OpenAI’s hunger for computing power has reached staggering new levels. The company has reportedly struck a massive deal with Broadcom to co-develop custom AI chips capable of handling the enormous energy demands of its products, including Sora 2 and ChatGPT. According to CNN, the agreement represents an additional 10 gigawatts of power capacity – roughly enough electricity to power eight million U.S. homes.
The partnership is part of OpenAI’s broader strategy to secure its own supply chain for both chips and power. With models like Sora 2, which generates high-definition video from text, and ChatGPT, which continues to grow in scale and complexity, the company’s need for computing infrastructure has exploded. Each new generation of these models requires exponentially more processing power and energy to train and run.
Until now, OpenAI has relied heavily on Nvidia’s GPUs for its data centers. But with demand outstripping supply and costs skyrocketing, the company is moving toward designing its own chips. Broadcom, one of the world’s leading semiconductor firms, will provide the design expertise and fabrication capabilities. The goal is to reduce dependence on Nvidia while improving power efficiency and performance for future AI systems.
The deal comes as concerns mount over the environmental and economic impact of AI energy consumption. Training and operating large models consumes immense amounts of electricity, both for computation and cooling. Analysts estimate that global data center energy use could double by 2030, driven largely by AI workloads. OpenAI’s new deal suggests it expects to be one of the primary drivers of that demand.
Energy analysts have noted that 10 gigawatts is an extraordinary figure. It’s not just a measure of hardware investment but of raw electricity consumption that rivals entire national grids. It underscores how the race to dominate artificial intelligence has become inseparable from access to energy and infrastructure.
OpenAI’s expansion also highlights the growing tension between innovation and sustainability. As governments push for greener energy policies, companies like OpenAI must find ways to meet their power needs without overwhelming grids or increasing emissions. Broadcom’s focus on chip efficiency could be key to balancing those priorities.
This deal marks a turning point for the AI industry. The competition is no longer just about smarter algorithms – it’s about who can power them. And with 10 gigawatts now in play, OpenAI is betting big that control over energy and silicon will define the next era of artificial intelligence.