- The Nvidia CEO confirms the names of Blackwell Ultra and Vera Rubin
- Both have revealed in GTC 2025 in March 2025
- Nvidia considers that general income grow 114% year after year as the rings of the demand for AI
The CEO of Nvidia, Jensen Huang, has made fun that the company’s next flagship chip will be shown in the coming weeks, and apparently officially confirms its name.
In statements to the analysts of the fiscal earnings call of the fourth quarter of the fourth quarter of the company, Huang revealed that the new hardware will lead to Blackwell Ultra nickname, and will be announced in its Nvidia GTC 2025 event in March 2025.
“Come to GTC and I will talk about Blackwell Ultra, Vera Rubin, and then I will show you the only click after that,” Huang said.
Nvidia Blackwell Ultra … and more
Huang then added that Blackwell Ultra will officially launch in the second half of 2025, and offers updates in processors, networks and memory, but will be built on the same architecture of the system as Blackwell.
Both Blackwell Ultra and Vera Rubin were bothered in a company’s roadmap document in Computex 2024, but Huang has now apparently confirmed the name of both.
Nvidia presented Blackwell in GTC 2024, promising a great step forward in terms of power and efficiency of AI.
Since then, he has seen a series of launches, including the first Blackwell “Superchip”, the GB200, which has the ability to climb from a single shelf to a complete data center, since Nvidia seeks to move forward with her leadership in AI’s career.
Blackwell contains 208 billion transistors (compared to 80 billion in Hopper) in its two GPU trochers, which are connected by 10 TB/Second Link from chip to chip in a single unified GPU, which achieves up to 30 times faster than Hopper when it comes to inference tasks AI, which offers up to 20 Petaflops of Power Power FP4, far ahead of anything else in the market today.
Despite this, Nvidia says that Blackwell can reduce the consumption of costs and energy up to 25 times, giving the example of training a parameter model of 1.8 billion billion, which would have previously taken 8,000 GPU of hopper and 15 megawatts of power, but now you can do only 2,000 Blackwell gpu that consumed only four megawatts.
Vera Rubin is the * next * step forward for Nvidia after Blackwell Ultra, with an expected release in 2026, with CPU and fastest GPU products, including a vera Rubin plate that combines the GPU and the CPU in a “superchip”.
Huang’s provocation occurred when Nvidia revealed its most recent set of financial results, and the company once again saw record yields, since it benefits from the large increase in the demand for AI.
The general income of NVIDIA doubled more year after year, reaching $ 130.5 billion for the financial year, an increase of 114%, largely helped by the income of the record data centers of $ 35.6 billion, 16% more than the previous quarter and 93% more than a year ago.
“Blackwell’s demand is surprising since AI reasoning adds another scale law: increasing the calculation for training makes the smartest models and the increase in computation for prolonging thinking that the answer is smarter,” Huang said about the results.
“We have successfully increased the large -scale production of the Blackwell supercomputers, reaching billions of dollars in sales in its first quarter. The AI advances at the speed of light, since the agent and physical the AI prepare the scenario for the next wave of Ia to revolutionize the largest industries. “
Through The edge
