Wall Street loves a good drama, and the latest NVDA earnings call live didn't disappoint. If you were watching the tickers or listening to the webcast, you know things felt a bit different this time. It wasn't just about beating estimates anymore; it was about whether the "virtuous cycle of AI" Jensen Huang keeps talking about actually has a ceiling.
Honestly, the numbers were kind of ridiculous.
Nvidia pulled in a record $57 billion in revenue for the third quarter of fiscal 2026. That is a 62% jump from a year ago. To put that in perspective, the company grew its revenue by $10 billion in just three months. But if you look past the headlines, the real story is in the data center. That segment alone brought in $51.2 billion. Basically, 90% of Nvidia is now just AI infrastructure.
The Blackwell Reality Check
Everyone was waiting for updates on Blackwell. You've probably heard the rumors about production delays or cooling issues, but CFO Colette Kress was pretty blunt on the call. She basically said that based on what they are seeing, they might actually blow past their original $500 billion revenue goal for Blackwell and the upcoming Rubin architecture through 2026.
✨ Don't miss: What Really Happened With the Google Cloud Outage September 27 2025
Demand is "off the charts."
Jensen mentioned that cloud GPUs are essentially sold out. This is a weird problem to have. Usually, when you can't make enough of something, customers get annoyed and look elsewhere. But in the world of high-end AI, there isn't really an "elsewhere" that can do what the Blackwell NVL72 does.
Why the Stock Isn't Just Going Vertical
It is strange, right? Record earnings, huge beats on EPS ($1.30 vs the $1.26 expected), and yet the stock has been acting like a bit of a "laggard" lately compared to some other AI names.
Some analysts, like Chris Caso over at Wolfe Research, think this is because of the late launch of Blackwell and general jitters about whether Big Tech is going to keep spending $600 billion a year on data centers. People are terrified of a bubble. During the NVDA earnings call live session, Jensen tried to kill that narrative. He called OpenAI "once-in-a-generation" and argued that we are at a "tipping point" where AI starts doing everything for every industry.
- Total Revenue: $57 billion (Up 62% YoY)
- Data Center Revenue: $51.2 billion (Record high)
- Gaming Revenue: $4.3 billion (Up 30% YoY)
- Next Quarter Outlook: $65 billion
The gaming side of things is actually doing okay, which people forget about. Revenue was up 30% year-over-year. They launched DLSS 4 with Multi Frame Generation and have been pushing the RTX 50-series hard. But let’s be real—gaming is the side hustle now.
What Really Happened With Those Circular Deals?
One of the spicier moments of the call involved "circular deals." This is the idea that Nvidia is investing in startups like OpenAI and Anthropic, who then turn around and use that money to buy Nvidia chips. It looks a bit like a merry-go-round of cash.
Jensen dismissed this pretty quickly.
He argued that these investments are about expanding the entire ecosystem. If OpenAI succeeds, the entire market for AI compute grows. He’s basically betting that by funding his biggest customers, he's ensuring that the "sold out" status of his chips stays that way for years.
The Rubin Surprise
While everyone was obsessed with Blackwell, the team dropped some serious hints about Rubin. This is the platform coming after Blackwell. At CES 2026, Jensen confirmed that Rubin is already seeing orders.
It’s built using something called "extreme codesign."
Basically, they aren't just making a faster chip; they are designing the CPU, GPU, and networking together so they don't bottleneck each other. It’s supposed to give 5x the performance of Blackwell for some AI tasks. If Blackwell is a "thinking machine," Rubin is intended to be something even more autonomous.
Why Most People Are Misreading the Margins
There was some talk about margin pressure. GAAP gross margins were 73.4%. That’s down slightly from the 74.6% they had a year ago.
Input costs are going up.
Everything from HBM3e memory to the complex packaging at TSMC is getting more expensive. Colette Kress noted that while costs are rising, they expect to hold margins in the mid-70s for the rest of the fiscal year. They are essentially passing those costs onto the big cloud providers who have no choice but to pay.
Practical Steps for Following NVDA
If you are trying to make sense of all this for your own portfolio or just to stay smart at work, don't just look at the stock price the day of the call. Here is how you should actually track the NVDA earnings call live impact over the next few weeks:
- Watch the Hyperscaler CapEx: Keep an eye on the earnings reports from Microsoft, Google, and Meta. If they confirm they are spending more on "AI factories," Nvidia wins.
- Monitor the "Rubin" Timeline: Any news about Rubin silicon arriving at data centers ahead of schedule is a massive bullish signal.
- Check the Networking Revenue: Nvidia's networking segment (InfiniBand and Spectrum-X) grew 162%. This is the "glue" that holds AI supercomputers together. If this stays high, it means companies are building bigger clusters, not just buying individual chips.
- Ignore the "Bubble" Noise temporarily: Focus on the "Inference" stats. Jensen mentioned that inference is now growing exponentially. If companies are actually using AI to serve customers (inference) rather than just training it, the demand is real and not just speculative.
The next big milestone is the Q4 fiscal 2026 report where they expect to hit $65 billion in a single quarter. That’s a run rate that would have been unthinkable two years ago. Whether they can actually manufacture enough Blackwell chips to hit that number is the only question that matters right now.