Nvidia’s Trillion-Dollar Forecast and the Next-Gen Rubin Platform

At its recent GTC 2026 conference in San Jose, Nvidia made a bold revision to its market outlook. Chief Executive Jensen Huang now anticipates that cumulative demand for artificial intelligence chips will hit the $1 trillion mark by 2027. This projection represents a doubling of the company’s forecast from just a year ago, which had targeted that figure for 2026.

Financial Performance and Market Sentiment

The company’s financial results provide a strong foundation for this optimism. For the full fiscal year 2026, Nvidia reported total revenue of $215.9 billion, a 65% increase year-over-year. Its data center segment was the primary driver, contributing $192 billion. Looking ahead, management has guided for first-quarter fiscal 2027 revenue of approximately $78 billion, which would equate to growth of 77%.

Despite these robust figures and the ambitious new forecast, the market’s immediate reaction was measured. Following the keynote, Nvidia shares closed with a gain of 1.65%—a solid but not spectacular move. Analyst commentary, however, remained largely bullish. Goldman Sachs reaffirmed its Buy rating with a $250 price target, citing improved visibility for data center orders through 2027. Bank of America was even more optimistic, setting a $300 valuation and emphasizing Nvidia’s strength as a fully integrated platform provider. JPMorgan analysts noted that the tight integration of CPUs, GPUs, and networking components is becoming increasingly difficult for competitors to replicate.

Morgan Stanley offered a more tempered perspective, observing that with a forward price-to-earnings ratio of about 21.8, the market is currently valuing Nvidia only slightly above the broad S&P 500 index—an unusual scenario for a company exhibiting such dynamic growth.

Should investors sell immediately? Or is it worth buying Nvidia?

Introducing the Vera Rubin Architecture

Central to Nvidia’s future roadmap is the unveiling of its next-generation platform, named Vera Rubin. This architecture succeeds the current Blackwell generation and is comprised of seven chips. These include new GPUs, a novel ARM-based Vera CPU featuring 88 cores, and dedicated networking components. A standard rack system built on Rubin will integrate 72 GPUs and 36 Vera CPUs, delivering roughly 3.6 exaFLOPS of computing performance. Shipments of these systems are scheduled to begin in the second half of 2026.

The company claims the Vera CPU achieves double the energy efficiency of current x86 processors for AI workloads. Notably, the portfolio also includes the “Space-1 Vera Rubin Module,” designed for orbital data centers, which promises 25 times the inference performance of the previous H100 generation.

The Strategic Pivot to Inference and Autonomous Agents

A key theme of the presentation was the industry’s shift from AI training to what Huang termed the “Inference Inflection.” He framed the next phase of the AI boom as an era of autonomous agents capable of independently executing tasks. To power this transition, Nvidia is integrating technology from Groq into its hardware stack. The new Groq 3 LPX inference chip, manufactured by Samsung, is slated for delivery starting in the third quarter of 2026.

On the software front, the company launched its “Agent Toolkit.” This suite includes NemoClaw, an enterprise-grade version of the open-source framework OpenClaw, and the OpenShell runtime environment. This software is designed to enable local AI agents to operate on everything from RTX-powered PCs to DGX supercomputers.