Artificial Intelligence Boom May Be Here Stay

The fear of an artificial intelligence bubble may be exaggerated, but a quieter change in the software sector is already happening—and it holds the most immediate importance.

Valuations don’t resemble the dot-com era

Comparisons to the late 1990s tech bubble overlook key differences. In 1999, leading companies traded at 60-70x forward price-to-earnings multiples. Today, many AI-related stocks trade at 20-25x, sometimes lower than established consumer brands.

Adoption has also taken a different shape. During the dot-com boom, spending outpaced actual demand. AI adoption, however, has already reached widespread use. Most businesses now integrate AI internally to improve efficiency, reversing the pattern seen in 1999.

Funding models have evolved as well. Hyperscalers like Microsoft, Google, and Amazon finance AI expansion through operating cash flow rather than speculative IPOs. Private credit fills the remaining gaps, creating a more stable foundation than the late-90s frenzy.

Demand is expanding

By the end of Q1 2026, the combined backlog of contracted cloud commitments across major hyperscalers will surpass $2 trillion. Some providers have stated they cannot meet current demand.

Capital expenditures are rising sharply. Hyperscaler spending is projected to exceed $700 billion this year and near $1 trillion by 2027. About half of next year’s spending will target semiconductors, shifting profits toward AI computing at an unprecedented rate.

The surge in demand has exposed supply-chain constraints, creating new investment opportunities. CPUs, memory, and power generation face significant pressure. Natural gas turbine manufacturers are fully booked through the decade, creating room for innovative solutions, including off-grid.

The software industry faces new challenges

Companies are redirecting software budgets toward agentic AI tools that work across their current stacks. A large enterprise is more likely to invest in AI agents this year than evaluate a new customer relationship management system. Meanwhile, AI models can replicate years of engineering work in far less time, weakening what were once strong competitive advantages.

Most software companies haven’t missed earnings yet, but the market anticipates a shift. With more compelling opportunities in the AI value chain, investors must reconsider long-term valuations.

For office workers, the next software upgrade may not come from a familiar brand. It could arrive as a lightweight AI agent that automates tasks across multiple platforms without requiring a full system overhaul. The tools people use daily won’t vanish, but their function is evolving.

Related: A Comprehensive Overview

Execution, not demand, poses the greatest risks

The main challenges involve whether infrastructure can keep pace. Supply constraints exist at every level—from semiconductors to power generation—but these limitations also drive innovation.

Three risks stand out. The first is the financial health of major AI labs, which are spending capital at record rates. The second is political resistance to rising electricity costs and water usage near data centers. The third is the unit economics of AI tokens, which will determine whether current spending levels remain sustainable.

The most frequent investor mistake is treating AI as a single, overvalued trade. In practice, it represents a multi-year cycle funded by some of the most profitable companies in history, supported by contracted enterprise revenue. The opportunity extends beyond what the market currently recognizes, and the divide between winners and losers is growing.

This isn’t a bubble. It marks a fundamental change, with the software industry experiencing its effects first.