During the annual GTC event in San Jose, California, on Tuesday, Nvidia (NVDA) CEO Jensen Huang unveiled the company’s next-generation Blackwell Ultra AI chip. Alongside the Blackwell Ultra, Nvidia introduced the GB300 superchip, which integrates two Blackwell Ultra chips with the company’s Grace central processing unit (CPU). These chips are engineered to enhance AI systems for a diverse clientele, including tech giants like Amazon (AMZN), Google (GOOG, GOOGL), Microsoft (MSFT), and Meta (META), as well as research laboratories globally. As of the close on March 18, Nvidia’s stock stood at 115. 43, reflecting a decline of 3. 43%. According to Nvidia, the Blackwell Ultra provides 1. 5 times the performance of its predecessor, Blackwell, and offers a staggering 50-fold increase in revenue potential for data centers compared to its Hopper chip, aided by its advanced AI features. Nvidia notes that the Blackwell Ultra is tailored for the "age of AI reasoning, " a processing style that emulates human thought processes. This approach gained traction when DeepSeek launched its R1 AI model, with OpenAI's o1 and Google's Gemini 2. 0 Flash Thinking also being examples of reasoning models. DeepSeek initially caused a stir on Wall Street by claiming it developed its AI models far more economically than major Silicon Valley firms, utilizing less powerful chips. However, Nvidia has countered this claim, asserting that reasoning models gain significantly from leveraging robust GPUs, which result in quicker, more accurate responses to user inquiries. Like its predecessor, the Blackwell Ultra will fit into Nvidia’s expansive NVL72 rack server, which accommodates 72 GB300 superchips, promising enhanced efficiency and serviceability. The GB300 NVL72 is reported to handle 1, 000 tokens per second using DeepSeek’s R1 AI model, a major improvement from the 100 tokens per second achievable with the Hopper chip.

This advancement means the GB300 NVL72 can respond to user questions in approximately 10 seconds, compared to the 1. 5 minutes needed by Hopper, marking a significant upgrade from the older systems. Additionally, Nvidia plans to incorporate the GB300 into its DGX SuperPod, a powerful AI supercomputer that amalgamates multiple NVL72 servers into one AI powerhouse. Each SuperPod will feature 288 Grace CPUs, 576 Blackwell Ultra GPUs, and a remarkable 300TB of memory. Nvidia's Blackwell chip has entered full production, marking the fastest ramp-up in the company’s history. In its latest quarter, Nvidia reported that Blackwell contributed $11 billion to its total revenue of $39. 3 billion. Despite this strong quarterly performance, Nvidia's stock has faced challenges due to concerns that hyperscalers may be overspending on AI without adequate returns. Furthermore, President Trump’s potential 25% tariff on overseas-produced semiconductors and the possibility of additional export controls have further complicated matters. Year-to-date, Nvidia shares have fallen by 11%, albeit with an increase of 36% over the past year. For further inquiries, you can email Daniel Howley at dhowley@yahoofinance. com and follow him on Twitter @DanielHowley. Click here for the latest technology news that could influence the stock market. Stay updated with the latest financial and business news from Yahoo Finance.