NVIDIA founder Jensen Huang: OpenAI will become the next trillion-dollar mega company.

Written by: Techub News Compilation

Introduction

Recently, NVIDIA founder and CEO Jensen Huang appeared again on the well-known podcast "BG2" hosted by investors Bill Gurley and Brad Gerstner. More than a year has passed since their last conversation, and the AI world has undergone "century" changes. This in-depth conversation lasting 104 minutes focuses on the recent shocking industry strategic collaboration between NVIDIA and OpenAI, the exponential growth prospects of AI computing demand, NVIDIA's annual release rhythm and system-level competitive barriers, as well as the geopolitics of the AI era and the new connotations of the "American Dream." As a core engine driver of the AI revolution, Jensen Huang's statements provide critical insights into understanding the current arms race in AI infrastructure and the future technological economy landscape.

Summary

Jensen Huang asserts that OpenAI is highly likely to become the next multitrillion-dollar mega company, and NVIDIA's investment is based on this immense potential.
AI computing has three major scaling laws: pre-training, post-training (reinforcement learning), and inference (thinking). The demand for inference computing power will grow a billion times due to "thinking," which is a dramatic industry change that most people have not fully understood.
NVIDIA is transforming from a chip company to an AI infrastructure partner, responding to the dual exponential growth of computing demand through "extreme collaborative design" and an annual release cycle for systematic performance enhancement.
Jensen Huang believes that the era of general computing has ended, and the future belongs to accelerated computing and AI computing. The trillions of dollars of global computing infrastructure will be refreshed, which represents a historical opportunity for NVIDIA.
Regarding the US-China technology competition, he calls for maintaining confidence, attracting global talent, and supporting social innovation such as "Invest in America" to ensure the prosperity brought by AI can be widely shared.

OpenAI: The Birth of the Next Trillion-Dollar Giant

The conversation began with the most attention-grabbing industry event: the strategic collaboration between NVIDIA and OpenAI. Jensen Huang confirmed that the two parties are working together on multiple projects, including continuously building capabilities for Microsoft Azure, collaborating with Oracle Cloud Infrastructure (OCI) and SoftBank, and providing computing power through CoreWeave. The recently announced partnership is centered on NVIDIA helping OpenAI establish its own full-stack AI infrastructure for the first time.

"This involves directly collaborating with OpenAI on the chip level, software level, system level, and AI factory level to help them become a fully independently operating mega company," Jensen Huang explained. This means that OpenAI is shifting from relying on cloud vendors like Microsoft to build data centers to establishing direct, comprehensive procurement and collaboration relationships with NVIDIA, similar to Elon Musk's X.AI or Meta.

Why is this move so important? Jensen Huang provided a shocking prediction: "I think OpenAI is very likely to become the next multitrillion-dollar mega company." He compared it to Meta and Google, believing that OpenAI will simultaneously serve both consumer and enterprise services, growing into a new industry giant. Therefore, NVIDIA has the opportunity to invest before it becomes a giant, "this is one of the wisest investments we can imagine."

He further analyzed the "dual exponential" driving this growth: first, the exponential growth in customer numbers, as AI capabilities become increasingly strong and almost all applications are integrating AI; second, the exponential growth in computing consumption for each use case, as AI transitions from "one-time inference" to "thinking before answering." The combination of these two exponentials leads to a compounded explosive growth in OpenAI's demand for computing power, which is the fundamental reason for building multiple parallel projects and adding its own infrastructure.

The Three Scaling Laws and "Billion Times" Computing Demand

Jensen Huang reiterated and deepened his views from a year ago: the future growth of AI computing goes far beyond pre-training. He systematically proposed three scaling laws:

Pre-training scaling law: traditional model training in the conventional sense.
Post-training scaling law: primarily refers to reinforcement learning, where AI masters skills through extensive practice, requiring a huge amount of inference computing power.
Inference scaling law: inference in the new era is no longer "one-time answers," but rather "thinking." AI conducts research, verifies facts, and iterates before generating higher quality answers.

"The longer the thinking time, the higher the quality of the answers obtained," Jensen Huang emphasized. Today, AI is no longer a single language model but a system composed of multiple concurrently running models, potentially using tools, and is multimodal. The explosion of applications like video generation is a manifestation of surging computing demand.

Based on this, he is more confident than a year ago: the demand for inference computing power will grow by "a billion times." This prediction is not a fantasy but is based on the evolution of current agent systems and actual computing consumption trends. Wall Street analysts believe NVIDIA's growth will level off after 2027; however, Jensen Huang thinks they severely underestimate the scale of this industrial revolution transitioning from general computing to accelerated computing.

NVIDIA's Transformation: From Chips to AI Infrastructure

Faced with huge market opportunities, Jensen Huang clearly articulated the evolution of NVIDIA's positioning. "People mistakenly think we are a chip company... but NVIDIA is fundamentally an AI infrastructure company." He stated that NVIDIA is the AI infrastructure partner for its customers, and the collaboration with OpenAI perfectly embodies this role.

He proposed a three-layer logic to depict market opportunities:

Layer of physical laws: the end of the general computing era; the future is accelerated computing and AI computing. Trillions of dollars of global computing infrastructure need to be updated.
Layer of existing use case migration: AI's first large-scale applications are ubiquitous, such as search, recommendation engines, and shopping. These workloads, originally completed by CPUs, are migrating on a large scale to GPUs and AI computing. This alone involves hundreds of billions of dollars in the market.
Layer of future new applications: when the computing paradigm shifts to AI and accelerated computing, it will spur entirely new applications. This is similar to replacing manual labor with motors; now it is about using AI (generating tokens) to enhance human intelligence. Human intelligence accounts for about 50-60% of global GDP (about $50 trillion), this portion will be enhanced by AI. Even if only enhancing by 10% ($5 trillion), it will require enormous AI infrastructure (factories) to support.

