$100B in AI Funding Isn't a Bubble, It's the Biggest Buildout in History: Nvidia CEO

The year 2025 marked the largest investment year in venture capital (VC) history, with over $100 billion invested globally, most of it flowing into AI-native companies, said Jensen Huang, founder and CEO of NVIDIA Corporation.

He was speaking in a conversation with Laurence D. Fink, Chair and CEO of BlackRock, at the World Economic Forum 2026 in Davos on Wednesday.

According to Huang, this surge in investment reflects a structural shift in AI maturity. For the first time, AI models are good enough to support a full application layer, and those applications require large-scale infrastructure. As a result, significant capital is now being deployed to build the foundations needed to power the next phase of AI-driven growth.

On concerns of an AI bubble, Huang argued that the world is underinvesting. GPU shortages, rising compute prices, record venture funding, and companies shifting R&D budgets toward AI infrastructure all signal sustained demand. He described this as the largest infrastructure buildout in history and a once-in-a-generation wealth creation opportunity.

The Five-Layer Cake

Huang describes AI as a five-layer cake that works from the base to the top. At the base is energy, as AI systems require massive power to generate intelligence in real time. Above that are chips and computing infrastructure, which provide the raw processing capability. The third layer is cloud services, which distribute this compute capacity at scale. The fourth layer consists of AI models, where most public attention has been focused.

However, Huang argues that the most important layer sits at the top: applications. This is where AI delivers real value across industries such as finance, healthcare, and manufacturing. It is the application layer that turns infrastructure and models into economic outcomes.

He noted that AI requires all five layers to work together, a dynamic that has triggered what he described as the largest infrastructure buildout in human history. Hundreds of billions of dollars have already been invested, and this is still the early stage. The application layer is only now beginning to scale, which explains why 2025 became a breakthrough year for AI.

“There are trillions of dollars of infrastructure that needs to be built out, and it’s sensible,” Huang said. “All of these contexts have to be processed so that the models can generate the intelligence necessary to power the applications that ultimately sit on top.”

Real-World Infrastructure and Industry Trends

Beyond Nvidia’s narrative, the broader market reflects this infrastructure boom. Major technology firms including OpenAI, Microsoft, and Google are pouring billions into data centers, chips, and networking equipment as demand for AI compute continues to surge. Analysts estimate that up to 75% of companies may invest in agentic AI infrastructure in 2026, further driving demand for chips and data centers.

Chip manufacturers themselves are responding with record capital spending. Taiwan Semiconductor Manufacturing Company (TSMC) recently announced plans to increase its investment to $52–$56 billion in 2026 alone, anticipating annual AI chip revenue growth of over 55% through 2029.

At the same time, global supply chains are feeling pressure: memory-chip shortages have emerged due to soaring demand from AI workloads, pushing memory prices up and prompting major capacity expansions that won’t fully ease supply constraints until 2028.

Breakthroughs That Changed AI in 2025

Huang identified three major breakthroughs from 2025 that transformed AI from an experimental technology into a practical one.

The first was the shift from chatbots to agents. Early AI systems were interesting but unreliable, often hallucinating or producing inconsistent outputs. Over the past year, models became more grounded and capable of genuine reasoning. This marked a transition from basic language models to agentic AI systems that can act independently.

The second breakthrough was the rise of open models. Huang pointed to DeepSeek’s release as a watershed moment. As one of the first open reasoning models, it enabled companies, researchers, universities, and startups to build specialized AI systems without being locked into proprietary platforms. This openness expanded experimentation and accelerated adoption across industries.

The third breakthrough was the emergence of physical AI. AI systems are no longer limited to understanding language; they are increasingly capable of modelling the physical world.

Job Transformation, Not Elimination

Concerns that AI and robotics will eliminate jobs remain widespread. Huang rejected this view, arguing instead that AI is changing the nature of work at a time when the world is facing a labour shortage.

The expansion of chip factories, data centres, and AI facilities is creating strong demand for skilled trade workers such as electricians, plumbers, construction workers, steel workers, and network technicians. In the United States, wages for workers involved in building AI infrastructure have risen sharply, with many roles now offering six-figure incomes. These jobs do not require advanced degrees, yet employers are struggling to find enough workers.

Huang also pointed to healthcare to explain why automation often increases demand for professionals rather than reducing it. Radiology was once seen as vulnerable to AI replacement. Instead, AI has become a routine tool in radiology, while the number of radiologists has grown. AI handles image analysis faster, allowing doctors to focus on diagnosis, collaboration, and patient care.

At the core of Huang’s thinking is the distinction between purpose and tasks. Jobs exist for a purpose, while tasks are the activities used to fulfill that purpose. AI automates tasks, not purpose. By removing repetitive work, AI allows professionals to focus on what their roles are meant to achieve.

AI as National Infrastructure and Policy Context

For AI to transform the developing world, Huang said it must be treated as national infrastructure, similar to electricity, roads, or the internet. Countries should build AI capability locally rather than relying entirely on imports. With open models and accessible tools, training AI is no longer limited to a small group of global companies.

Countries can leverage their own languages, culture, and data to develop what Huang described as “national intelligence” tailored to local needs.

Policy developments are emerging alongside this buildout. The European Union’s AI Act, the world’s first comprehensive legal framework for AI, is now being implemented and enforced, with regulatory stress tests underway and data retention orders issued to major tech platforms. This represents one of the most advanced AI governance efforts globally and could shape how AI systems are developed and deployed.

AI’s Rapid Adoption

AI’s rapid adoption is driven by its usability. Tools like ChatGPT and Claude have reached close to a billion users within a few years because the barrier to entry is low. People can learn to use AI by prompting, directing, and evaluating it, much like managing digital workers. This accessibility helps close the technology divide, enabling people without formal computer science training to write code, build applications, and solve problems.

Huang also highlighted Europe as a major opportunity. While the U.S. led the software era, AI is less about writing code and more about teaching systems. Europe’s strength in industrial manufacturing and scientific research positions it well for physical AI and robotics, provided it invests in energy and infrastructure.