Huang Renxun prévoit un chiffre d'affaires de mille milliards de dollars pour les puces AI, annonce officielle du processeur Groq 3 et renforce l'écosystème de sécurité OpenClaw

Local time Monday, NVIDIA CEO Jensen Huang announced that driven by the adoption of tools such as Anthropic’s Claude Code and OpenAI’s OpenClaw, revenue in the artificial intelligence chip sector is expected to reach at least one trillion dollars within the next two years.

During the flagship GTC conference held in Silicon Valley by the semiconductor giant NVIDIA, with a market value of $4.5 trillion, Huang said: “Based on my current judgment, by 2027, the revenue in this field could reach at least one trillion dollars.” He also added that he is “confident” that actual demand will far exceed this figure.

Following Huang’s statement, NVIDIA’s stock price rose more than 2% at one point, but the rally was not sustained, and it ultimately closed below its pre-speech level.

In recent months, market skepticism about the returns on massive investments in AI infrastructure, conflicts in the Middle East threatening the semiconductor supply chain, and shortages of storage chips needed for NVIDIA’s products have all contributed to Wall Street’s continued disappointment with the company’s optimistic earnings forecasts.

The revenue outlook announced by NVIDIA this time far exceeds Wall Street consensus, signaling the company’s latest stance—that it is betting on the AI boom to continue heating up.

As the largest supplier of advanced AI chips, NVIDIA has thus become the world’s most valuable company by market capitalization.

Deepwater Asset Management partner Gene Munster said that investor concerns about NVIDIA’s long-term growth potential have created a “resistance wall” for the company’s stock.

He stated: “The core conclusion of Jensen Huang’s keynote is that market demand is so strong that it even exceeds the most optimistic expectations, yet investors still find it hard to accept this fact.”

Previously, based on the highly certain market demand and purchase orders for NVIDIA’s latest Blackwell and Rubin hardware series, Huang predicted that by the end of 2026, revenue from AI-related business would reach $500 billion.

The new forecast of $1 trillion in AI hardware revenue far surpasses Wall Street’s general estimates of NVIDIA’s total revenue.

According to data from Capital IQ, analysts’ combined estimates for NVIDIA’s total revenue for fiscal years 2027 and 2028 (ending January 2028) amount to approximately $835 billion.

Huang stated that the popularity of tools like Anthropic’s Claude Code, along with rising market demand for “reasoning computation”—the process of running AI models and related applications—has driven continuous growth in computing power.

In this two-hour keynote, Huang announced a series of new initiatives, including autonomous driving taxi collaborations and a chip designed specifically for orbital data centers; the concept of orbital data centers was proposed by Elon Musk as part of his vision to integrate SpaceX and xAI.

He also unveiled a new chip product—Groq 3 “Language Processing Unit,” designed to improve the response speed of AI systems to user commands.

This move indicates that NVIDIA is attempting to break through the previous development model relying solely on a single graphics processor to support AI workloads, exploring new chip architectures to strengthen its leadership in the AI chip market.

Huang said: “Groq 3 chips are now in mass production and are expected to ship in the second half of 2026, most likely in Q3.”

He also revealed that the chip will be manufactured by Samsung Electronics in South Korea—an attempt for NVIDIA, as previously its AI processors were typically produced by TSMC. This is the first new product launched after NVIDIA’s licensing agreement with startup Groq at the end of last year, which also involved hiring Groq founder Jonathan Ross, who previously participated in the development of Google’s AI chips.

Following the announcement, Samsung Electronics’ stock rose by 4% at one point on Tuesday in the Korean market.

Huang stated that as more companies adopt AI assistant tools like OpenClaw, the importance of reasoning computation will further increase.

In February this year, OpenAI hired Peter Steinberger, the founder of OpenClaw. OpenClaw is an open-source, free tool.

Huang compared OpenClaw to other open-source tools that support the tech industry, including the Linux operating system dominating data centers today and the Hypertext Transfer Protocol (HTTP) that enables browsers to load web pages. He said: “This is the next generation of computers.”

NVIDIA is developing a software layer called “NemoClaw” for the underlying AI assistant of OpenClaw. The company claims this software will build the missing privacy and security mechanisms in OpenClaw’s current standard product.