Concern is rising over how much the world’s biggest tech companies are pouring into artificial intelligence, but NVIDIA CEO Jensen Huang says the eye-popping numbers shouldn’t shock anyone. In his view, the current surge in AI capital expenditures is not reckless overspending—it’s a sustainable, once-in-a-generation infrastructure build-out that mirrors the early days of other major computing shifts.
The AI boom is accelerating as the industry heads toward 2026, and recent earnings updates from major players have reinforced the idea that AI is nearing an inflection point. That momentum helps explain why AI spending has exploded, with annual capital expenditures now topping levels many observers never expected and pushing beyond $100 billion. A large share of that money is going into the core plumbing of AI: data centers, advanced chips, networking, and the systems needed to train and serve large-scale models.
Skeptics aren’t only reacting to the size of the spending. The bigger question is whether these massive investments will deliver the kind of returns investors are hoping for. Over recent weeks, the combined AI-focused capital commitments discussed across leading tech giants have reached roughly $660 billion, well above many estimates. That gap between expectations and reality has fueled doubts about what’s happening behind the scenes—and whether the payoff will arrive fast enough to justify the buildout.
Huang’s argument is that demand is genuinely “sky-high” because AI has rapidly evolved from an interesting novelty into something broadly useful for businesses and consumers. He points to the idea that AI has crossed into an era where usage can be meaningfully monetized, highlighting that leading AI labs are already generating significant revenue. From this perspective, companies aren’t spending blindly—they’re responding to real demand signals and positioning themselves for the next major platform shift.
A key part of Huang’s optimism is his belief that the software landscape itself is changing. In the traditional view, software is a tool you operate—like a spreadsheet. In the AI era, software becomes something that can operate tools on your behalf. AI systems can use applications, workflows, and data sources to complete tasks, which opens the door to an enormous new wave of “software that does” rather than “software you use.” Huang describes this as the largest software opportunity in history, suggesting that this transition could justify the scale of today’s investment.
That shift is already showing up in the fast-moving “implementation layer” of AI, where new products are being built around agent-like systems that can plan, execute, and iterate. The idea is that AI isn’t limited to chatbots anymore—it’s becoming a practical engine for building apps, automating business processes, and generating real-world outcomes at speed. As these agentic AI experiences improve, the value of the underlying infrastructure becomes easier to argue.
Some industry voices have compared the current race to a modern-day gold rush, where companies like Amazon, Meta, Google, Microsoft, and others are digging aggressively to capture the next dominant computing platform. Supporters of this view say that pulling back too early could mirror past moments when investors underestimated transformative shifts—such as cloud computing—only to realize later that the winners had already built massive leads.
Still, concerns remain. Critics warn that AI could face a mismatch between capacity and near-term demand, echoing historical comparisons like the “dark fiber” overbuild that followed the dot-com era. In that scenario, too much infrastructure gets built too quickly, and it takes longer than expected for usage, pricing, and profits to catch up.
For now, Huang is firmly in the camp that the spending is sustainable and that the opportunity is large enough to support it. Whether the AI investment wave ultimately proves justified will depend on how quickly AI moves from early adopters to mass usage—and whether it delivers clear productivity gains and new revenue streams at scale. The next couple of years should reveal whether today’s AI infrastructure boom becomes a foundation for long-term growth or a costly overbuild that takes time to pay off.
