Analysts on Nvidia GTC Conference: "$1 Trillion Demand Outlook is a 'Floor, Not a Ceiling'"

Investing.com - NVIDIA’s annual GTC Developer Conference reinforced Wall Street’s view that AI spending is entering a new phase, with demand shifting from model training to inference and large-scale deployment.

During the event, Jensen Huang stated that GPU demand “is soaring,” driven by a nearly million-fold surge in computing needs over the past two years and expanding inference workloads.

Gain deeper insights into NVIDIA’s stock outlook with InvestingPro.

A key highlight of the conference was NVIDIA’s updated outlook. The AI chip giant indicated that by 2027, revenue opportunities from its AI chips could reach at least $1 trillion, and outlined plans to compete more aggressively in the rapidly growing real-time AI system market.

This forecast is higher than NVIDIA’s previous mention of a $500 billion opportunity for the Blackwell and Rubin chips by 2026.

At the conference, Jensen Huang announced a new central processing unit and an AI system integrated with Groq technology, a chip startup NVIDIA licensed its technology to last December for $17 billion.

These announcements are part of NVIDIA’s efforts to strengthen its position in inference computing—the process of answering queries—where its graphics processors face increasing competition from custom chips developed by CPUs and companies like Google. NVIDIA has so far dominated AI model training, which has been a major focus in recent years.

“The inference inflection point has arrived,” Huang said. “Demand continues to rise,” he added.

Huang explained that inference will be divided into two phases. NVIDIA’s Vera Rubin chips will handle the “pre-fill” step, converting user input into tokens used by the AI system, while Groq’s chips will manage the “decode” phase, generating responses.

Huang also outlined the Feynman roadmap, expected to launch after Rubin Ultra in 2028, though details are limited beyond including AI and networking chips.

Additionally, NVIDIA is eyeing the autonomous AI agent market, launching NemoClaw, which integrates with the OpenClaw platform, adding privacy and security controls to tools capable of executing tasks with minimal human input.

The company’s stock briefly rose after the announcement, then narrowed its gains to close up 1.65%.

Analyst Views on GTC 2026

Wolfe Research: “NVIDIA held its GTC keynote today. The most important disclosure was an update on the previously announced $500 billion revenue target in October. The new disclosure indicates upside potential for 2027 calendar year revenue, with demand still growing.”

“We believe this revenue figure is sufficiently vague to not reflect firm guidance but still leaves significant upside relative to market consensus. Therefore, we see this as a baseline rather than a ceiling.”

Bernstein: “$1 trillion (the same as the previous $500 billion) is just a current snapshot; there are seven more quarters before the end of 2027 to achieve more (the company hints at continued growth). More importantly, Colette confirmed to us that this figure only includes Blackwell and Rubin (and related networks); it excludes any other products (such as Groq LPU, CPX, CPU racks, etc.). So, we suspect the data center business will be far above this $1 trillion target and well beyond expectations.”

“NVIDIA’s roadmap looks very solid, their capability gap continues to widen, and new products should help solidify their position in inference, just as they dominate training. Order books indicate further upside, and given their positioning, the stock (in our view) is trading at an almost absurdly low valuation (about 15 times our 2026 calendar year / 2027 fiscal year EPS). We are buying.”

Goldman Sachs: “NVIDIA provides visibility into strong growth prospects for 2027, consistent with our estimates and well above market expectations—we believe this helps address investor concerns about the ‘peak capital expenditure’ in 2026. The launch of Groq’s LPX racks helps reinforce the company’s commitment to the inference market—we see this as a key but increasingly competitive growth area in AI infrastructure.”

Morgan Stanley: “The core message is that NVIDIA’s inference advantage per token is clear, and as Rubin improves, our survey agrees.”

“Our view is that the company’s market share will be more stable than the market thinks, and the momentum in AI spending will last longer.”

“(NVIDIA) remains the preferred stock in semiconductors, with clear opportunities to catch up with peers and AI supply chain stocks.”

Stifel: “The headline disclosure is that by 2027, the visibility of cumulative orders for the Grace Blackwell and Vera Rubin platforms reaches $1 trillion, reaffirming demand acceleration and indicating that the ‘AI factory’ construction is still speeding up. The strategic shift involves unbundling CPUs and network stacks, integrating Groq LPU to capture inference advantages, and launching OpenClaw/NemoClaw, which Huang described as the ‘HTTPS moment’ for AI agents.”

This article was translated with the assistance of artificial intelligence. For more information, see our Terms of Use.