Nvidia has introduced its Blackwell B300 GPU architecture, the latest addition to its AI hardware lineup, as companies continue to expand their use of generative artificial intelligence. The company says the new chip is designed to improve inference performance for large language models, a key measure of how quickly AI systems can generate responses.

According to Nvidia, the Blackwell B300 can deliver up to 40% better inference performance than previous versions, positioning it as a major upgrade for enterprise customers. The announcement comes as businesses across sectors race to deploy AI tools for customer service, software development, data analysis, and other commercial uses.

The launch underscores how competition in the AI chip market remains intense, with hardware makers trying to meet rising demand for faster and more efficient processing. Nvidia has been a dominant supplier for AI training and deployment, and the new product is likely aimed at preserving that lead as adoption grows.

The company’s latest release adds to a fast-moving technology market where performance gains can influence spending decisions for large organizations. As generative AI moves deeper into mainstream business operations, chips that improve speed and efficiency are becoming increasingly important.