Nvidia has introduced Blackwell Ultra, a new AI chip designed to push training performance beyond its B200 predecessor. The company says the updated processor delivers about 30% better performance, a move aimed at meeting surging demand for advanced AI hardware.

The launch comes as cloud providers, chip buyers and enterprise developers continue to pour money into AI systems that require more powerful and efficient training infrastructure. Nvidia has remained the dominant supplier in that market, and the Blackwell Ultra release is meant to reinforce that position.

According to the company, the new chip builds on the Blackwell family with improved speed for large-scale model training. Nvidia did not present the product as a radical redesign, but rather as a targeted upgrade for customers seeking incremental gains in performance and efficiency.

The announcement adds to the competitive pressure in the AI chip race, where rivals are trying to narrow Nvidia’s lead. For now, the company appears focused on extending its advantage by offering faster hardware to data centers and AI developers racing to expand their systems.