NVIDIA releases TensorRT 7 inference optimization software while Intel acquires Habana Labs for $2 billion, signaling competing software-versus-hardware strategies.

NVIDIA and Intel are making significant moves in the AI infrastructure space, though with different strategic approaches. NVIDIA released TensorRT 7, its latest inference software library, which brings significant performance improvements that the company says enable conversational AI. Meanwhile, Intel is stepping up its game by acquiring Habana Labs, an AI chip startup that promises top performance on the hardware level.

At GTC China, NVIDIA made a series of announcements highlighting both software and hardware developments. The company announced partnerships with major players including Alibaba and Baidu, demonstrating the scale of its influence in the AI market. Beyond these partnership announcements, NVIDIA unveiled Orin, a new system-on-a-chip (SoC) designed specifically for autonomous vehicles and robots. Accompanying this hardware release is Nvidia Drive AGX Orin, a new software-defined platform powered by the Orin SoC.

These moves represent NVIDIA's expanding interest and demonstrated progress in domain-specific applications, coupled with an effort to foster a broader ecosystem around its technologies. The company's dual focus on both software optimization through TensorRT 7 and specialized hardware through Orin reflects a comprehensive strategy to capture different segments of the AI infrastructure market.