Nvidia AI Chips: Jensen Huang Responds to Skeptics
Nvidia CEO Jensen Huang has responded to critics who doubted the need for Nvidia’s costly AI chips after the impressive performance of DeepSeek.
“Almost the entire world got it wrong,” Huang remarked during the conference, which he described as “the Super Bowl of AI,” as reported by Reuters.
“The computation required due to agentic AI and reasoning is now estimated to be 100 times greater than previous projections from last year,” he explained, highlighting the functionality of autonomous AI agents that operate with minimal human input.
Only a few months ago, the necessity for Nvidia’s high-priced hardware was questioned when DeepSeek, a Chinese AI firm, successfully used older, more affordable AI chips to train its models. This shift has prompted AI companies to move from extensive data training to a focus on “inference,” where AI models leverage their intelligence to generate responses.
Nvidia’s substantial revenue of $130.5 billion last year was largely driven by its AI chip sales, with some units costing tens of thousands of pounds each. Despite Huang’s optimistic presentation, investor confidence appeared to falter, as Nvidia’s shares declined by 3.4% on Tuesday.
Huang also addressed criticisms regarding DeepSeek, stating that Nvidia’s Dynamo open-source software could significantly enhance the speed of DeepSeek’s R1 model.
“Dynamo can harness this advantage and provide 30 times higher performance with the same number of GPUs and architecture for reasoning models like DeepSeek,” explained Ian Buck, NVIDIA’s vice president and general manager of hyperscale and HPC computing.
