Gartner Forecasts AI Spending Surge as Bubble Debate Deepens

Key Takeaways

• Worldwide AI spending is forecast to reach $2.7 trillion in 2026, even as enterprise-wide profit gains remain modest.
• Venture funding is increasingly concentrated among OpenAI, Scale AI, Anthropic, Project Prometheus, and xAI.
• The personal computer boom may offer a more useful comparison than the dot-com crash because infrastructure demand can outpace near-term application returns.

The artificial intelligence market has reached an awkward stage. Capital spending, venture investment, and product deployment are accelerating, while evidence of broad financial returns remains patchy. That gap is giving prominent investors plenty of material for bubble warnings. It does not, by itself, establish that the entire market is detached from economic reality.

Gartner forecasts worldwide AI spending will reach $2.7 trillion in 2026, representing a 49.5% year-over-year growth. Infrastructure investment and the inclusion of AI features in software and services are the main drivers. Those categories matter because they capture more than corporate purchases of standalone generative AI applications. Spending also includes the computing capacity, data systems, cloud services, and product upgrades supporting adoption.

Markets can overfund a technology that ultimately becomes foundational.

That is where the growth of personal computer sales may provide a better comparison than a simple replay of the dot-com era. Early PC demand created durable markets for processors, operating systems, storage, business applications, and technical services. Not every supplier survived, and valuations did not move in a straight line. Yet the underlying computing shift continued even when particular companies stumbled.

AI could follow a similarly uneven path. Nvidia benefits from the infrastructure buildout as customers purchase computing capacity, regardless of whether every application provider produces an attractive return. OpenAI and Anthropic, meanwhile, occupy a different part of the market, where model costs, pricing, customer retention, and product differentiation can have greater influence on long-term economics.

Venture capital concentration adds another layer of risk. Crunchbase reported that AI-related startups attracted $211 billion during 2025, an 85% increase from 2024 and roughly half of global venture funding. OpenAI, Scale AI, Anthropic, Project Prometheus, and xAI raised a combined $84 billion, equivalent to 20% of worldwide venture funding.

Such concentration can produce striking valuations and crowded expectations. It can also reflect the extraordinary capital requirements involved in training models, securing computing resources, hiring specialized talent, and serving customers at scale. The uncomfortable question is whether those requirements form a durable barrier to entry or merely an expensive race with uncertain margins.

Enterprise results remain the main counterweight to the investment enthusiasm. According to McKinsey data from 2025, 64% of organizations said AI was enabling innovation, but only 39% reported an enterprise-level EBIT impact. Most organizations in that smaller group attributed less than 5% of EBIT to AI. Experimentation is widespread; material profit contribution is not.

Still, adoption often moves through departments before it appears clearly in consolidated financial statements. An assistant that reduces time spent drafting documents may deliver a local productivity gain without changing revenue or headcount. A customer-service system might improve response times while creating new integration and oversight costs. These are real benefits, but they can be difficult to translate into enterprise-wide EBIT.

That said, weak measurement should not become an excuse for indefinite spending. Business leaders can separate infrastructure enthusiasm from operating value by assigning owners, establishing baselines, tracking total costs, and reviewing whether pilots advance into production. Governance also belongs in that calculation because unreliable outputs, data exposure, and poorly supervised automation can erase apparent savings.

The longer-range case remains substantial. IDC projects $22.5 trillion in cumulative economic value from AI between 2025 and 2031. It also identifies inflation and uncertainty surrounding the timing of enterprise "second-wave" spending as downside risks. In other words, a large destination does not imply a smooth journey.

For corporate buyers and investors, "bubble or no bubble" may be too blunt a choice. Infrastructure demand, application economics, venture valuations, and enterprise productivity can move in different directions at the same time. The PC analogy suggests that a technology cycle can produce lasting change alongside failed vendors, excessive funding, and painful corrections. AI may be no different.