Yet seasoned investors are uneasy, and cracks are beginning to appear in what many had assumed was a bulletproof market.

Just a few years ago, a global semiconductor shortage reminded the world of the fragility of modern supply chains. Around 11 million cars went unproduced, Apple scaled back iPhone production, and economists estimated that the U.S. alone absorbed a staggering US$240 billion in losses in 2021 as production lines ground to a halt. That crisis underscored the critical importance of infrastructure to the functioning of contemporary economies.

Now, industry leaders are warning of a threat potentially far more disruptive than chip shortages or assembly-line delays: a looming shortage of power, driven not by factories or vehicles, but by the meteoric rise of AI.

Financial titans caution of looming AI risk

Warren Buffett, spoke candidly recently about the scale of AI spending. “We’re sitting here in late 2025, and I’m watching the biggest tech companies in America spend $400 billion on AI infrastructure in a single year,” he said.

“Microsoft, Amazon, Google, Meta… they’re all in a race to build data centres that could cover Manhattan. They’re buying chips like there’s no tomorrow, hundreds of billions with companies that don’t even have revenue yet.”

Buffett’s concerns are as much about market psychology as technology. “I’m getting that same sinking feeling I had back in 1999. I’ve lived through enough market cycles to know one simple truth: when everyone’s running in the same direction, look behind you to see what’s chasing them. More often than not, they’re running toward a cliff.

“The AI revolution is real, I won’t deny that. But what’s happening in the markets right now that’s not investing, that’s speculation dressed up in a three piece suit and history as a way of punishing speculation, no matter how smart the people involved like they are.”

Energy: The true bottleneck

Buffett’s caution is echoed by the tech elite themselves. Elon Musk, chief of Tesla and xAI, has repeatedly warned that the next major shortage will not be chips, but electricity. He warns that AI’s exponential appetite for power could soon outstrip available capacity, suggesting that the tipping point might arrive as early as mid-2026, when AI data centres could consume more electricity than some entire nations.

Mark Zuckerberg, who is steering Meta’s future around generative AI, has also sounded the alarm. Speaking on a recent podcast, he said: “Before we hit capital constraints, we’ll run into energy constraints.” His comments illustrate a point that is often overlooked: the cost and availability of electricity could ultimately dictate the pace of AI adoption.

Amazon CEO Andy Jassy has joined the chorus, highlighting the power demands of large language models like ChatGPT. “Every new model demands exponentially more compute, and exponentially more electricity,” he told investors, hinting at what could become one of the largest infrastructure challenges of the decade.

The grid under pressure

Media coverage of the issue has been increasingly stark. The Wall Street Journal warned that AI expansion could push the nation’s power grid to a “breaking point,” while the Washington Post cautioned that America is running out of power.

The problem is clear: massive AI server farms, often the size of small cities, are being constructed faster than utilities can upgrade transmission lines to feed them. The consequences are already tangible: local brownouts, rising electricity prices, and warnings from grid operators that demand may soon exceed supply in multiple regions.

Even Nvidia, the trillion-dollar chipmaker at the heart of the AI revolution, has recognised the energy challenge. CEO Jensen Huang told Bloomberg that the industry is entering a “power-limited” era.

“The world needs to build energy infrastructure as fast as it builds AI infrastructure,” he said. “Every single data centre in the future will be power-limited. Your revenues will associate with the power you have to work with.”

The AI power crunch

Experts are already referring to this emerging challenge as the great AI power crunch, a collision between human ambition and the physical limitations of our infrastructure.

The rush to develop ever-larger AI models has triggered an unprecedented surge in electricity demand. Some estimates suggest that data centres could account for more than 10% of global power consumption within the next decade, up from just 2% today.

This exponential increase in energy use carries profound implications. Economically, it may slow AI adoption in regions where electricity is expensive or unreliable. Environmentally, the surge in power consumption could significantly impact emissions unless renewable energy infrastructure scales at an equally rapid pace.

Politically, governments and utilities may be forced to prioritise energy allocation, raising questions about equity and access in the AI-driven economy.

Will ambition burn against reality?

For now, servers hum, GPUs burn, and algorithms grow smarter by the second. But between the bright promise of artificial intelligence and the dark limits of the power grid, a new kind of crisis is emerging – one that code alone cannot solve.

The challenge is both technical and societal. Building AI infrastructure is relatively straightforward for companies with deep pockets. Scaling the energy grid, on the other hand, requires years of planning, regulatory approval, and capital-intensive investment. Without careful coordination, the race for AI supremacy could outpace the very power needed to sustain it.

This tension also raises questions about the sustainability of AI-driven growth. Will tech giants continue to invest at current rates if electricity becomes the limiting factor? Will regions with stronger, more reliable grids become the exclusive homes of advanced AI research and deployment? And how will governments respond if AI’s energy consumption begins to threaten broader economic stability?

Is the end nigh?

The warnings from Musk, Zuckerberg, Jassy, Huang, and Buffett converge on one sobering point: AI’s progress is no longer limited by imagination, innovation, or even capital, it is constrained by the laws of physics. Electricity, once an assumed utility in the tech equation, has become a critical bottleneck, dictating which companies and nations can realistically participate in the AI revolution.

Some industry observers suggest that the next wave of AI investment may shift from models and data centres to energy infrastructure, particularly renewable sources and grid upgrades. Others argue that more efficient algorithms and hardware could partially mitigate the problem, though these solutions are unlikely to keep pace with exponential growth.

Regardless of the approach, one truth is unavoidable: the AI era is colliding with the real-world limits of energy. The sector’s leaders are now forced to confront a question that transcends software, chips, or venture capital: can human ambition outgrow the physical world?

The AI boom has captivated investors, technologists, and the public alike. Yet as Silicon Valley races ahead, the infrastructure needed to power this revolution lags behind. Massive server farms, exponential growth in computing power, and energy-hungry AI models are creating a perfect storm.

RELATED STORIES

Big tech’s AI spending spree: Boom, bubble, or bold bet?

AI data centre boom strains power grids and cooling supply chains: report