Facing concerns over AI computing power consumption, NVIDIA promotes the high energy efficiency of the Blackwell chip
NVIDIA promoted its latest Blackwell chip at the AI Summit DC conference in Washington, emphasizing its high energy efficiency to meet the power demands of AI computing. The chip only requires 3 gigawatts of power to run OpenAI's GPT-4 software, a significant decrease from 5500 gigawatts ten years ago. NVIDIA Vice President Bob Pette stated that energy efficiency was a key focus during the design of the Blackwell platform. In addition, NVIDIA showcased its progress in the software field, helping businesses quickly deploy customized AI services. NVIDIA's stock price rose by 4.1% to $132.92
Facing concerns about AI computing power consumption, NVIDIA promoted the energy efficiency of its latest chips at a conference in Washington on Tuesday. At the event called "AI Summit DC," NVIDIA stated that its Blackwell chip, which it began offering to customers this year to develop OpenAI's GPT-4 software, requires 3 gigawatts of power. The chipmaker said that a decade ago, this process required as much as 5,500 gigawatts of power. NVIDIA Vice President Bob Pette said in a briefing before the event, "Our Blackwell platform was designed with energy efficiency in mind from the beginning." The California-based company also highlighted its progress in the software field, including creating "agent blueprints" that allow enterprises to quickly deploy customized AI services. Agents are software that handle real-world tasks such as processing customer requests, assisting in device design, or accelerating drug development