Nvidia is on a roll, especially in its datacentre business, thanks to the booming demand for AI workloads. In its latest quarterly report, the company announced revenue of $35.1 billion, marking a 17% increase from the previous quarter and an astonishing 94% rise from last year. The datacentre division contributed a hefty $30.8 billion to that total, highlighting its critical role in Nvidia’s success.
Jensen Huang, Nvidia’s founder and CEO, emphasized the shift towards AI, saying, “The age of AI is in full steam.” He pointed out the incredible demand for their next-gen chips, Hopper and the soon-to-be-released Blackwell, as companies ramp up their AI capabilities.
Huang remarked on the sweeping changes AI is bringing to all sectors. “Industries are embracing agentic AI to redefine workflows,” he noted. He also highlighted a surge in investments in industrial robotics and the global recognition of the need for strong national AI frameworks.
Nvidia’s Chief Financial Officer, Colette Kress, shared some impressive numbers, stating that sales of H200 GPUs have soared into the “double-digit billions.” She described this as the fastest product ramp in Nvidia’s history, with cloud service providers accounting for roughly half of the datacentre revenue.
During the earnings call, Huang discussed advancements in enhancing large language models. He explained that scaling the training process is still viable. “AI foundation model pre-training scaling is intact and continues,” he stated, emphasizing that it’s an empirical observation rather than a hard rule.
He also touched on the evolution of post-training techniques, noting a shift from human feedback to AI-generated insights. “The longer it thinks, the better the response,” he said, explaining that it’s about refining answers through careful consideration, similar to how we think before speaking.
To make AI more accessible, Huang pointed out that boosting GPU performance helps lower training and inference costs. “We’re reducing the cost of AI,” he said.
When asked about potential challenges to growth, Huang acknowledged the scale of datacentres. “Most are around a hundred megawatts or more, but we aim for gigawatt facilities,” he explained. He added that success hinges on delivering maximum performance per watt, which directly impacts revenue.