Nvidia Corp.'s
What's untold is just how much more powerful GB200 is compared to the first AI supercomputer, the DGX-1, that Nvidia donated to OpenAI back in 2016.
What Happened: Nvidia unveiled the GB200 "miracle" chip at the GTC event on Monday.
CEO Jensen Huang and OpenAI CEO Sam Altman - both at the forefront of AI - believe that "compute" is the future, and the new GB200 superchip looks to be a promising leap in this crucial aspect.
According to the details revealed by Nvidia, GB200 can deliver up to 1.44 exaflops of AI compute.
On the other hand, the DGX-1 supercomputer that Nvidia donated to OpenAI in 2016, was rated to deliver up to 170 petaflops of compute.
To put this into context, there are 1000 teraflops in one petaflop and 1000 petaflops in one exaflop.
Purely in terms of computing capabilities, GB200 delivers more than 8,470 times the computing performance of DGX-1.
Huang hand-delivered the DGX-1 supercomputer to OpenAI, in the presence of co-founders Musk, Altman, Ilya Sutskever, and Greg Brockman, among others.
"The DGX-1 is a huge advance," OpenAI's chief scientist Sutskever said at the time.
"It will allow us to explore problems that were completely unexplored before, and it will allow us to achieve levels of performance that weren't achievable."
Why It Matters: Huang thinks the world is just getting started when it comes to computing power in the AI industry.
"We're in the beginning of this AI computing ramp," Huang said. "And we're in the beginning of the accelerated computing ramp. It's going to last a few years."
"We created a brand-new way of doing computing."
OpenAI's Altman also echoed similar sentiments on the Lex Fridman Podcast, saying "compute" is the future currency, not fiat or Bitcoin.
