. Ultimately, it is a B2B structure that benefits companies or public institutions that operate these devices.

B2C/B2B/B2G: B2B (targeting manufacturers who want to embed the chip in finished products). In some cases, it may be indirectly adopted in public sector (B2G) projects such as smart cities and security.

Environmental/Regulatory Constraints: It can be applied to the global edge device market without being tied to specific countries or regulations. Offline AI processing is possible even in environments where cloud connection is difficult (e.g., agriculture, industrial sites), reducing geographical constraints.

Industry/Market Expansion: It can be expanded to various fields such as smart cameras, IoT, AR/VR, industrial equipment, and agricultural technology, and TekStart's ChipStart business unit can provide this chip design customized to various companies, enabling entry into vertical markets. As the demand for edge computing increases, the scope of application is expected to expand to automobiles and home appliances in the future.

CES Award Context: Recognized for design quality and technical impact with the Innovation Award (Honoree) in the Embedded Technologies category. The judges highly praised the innovation and design completeness specialized for edge AI and noted that this chip is a solution to resolve the bottleneck of edge computing.

Technical Completeness: It is evaluated as a highly feasible result of combining ChipStart's semiconductor design expertise and TekStart's market-oriented approach. It is expected that the balance between ultra-low power design and high computing performance could partially change the landscape of the industry if verified in actual application cases.

Market Expectations: As the importance of edge AI grows, demand for real-time processing of on-site data is increasing, and it is expected that there will be many applications for this chip. In particular, there are positive prospects for performance improvements in security surveillance cameras, autonomous drones, etc.

Overestimation/Underestimation Factors: However, market entry strategy compared to competitors and whether actual commercial chips will be produced remain uncertain. As large companies such as NVIDIA and Google are also entering the edge AI chip market, how quickly TekStart will build an ecosystem with its scale can be a variable. The innovation of the technology itself is highly evaluated, but there are also cautious opinions about the actual commercialization speed.