Jensen Huang cited internal company practices as an example: NVIDIA equips every software engineer and chip designer with AI assistants, achieving 100% coverage. This has increased productivity, revenue, and profits, and enabled the company to hire more people and pursue more creativity. "Scaling NVIDIA’s story to global GDP, this is the future that could happen."

Annual Releases, Extreme Collaborative Design, and Competitive Barriers

To meet the exponential growth of computing demand, NVIDIA has accelerated its chip release rhythm to an annual cycle (Hopper, Blackwell, Rubin, Ultra, Fermi...). Jensen Huang explained that the fundamental reason driving this transformation is "the token generation rate is exponentially rising." If performance does not improve at an astonishing speed, the cost of generating tokens will continue to rise due to the failure of Moore's Law.

"From Hopper to Blackwell, we achieved a 30-fold performance improvement through technologies like NVLink. This is not reliant on Moore's Law, but rather on extreme collaborative design." He elaborated that extreme collaborative design means optimizing the models, algorithms, systems, and chips simultaneously, innovating outside of Moore's Law's "box." NVIDIA is simultaneously innovating CPUs, GPUs, networking chips, NVLink (vertical scaling), and Spectrum-X (horizontal scaling), optimizing them as an integrated system.

This model builds exceptionally high competitive barriers:

Technical barriers: requiring the simultaneous design of multiple chips and optimization of the entire software stack, resulting in extremely high complexity.
Scale and supply chain barriers: customers often deploy gigawatt-level data centers containing hundreds of thousands of GPUs, requiring the supply chain to plan years in advance and invest hundreds of billions of dollars. Only companies like NVIDIA, with mature architectures and a solid track record, can gain supply chain and customer trust for such large-scale upfront investments.
System thinking barriers: competitors may still be designing single-function ASICs, while NVIDIA is building and optimizing the entire "AI factory" system and beginning to launch dedicated processors for specific workloads (like CPX for contextual processing).

Regarding competitors like Google TPU, Jensen Huang expressed respect but also pointed out the challenges: early market entrants have a first-mover advantage, but now AI infrastructure has become "gigantic and complex," and it continues to evolve rapidly. As a latecomer, it is extremely difficult to enter such a vast and variable market. He even made a sharp point: even if competitors' chips were provided for free, customers should still choose NVIDIA's systems. The reason is that data centers are limited by power, while NVIDIA systems' "performance per watt" (tokens per watt) is far superior to rivals, using free but inefficient chips will incur significant opportunity costs.

Geopolitics, Talent Competition, and the American Dream

As a key provider of global AI infrastructure, Jensen Huang inevitably spoke about geopolitics, especially US-China relations. He first emphasized: "No one needs an atomic bomb, but everyone needs AI." AI is the reshaping of modern computing, and all countries need to participate, which has given rise to the demand for "sovereign AI."

He candidly admitted that the current geopolitical environment has put NVIDIA in a "very difficult" position. On one hand, he understands and supports the necessity of protecting US national security; on the other hand, he is deeply concerned that excessive restrictions will harm the foundation of US innovation—the ability to attract and retain global talent.

"I heard that three years ago, 90% of top AI researchers graduated from top universities in China who hoped to and indeed came to the US. Today, this proportion may have dropped to 10% to 15%." Jensen Huang believes this is an early indicator of a "survival crisis." He criticized the so-called "hawkish" label against China, arguing that cutting off the talent pipeline is "not patriotic at all."

"We need to have the confidence of a great nation... our attitude should be: bring it on." Jensen Huang said he believes in the people, culture, systems, and institutions of the US. He observed former President Trump expressing respect when discussing China, and he never used the term "decoupling"; he sees this as the correct direction. "You cannot decouple from one of the most important bilateral relationships of the next century; it makes no sense."

When discussing the "American Dream," Jensen Huang linked it to "rising rights." He praised and supported the "Invest in America" plan—starting in 2026, every child born in the US will receive an initial investment account for investing in outstanding American companies. "This ensures that every child is a stakeholder in America's future." He believes that in the era where AI brings exponential advancements, the social contract needs to evolve in tandem to ensure prosperity is widely shared.

In response to concerns that AI could lead to unemployment, Jensen Huang holds an optimistic view. He believes AI is a "great equalizer," eliminating the technological gap (now you only need to interact with AI using human language). He dismissed the assumption that "AI will lead to massive unemployment," as it is based on the erroneous premise that "humans have no new ideas." "Intelligence is not a zero-sum game. The more geniuses around me, the more creative ideas I can think of, the more problems we can solve, and the more jobs we create."

Future Outlook: Embrace the Exponential, Get On the Train

Looking ahead, Jensen Huang predicts that significant breakthroughs will occur in the integration of artificial intelligence and mechatronics (robotics) within the next five years. Everyone will have a cloud-exclusive AI, possibly embodied in vehicles, robots, and other entities. Digital twin technology will be widely applied in healthcare to predict health risks and diseases.

When asked how businesses and individuals should respond to this accelerating world of change, Jensen Huang provided a vivid metaphor: "If you have a train that is getting faster and accelerating exponentially, the only thing you really need to do is get on it. Once you're on, you'll naturally figure everything out along the way." Attempting to predict the future location of this exponentially accelerating train and intercepting or waiting for it in advance is impossible. The best strategy is to board while it's still relatively slow and then grow exponentially with it.

"Don't scare people; lead them forward together." Jensen Huang concluded the conversation in this way. This may very well be the core initiative this leader of the AI revolution proposes for the industry, society, and even the nation in this era of technological frenzy.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。