Nvidia has unveiled a âsuperchipâ for training artificial intelligence models, the most powerful it has ever produced. The US computing firm, which has recently rocketed in value to become the worldâs third-largest company, has yet to reveal the cost of its new chips, but observers expect a high price tag that will make them accessible to only a few organisations.
The chips were announced by Nvidia CEO Jensen Huang at a press conference in San Jose, California, on 18 March. He showed off the companyâs new Blackwell B200 graphics processing units (GPUs), each of which has 208 billion transistors â the tiny switches at the heart of modern computing devices â compared with the 80 billion transistors of Nvidiaâs current-generation Hopper chips. He also revealed the GB200 Grace Blackwell Superchip, which combines two of the B200 chips.
âBlackwell is just going to be an amazing system for generative AI,â said Huang. âAnd in the future, data centres are going to be thought of as AI factories.â
Advertisement
GPUs have become coveted hardware for any organisation seeking to train large AI models. During AI chip shortages in 2023, Elon Musk spoke of GPUs being âconsiderably harder to get than drugsâ and some academic researchers without access bemoaned being âGPU poorâ.
Nvidia claims its Blackwell chips can deliver 30 times performance improvement when running generative AI services based on large language models such as OpenAIâs GPT-4 compared with Hopper GPUs, all while using 25 times less energy.
It says that whereas GPT-4 required approximately 8000 Hopper GPUs and 15 megawatts of power to perform 90 days of training, the same AI training could be done using just 2000 Blackwell GPUs consuming 4 megawatts of power.
The company hasnât yet revealed the cost of the Blackwell GPUs, but the price tag is likely to reach eye-watering levels, given that the Hopper GPUs already cost between $20,000 and $40,000 each. This focus on developing more powerful and expensive chips means they âwill only be accessible to a select few organisations and countriesâ, says at Hugging Face, a company that develops tools for sharing AI code and datasets. âApart from the environmental impacts of this already very energy-intensive tech, this is truly a Marie Antoinette, âlet them eat cakeâ moment for the AI community,â she says.
The electricity demand from data centre expansions â largely driven by the generative AI boom â is expected to double by 2026, matching the energy consumption of Japan today. That can also come with steep rises in carbon emissions if the data centres supporting AI training continue to rely on fossil fuel power plants.
Global demand for GPUs has also meant geopolitical complications for Nvidia amid growing tensions and strategic competition between the US and China. The US government has implemented export controls on advanced chip technologies to delay Chinaâs AI development efforts in a move that it describes as vital to US national security â and that has forced Nvidia to create less powerful versions of its chips for Chinese customers.
