Nvidia Q1 2027 Earnings Beat but Stock Falls

CNBC reported Wednesday that Nvidia delivered a stronger-than-expected fiscal first quarter, with revenue hitting $81.62 billion against analyst forecasts of $78.86 billion. Adjusted earnings per share came in at $1.87, topping the $1.76 consensus estimate. Despite the beat, shares fell in extended trading, marking a fourth consecutive post-earnings decline.

Data Center Drives Record Quarter

Nvidia data center revenue nearly doubled year over year, cementing it as the engine powering the company’s overall growth. CEO Jensen Huang told analysts that agentic AI has fully arrived and described the broader AI infrastructure buildout as accelerating at an extraordinary pace. Nvidia also announced an $80 billion share buyback program and raised its dividend, while guidance for the coming quarter also surpassed Wall Street expectations.

Also Read: What Is Agentic AI and Why Does It Matter for Enterprise Tech?

A New Way to Report the Business

Huang used the call to explain a significant overhaul of how Nvidia will break down its quarterly results. Going forward, the company will report two primary segments: data center, split into two subgroups, and edge computing. Huang said the change reflects how differently Nvidia’s customers operate. Hyperscalers like Meta and Alphabet have massive GPU-driven buildout programmes. Governments, robotics researchers, and smaller enterprises each have distinct computing needs. Huang said the restructured reporting was simply meant to give investors a clearer picture of each business unit.

Also Read: Nvidia GTC 2026 Highlights: Vera Rubin and the Next Wave of AI Infrastructure

Vera Rubin and the Next Architecture Cycle

Huang expressed confidence in Nvidia’s next rack-scale AI system, Vera Rubin. He said it is off to a strong start and expects it to outperform the Grace Blackwell system it succeeds. Nvidia is growing inference market share rapidly as the number of frontier model companies expands. Anthropic was cited as a notable new customer this year, accessing Nvidia compute through Microsoft Azure, Amazon Web Services, and CoreWeave.

Groq Chips Remain a Narrow Play

Not every product line drew the same optimism. Huang described Nvidia’s Groq language processing unit, acquired through a roughly $20 billion deal, as likely to remain a niche offering for the foreseeable future. The LPX rack system built around Groq chips is designed for low latency and fast token generation but carries limited throughput. Huang acknowledged the use case is not broad. The comments come as the custom ASIC market heats up, with competitors like Cerebras, SambaNova, and in-house chip efforts at Google, Amazon, Meta, and Microsoft all vying for GPU alternatives.

Read Next: Fed Holds Rates Steady as Inflation Data Complicates Cut Timeline