Nvidia CEO Jensen Huang recently introduced the Blackwell chip, the latest advancement in AI technology succeeding the highly sought-after H100. Named after renowned statistician and mathematician David Blackwell, the Blackwell chip is reported to be at least twice as fast as its predecessor, the H100. Huang made this announcement during his keynote address at the GPU Technology Conference (GTC) in San Jose, California.
One of the key features of the Blackwell chip is its enhanced speed, which is further amplified by its network computation capabilities. Huang highlighted that the chip can perform tasks such as converting speech into 3D video, showcasing its versatility and power.
The Blackwell chip is equipped with an impressive 208 billion transistors, significantly more than the Hopper chip. Nvidia claims that the Blackwell chip offers five times the AI performance of the Hopper while also reducing costs and energy consumption by up to 25 times.
Huang emphasized the seamless design of the Blackwell chip, stating that it eliminates memory locality and cache issues by functioning as one cohesive unit. He expressed confidence in the chip's capabilities, noting that the engineers behind it pushed the boundaries of physics to achieve remarkable results.
Although the price of the Blackwell chip was not disclosed during the announcement, Huang hinted at its premium nature, joking about the expense of its functioning board. He positioned the Blackwell chip as a catalyst for a new era of computing, heralding it as the engine to power the next industrial revolution.
The demand for Nvidia's H100 chips has been substantial, with prices reaching upwards of $40,000 per unit. Tech giants like Mark Zuckerberg have been stockpiling these chips, contributing to Nvidia's soaring sales figures. In its recent fourth-quarter earnings call, Nvidia reported a record quarterly revenue of $22.1 billion.
Despite its success, Nvidia has faced challenges in the Chinese market due to stringent export controls. Additionally, competitors such as Meta, Microsoft, Google, and AMD are actively developing rival AI chips to compete in the rapidly evolving landscape of AI technology.
