¿Es la IA un motor de crecimiento sostenido, o una situación que podría convertirse en una “burbuja”?

The artificial intelligence (AI) boom of the 2020s has sparked intense debate among investors, economists, and technologists. Is this a transformative growth engine with long-term economic potential, or a speculative bubble primed for a correction? To answer this, we must dissect the structural differences between the current AI surge and historical financial bubbles, such as the dot-com crash of 2000 and the 2008 housing crisis. By analyzing economic fundamentals, investment patterns, and sector-specific risks, this article evaluates whether AI represents a durable inflection point or a precarious overbuild.

Structural Similarities and Differences with Historical Bubbles

The AI boom shares familiar hallmarks with past speculative episodes: rapid capital inflows, soaring valuations, and a concentration of market performance among a handful of dominant firms. For instance, NVIDIA, Microsoft, and Meta have seen their valuations soar on the back of AI-driven growth, with many AI-related firms trading at price-to-sales multiples exceeding 20x. This mirrors the dot-com era, where companies with no revenue commanded stratospheric valuations. However, a critical divergence exists: today's leading AI firms are not merely speculative plays. Unlike the dot-com companies of the late 1990s, which often lacked revenue models, modern AI leaders generate tangible returns. Microsoft's Azure and NVIDIA's data-center chips, for example, are already embedded in core business operations, providing recurring revenue streams.

Infrastructure investment is another key similarity. Global capital expenditure on high-performance computing and data centers is projected to reach $6.7 trillion by 2030. While this could drive productivity gains, it also carries risks of overbuilding. The dot-com bubble saw unused fiber-optic networks become white elephants; a similar fate could await underutilized AI infrastructure if productivity gains fail to materialize. Howard Marks of Oaktree Capital has warned of a "capex winter" in the AI sector, where overleveraged firms could face dislocations if demand slows.

The Role of Cash-Rich Hyperscalers and Circular Investment Risks

The current AI ecosystem is dominated by cash-rich hyperscalers-Amazon, Alphabet, Meta, and Microsoft-which are pouring unprecedented sums into AI infrastructure. In 2025, their combined capital expenditures reached $300 billion, with projections of $490 billion in 2026. These firms, with their strong free cash flow, are funding a self-reinforcing cycle: NVIDIA's chips power data centers, which in turn host AI models developed by companies like OpenAI, which then rely on NVIDIA's hardware. This circularity creates a feedback loop that resembles the speculative dynamics of the dot-com era, but with a crucial difference: today's investments are backed by balance sheets with ample liquidity.

However, this circularity also raises concerns. For example, NVIDIA's $100 billion commitment to OpenAI, which in turn invests in data centers powered by NVIDIA's chips, creates a closed-loop ecosystem that could obscure financial risks. While this model drives innovation, it also risks inflating valuations without generating broad-based economic returns. As of 2025, the top four hyperscalers are projected to generate only $30–50 billion in AI-related revenue despite spending over $300 billion in capex. This spending-to-revenue gap highlights the need for caution.

Infrastructure ROI and Energy Constraints

The economic viability of AI hinges on infrastructure ROI. A Bain & Company analysis estimates that the AI sector must generate $2 trillion in new annual revenue by 2030 to justify current investments. Yet, enterprise adoption remains uneven. While average AI investment rose by 14% in 2025, with enterprises spending $130 million on average, challenges persist. Gartner warns that organizations often miscalculate AI costs by 500% to 1,000%, with proof-of-concept phases alone costing $2.9 million for large enterprises.

Energy constraints further complicate the equation. Data centers are projected to consume 12% of U.S. electricity by 2028, straining power grids and driving up costs. The U.S. Department of Energy has flagged energy availability as a critical bottleneck, with AI's computational demands reshaping national energy strategies. For example, Virginia's data centers already consume 25% of the state's electricity, forcing utilities to rethink efficiency models. This energy-intensive model raises questions about sustainability, particularly as AI's environmental footprint grows.

Normalization Phase: Winners, Survivors, and Casualties

As the AI boom enters a normalization phase, the focus shifts from hype to execution. Winners will be companies that successfully monetize AI while managing infrastructure costs. Renewable energy firms, for instance, are gaining momentum as AI-driven efficiency measures optimize grid performance and reduce waste. Solar PV capacity is expected to account for 80% of global renewable growth over the next five years, with AI playing a pivotal role in this transition.

Survivors will be those that adapt to regulatory and energy constraints. The U.S. Department of Energy has highlighted gaps in AI governance and cybersecurity, urging policymakers to develop frameworks that balance innovation with oversight. Companies that prioritize data governance, cross-functional collaboration, and regulatory foresight-such as Microsoft and NVIDIA-are better positioned to navigate this landscape.

Casualties, meanwhile, will include traditional energy producers and speculative startups. Oil and gas companies struggled in 2025 as AI and renewables outpaced their growth, with energy stocks returning only 9% compared to the S&P 500's 16.4%. Similarly, venture capital monoculture-where investors pour billions into unproven AI startups- risks a correction. In 2025, $192.7 billion in VC funding flowed into AI startups, many of which lack revenue. A targeted correction in this segment is likely, particularly among hardware vendors and data-center developers.

Actionable Investment Perspective

For investors, the AI landscape demands a nuanced approach. While the sector's long-term potential is undeniable, near-term risks require careful navigation. Here are key takeaways:

1. Prioritize Infrastructure with Tangible ROI: Focus on hyperscalers and renewable energy firms with clear monetization strategies. Avoid speculative plays lacking revenue visibility.
2. Monitor Energy and Regulatory Developments: Energy constraints and policy shifts will shape AI's trajectory. Invest in companies addressing these challenges, such as those optimizing grid efficiency or developing sustainable data-center solutions.
3. Diversify Across the AI Ecosystem: Balance exposure to core AI leaders (e.g., NVIDIA, Microsoft) with emerging innovators in niche applications (e.g., AI-driven healthcare or logistics).
4. Prepare for Sector-Specific Corrections: The secondary ecosystem-hardware vendors, data-center developers-faces higher risks of overbuilding. Allocate capital cautiously in these areas.

Conclusion

The AI boom is neither a classic bubble nor a guaranteed growth engine. It sits in a gray zone: a transformative force with the potential to drive productivity gains, but also structural risks tied to overinvestment, circular dependencies, and energy constraints. For investors, the key lies in distinguishing between durable innovation and speculative excess. By focusing on companies with strong fundamentals, sustainable infrastructure, and regulatory agility, investors can position themselves to capitalize on AI's promise while mitigating its perils.