One Statement That Settled the AI Bubble Debate

A session at the 2026 Davos Forum featuring Larry Fink, Chairman of BlackRock (the world's largest asset manager), and Jensen Huang, CEO of NVIDIA (the company with the largest market capitalization), has drawn significant attention. At this event, Jensen Huang presented a perspective that decisively addressed the "AI bubble" debate.

Here's the key point: AI is no longer simply a technology—it's becoming essential infrastructure, like electricity, roads, and telecommunications. Once you accept this premise, the ongoing investment appears not as overheating but as "the largest infrastructure build-out in modern history."

NVIDIA's Growth by the Numbers

Larry Fink opened the conversation with an intriguing comparison. BlackRock and NVIDIA went public almost simultaneously in 1999, but their trajectories have been strikingly different. NVIDIA has achieved an annualized compound return of 30-37%, and Fink remarked, "Imagine what it would have meant if pension funds had participated in the IPO back then."

Jensen Huang shared an anecdote: shortly after the IPO, when the market cap was $300 million, he sold some shares to buy his parents a Mercedes S-Class. He joked that it was "the most expensive car in the world at the time," but from today's perspective, it sounds like a story from an entirely different economic era.

Platform Shift: Why This Cycle Is Different

Fink's question was fundamental: "Why can AI be such a powerful growth driver? What makes it different from previous technology waves?"

Jensen Huang responded not with marketing slogans but with a "return to first principles." His core argument is clear: this is a platform shift similar to PC, internet, mobile, and cloud—and with each shift, the computing stack was reinvented and new classes of applications emerged.

Fundamental Change in Programming Paradigm

However, the essence of this change lies in the nature of programming itself.

Jensen Huang described SQL as "the most important database engine the world has known," emphasizing that nearly everything operated on top of it. But now, a new type of computer has emerged that can understand and reason about unstructured data.

Five-Layer AI Stack and $1 Trillion Infrastructure Build-Out

The "five-layer AI stack" concept presented by Jensen Huang was the core framework of the session. Although not all layers were detailed in the original discussion, he emphasized that each layer requires new infrastructure.

• Hardware Layer: AI accelerators like GPUs and TPUs
• Network Layer: Ultra-high-speed data transmission infrastructure
• Software Layer: AI frameworks and development tools
• Application Layer: Actual AI services
• Data Layer: Training and inference datasets

Jensen Huang described the investment scale needed to build this stack as a "trillion-dollar build-out." This means infrastructure investment exceeding $1 trillion is required.

Structural Causes of GPU Shortage

The conversation naturally led to GPU supply shortages. From Jensen Huang's perspective, this isn't simply a supply-demand imbalance.

It's a structural phenomenon occurring because the entire world is simultaneously building new computing infrastructure. Just as with electrical grids or road networks, supply struggles to keep pace with demand during the initial construction phase.

Comparison with Previous Technology Waves

Jensen Huang explained that the computing stack was reinvented with each platform shift, and massive infrastructure investment was required each time. The AI era follows this pattern but is different in scale and speed from anything before.

Larry Fink's Investor Perspective

As BlackRock's chairman, Fink connected this discussion to asset management perspectives. He asked how pension funds and institutional investors should view this change, and Jensen Huang's response consistently returned to "infrastructure investment."

Investment in AI is not speculation but essential infrastructure construction, fundamentally the same in nature as 20th-century power grid or highway construction.

[AI Analysis] Future Outlook and Implications

The implications of the Davos conversation between Jensen Huang and Larry Fink are clear.

First, the AI investment cycle is likely to continue for at least 10 years or more. PC, internet, and mobile each underwent 10-15 years of infrastructure build-out, and AI is expected to have an even broader impact.

Second, GPU shortages are a structural problem unlikely to be resolved quickly. With the entire world simultaneously building AI infrastructure, supply expansion will take time. Competition will intensify not only among NVIDIA, AMD, and Intel but also among big tech companies developing their own chips.

Third, the change in programming paradigm will reshape the entire software industry. The shift from SQL-centered structured data processing to neural network-based unstructured data processing means not just a technical upgrade but the birth of new developer ecosystems and business models.

Fourth, the "AI bubble" debate may itself be based on a flawed premise. During essential infrastructure construction phases, what appears to be overinvestment often proves appropriate in hindsight. However, at the individual company level, there will still be winners and losers.

This Davos 2026 session elevated the level of AI discourse. The conversation's center of gravity has shifted from "Are chatbots good or bad?" to "How do we efficiently build $1 trillion in infrastructure?"