Nvidia Stock Swings Higher Post-Earnings as Jensen Huang Teases AI Agent Transition
Nvidia, the world's most valuable company, released its latest earnings report last night, and after the release, the stock fell initially nearly 2% until CEO Jensen Huang jumped on the call with optimism, which pumped it 4% up in after-hours trading. Pre-market, it went as high as 7% up, and at the time of writing, it is 5.5% up pre-US cash open.
Traders brace for the market open to trade Nvidia shares, NQ / Nasdaq, S&P 500, and even companies looking to position their risk just like a macro event.
Nvidia released its Q2 FY2027 earnings on August 26, 2026, delivering another "double beat" that surpassed consensus estimates across top- and bottom-line metrics.
Key Financial Results
Record Revenue: $96.22 billion (up 106% YoY), beating Wall Street expectations of ~$92.16 billion.
Adjusted EPS: $2.22 per share, topping estimates of $2.08–$2.09.
Data Center Revenue: $89.0 billion (up 117% YoY), accounting for over 92% of total sales.
Gross Margins: Held strong at 75.0%.
Future Guidance & Takeaways
Q3 Revenue Guidance: Projected to reach $108.0 billion (±2%), topping the ~$104.6B consensus.
Platform Ramping: CEO Jensen Huang emphasized that demand for the next-generation Vera Rubin platform is ramping heavily across major cloud providers and AI-native startups, noting that physical AI and agentic workflows are driving compute needs 15x to 100x higher than human-driven tasks.
China Outlook: The Q3 guidance conservatively assumes zero Data Center compute revenue contribution from China.
The Negatives & Bottlenecks
Severe Supply Constraints: Jensen candidly stated that supply limits FY28 growth to ~70%, despite customer demand actually calling for 100%+ growth. Nvidia is completely supply-bound for the foreseeable future.
Gross Margin Compression Ahead: CFO Colette Kress warned that gross margins will step down to ~74% in Q3 and dip to 71–72% in Q4 due to high component/memory squeeze costs before recovering in FY28.
Rising Competition / In-House Chips: Analysts repeatedly pressed Jensen on Big Tech making custom chips (XPUs). Jensen defended Nvidia as a "full-stack platform" rather than just a chip maker, but custom silicon remains an ongoing market overhang.
Expectations are certainly slowing down from quarter to quarter, yet Jensen Huang at Nvidia is certainly still hitting the ball out of the park. But we need to remember as Nvidia has capitalized on the LLM chatbot era, they are now moving into the AI agent era as we move from models like the Nvidia Blackwell to the Nvidia Vera Rubin, which he announced earlier this year in his fantastic presentation.
Blackwell (B200 / B300): Designed primarily as the foundational powerhouse for training and running massive LLMs (Large Language Models) and Mixture-of-Experts (MoE) models. It was engineered to scale raw compute density and memory bandwidth to train trillions of parameters efficiently.
Vera Rubin: Specifically engineered as Nvidia's first "Agentic AI Native" architecture.
Why Vera Rubin is called the "AI Agent" chip: AI agents operate differently than standard LLMs. Instead of just taking a prompt and outputting a single text response, agents run continuously, execute multi-step reasoning loops, call external tools, and manage massive token context histories over long sessions.
To handle agentic workloads without hitting a bottleneck, the Vera Rubin platform introduces:
Extreme Memory Bandwidth (HBM4): Delivers up to 22 TB/s of bandwidth (nearly triple Blackwell), which is critical because AI agents executing long reasoning chains are heavily memory-bound rather than just compute-bound.
1M+ Token Context Support: Specifically built to process massive long-context memory windows so agents don't "forget" prior actions or code iterations.
Massive Efficiency Gains: Nvidia targets up to 10x to 30x higher throughput per megawatt over Blackwell when running sustained agentic coding and multi-agent workflows.
Companies like AMD, as a competitive chip maker, try to shorten the gap, yet Nvidia is so many years ahead when it comes to graphics, as they spent years in the darkness and still hold the monopoly board—don't pass go, don't collect £200.
Another thing worth thinking about is... they have had the LLM AI chatbot build out, and they are now having their AI agent build out glory, but what about when that is built out or at least priced in? Is the hype over? Or will humanoid robotics fill that gap? That's debatable, as it's well known that there is a big lag between software and hardware when it comes to robotics, and then where does that leave Nvidia's share price? Is that what they call in markets "the top"? Maybe not just for Nvidia shares, but for the NASDAQ stock index and also the economy itself. It's all speculation at the end of the day, and traders don't make prophecies—they make dollars.