Inventec (2356) spent more than three years to invest in AI chip design and R&D outputs, and then officially announced the latest embedded neural network processor IP "VectorMesh" AI accelerator series in the first quarter of this year, announcing that it has entered the upstream field of IC design. Now, Inventec is further locking in the needs of in-vehicle applications and launching MCU single-chip software IP applications.

In addition to the advantages of low power consumption, high performance, and highly elastic architecture, Inventec's VectorMesh AI accelerator series can also support advanced AI inference operations, from model training, design, SoC integration to chip mass production, providing one-stop and customized integration services, which can not only greatly shorten the customer's product development time, but also enhance the market competitiveness of its products.

The VectorMesh AI accelerator series has also won the cooperation of several of Taiwan's top 10 IC design companies, and it is expected that the chips will be rolled off the production line from the fourth quarter of this year.

The VectorMesh AI accelerator series has also won the cooperation of several top 10 IC design companies in Taiwan, and it is expected that the chip production will start in the fourth quarter of this year.

Chen Weichao, senior vice president of Inventec, pointed out that the implementation of the Minima series can be used for object detection, feature detection, face detection, face recognition, posture recognition, image segmentation, image beautification, noise filtering and other applications.

Inventec also announced that it will successively launch the Parva and Magna series with higher complexity and greater terminal computing power, which also have three characteristics: low power consumption, high performance, and highly elastic architecture, and are applied to AI inference at the edge to meet the diverse needs of the market.