Nvidia has unveiled new details about its latest AI chip platform, Blackwell Ultra, which aims to improve AI’s ability to reason and act on a user’s behalf. The upgraded platform, revealed at Nvidia’s annual GTC conference, builds upon the existing Blackwell chip, offering greater computing power to help AI models break down complex queries and analyze multiple possibilities more efficiently.

Advancing AI Beyond Chatbots

AI has seen rapid advancements since the release of OpenAI’s ChatGPT in 2022, leading to a surge in demand for Nvidia’s chips. These chips power AI-driven data centers used by major tech firms like Microsoft, Amazon, and Google. However, competition is increasing. The emergence of DeepSeek, a Chinese tech startup, has sparked industry discussions about whether high-performing AI models can operate without expensive hardware. DeepSeek’s R1 model, recognized for its strong reasoning capabilities, has raised questions about the necessity of costly chips.

Despite these concerns, Nvidia remains dominant in the AI chip market. The company’s recent earnings exceeded Wall Street’s expectations, reflecting sustained demand for its hardware. Nvidia claims that a query requiring 90 seconds on its previous-generation Hopper chip would take only 10 seconds with Blackwell Ultra, significantly enhancing AI processing speeds.

Industry Adoption and Future Outlook

Several major companies, including Cisco, Dell, HP, Lenovo, and Supermicro, are developing new servers based on Blackwell Ultra. The first products featuring the upgraded platform are expected to launch in the second half of 2025.

The Importance of AI Reasoning

Experts believe that improving AI’s ability to reason will enable applications to handle more complex queries. Unlike conventional chatbots that provide single responses, reasoning-based AI can analyze questions from multiple angles and generate diverse solutions. For instance, Nvidia demonstrated how a reasoning model could create a wedding seating plan that considers various guest preferences, such as seating parents and in-laws appropriately while ensuring the bride is positioned correctly.

“The models are now starting to mimic a little bit of human-like behavior,” said Arun Chandrasekaran, an AI analyst at market research firm Gartner.

The Rise of AI Agents

Beyond Nvidia, other companies are also advancing AI reasoning. Google recently enhanced its Gemini models, while Anthropic introduced the Claude 3.7 Sonnet hybrid reasoning model in early 2025. Experts suggest that these developments could lead to AI “agents” capable of performing tasks rather than just answering questions. Companies like Google, Amazon, and Qualcomm have shared visions of AI-powered assistants that could book vacations based on user preferences instead of merely providing travel suggestions.

“Agentic AI excels at multitasking,” said Gene Munster, managing partner at Deepwater Asset Management. “Enhancing reasoning capabilities will make these AI agents even more powerful.”

With Blackwell Ultra, Nvidia is positioning itself at the forefront of AI reasoning technology, ensuring its chips remain integral to the next generation of AI applications.

Source