Nvidia is really holding its ground in the AI datacentre space. In their latest quarterly report, they revealed a 16% growth in revenue, which is a staggering 93% more than the same time last year. Their datacentre segment alone brought in $35.6 billion this quarter, pushing their annual revenue to $115 billion—up by 142% compared to last year.
CEO Jensen Huang shared some strong insights during the earnings call. He said that the demand for their Blackwell AI supercomputers is off the charts. He explained, “Increasing compute for training makes models smarter, and more compute for deep thinking leads to smarter answers.” It’s clear they’re pumping out these Blackwell supercomputers in huge volumes, raking in billions in just the first quarter. Huang is optimistic about the future, highlighting how agentic AI and physical AI are ready to transform big industries.
Analysts pushed back with questions about DeepSeek, which doesn’t need super powerful GPUs, and the trend of cloud service providers like Microsoft creating their own custom chips for AI. Nvidia’s business relies heavily—about half—on these cloud providers. Still, Huang pointed out growing interest from enterprise customers, claiming that this could lead to long-term GPU sales opportunities.
Huang believes that new AI models will keep driving demand, even though they’re getting more computationally efficient. He emphasized, “The more the model thinks, the smarter the answer.” He talked about reasoning models like OpenAI and DeepSeek-R1, stating they can use up to 100 times more compute. He also mentioned that future models might demand even more.
When it came to concerns about cloud providers developing their own application-specific integrated circuits (ASICs) instead of relying on Nvidia’s GPUs, Huang noted the complexity of the technology involved. He hinted that manufacturing custom chips would be a tough challenge. “The software stack is incredibly hard,” he stated, making it clear that they face similar hurdles in their operations.
According to Huang, the technology environment built around Nvidia’s architecture is ten times more complex than it was just two years ago. His point is that as software development accelerates, especially in AI, managing multiple chips together becomes even more complicated.
Forrester analyst Alvin Nguyen commented on Nvidia’s results, saying it’s impressive to see yet another record performance. He attributed this to the ongoing demand for Nvidia’s AI products and noted that stressing the need for reasoning models that require more computation counters worries about alternatives like DeepSeek.
However, Nguyen felt Huang’s take on the competition with custom chips was somewhat dismissive. He argued that ignoring the potential for companies like Amazon and Google to seek options beyond Nvidia for their specific AI needs could backfire.