Artificial intelligence is no longer just a technology story. It has become a capital markets story, an energy story, an infrastructure story and, increasingly, a national competitiveness story. The question now is whether the world is witnessing a rational build-out of the next general-purpose technology – or whether investors are once again pricing perfection into the future.

The phrase “AI bubble” refers to the concern that the valuations of artificial intelligence-related companies, the scale of capital being deployed into AI infrastructure, and the expectations around future productivity gains may have moved ahead of the actual cash flows being generated today.

Unlike many speculative manias, today’s AI boom is built on real demand, revenues and infrastructure. Nvidia, the company at the heart of the AI hardware cycle, reported full-year fiscal 2026 revenue of USD215.9 billion, up 65 percent, while its data centre revenue reached a record USD193.7 billion, up 68 percent for the year. In its fourth quarter alone, Nvidia reported USD68.1 billion in revenue and USD62.3 billion from data centres. These are not imaginary numbers; they represent one of the fastest corporate revenue expansions in modern technology history.

Yet that is exactly why the bubble debate has become more serious. When a real technology attracts real capital at extreme speed, the risk is not that nothing is happening. The risk is that too much is being priced in too early.

The scale of the AI boom is extraordinary

The numbers behind the AI build-out are now large enough to influence global capital allocation. Stanford University’s 2026 AI Index reported that global corporate AI investment reached USD581.7 billion in 2025, up 130 percent from the previous year. Private AI investment alone reached USD344.7 billion, up 127.5 percent. The United States remained the dominant market, with USD285.9 billion in private AI investment in 2025, which is more than 23 times China’s USD12.4 billion, although Stanford notes that private investment figures may understate China’s total AI spending because of the role of government-guided funds.

Our analysis shows that the capital expenditure story is even more striking . Goldman Sachs estimates that AI infrastructure could require approximately USD7.6 trillion in cumulative capital expenditure between 2026 and 2031 across compute, data centres and power. Its baseline model implies annual AI capex of USD765 billion in 2026, rising to USD1.6 trillion in 2031.

A separate Goldman Sachs analysis estimates that large hyperscalers could spend USD5.3 trillion on AI and data centres from 2025 through 2030. The same report notes that private infrastructure funds had more than USD1.7 trillion in assets under management and USD400 billion of dry powder as of September 2025, highlighting how AI is pulling private capital deeper into digital infrastructure.

This is where the AI story begins to resemble previous technology booms. The internet in the late 1990s was transformative, but many companies built around it failed because the market had capitalised the opportunity too aggressively. Railways, telecom networks, fibre-optic cable and renewable energy have all gone through similar cycles: the technology was real, but the financing cycle became overheated.

Big Tech is spending like AI is an arms race

AI infrastructure is now being treated by the largest technology companies as a strategic necessity. Alphabet said in its June 2026 investor presentation that it expects CAPEX of USD180 billion to USD190 billion this year, around six times its 2022 level and double the previous year, with the “overwhelming majority” going into technical infrastructure.

Meanwhile, Meta reported capital expenditures, including principal payments on finance leases, of USD72.22 billion for full-year 2025. The company also said that 2026 expense growth would be driven largely by infrastructure costs, including third-party cloud spend, higher depreciation and higher infrastructure operating expenses.

Microsoft’s 2025 annual report similarly states that the company will continue investing in capital expenditures to support cloud growth and AI infrastructure and training. The company specifically references data centres, computer systems and AI-related infrastructure as part of its planned uses of capital.

This level of spending is logical if AI becomes the next foundational layer of the global economy. It is risky if monetisation is slower, competition compresses margins, or enterprises fail to generate enough measurable returns from AI adoption.

Adoption is high, but measurable returns remain uneven

The strongest argument against calling AI a simple bubble is – inherently – adoption. McKinsey’s 2025 State of AI survey found that 88 percent of respondents said their organisations were regularly using AI in at least one business function, up from 78 percent a year earlier. And more than two-thirds said their organisations were using AI in more than one function.

However, adoption is not the same as transformation. McKinsey also found that only about one-third of organisations had begun scaling AI across the enterprise, while just 39 percent reported any enterprise-level EBIT impact from AI. Among those reporting EBIT impact, most said the contribution was less than 5 percent of total EBIT.

That gap is central to the AI bubble debate. Businesses are experimenting with AI rapidly, but the market is valuing parts of the AI ecosystem as if widespread productivity gains are already locked in.

MIT NANDA’s 2025 “GenAI Divide” report reached an even sharper conclusion. Based on a review of more than 300 public AI initiatives, interviews with 52 organisations and survey responses from 153 senior leaders, the report said that despite USD30 billion to USD40 billion in enterprise investment into generative AI, 95 percent of organisations were getting zero return, while only 5 percent of integrated AI pilots were extracting millions in value. The report argued that the problem was not primarily model quality or regulation, but poor integration with workflows and day-to-day operations.

Again, this cannot be read into as AI failing. It simply means the easy phase – using chatbots, copilots and AI assistants for individual productivity – may not automatically translate into enterprise-wide profit growth.

For investors, that distinction is critical.

The energy system is becoming part of the AI trade

The AI boom is also moving from software into physical infrastructure – and this trend will continue until the technology plateau. The International Energy Agency (IEA) estimates that data centres consumed around 415 terawatt-hours of electricity in 2024, about 1.5 percent of global electricity consumption. It projects that data centre electricity consumption could double to around 945 TWh by 2030, reaching just under 3 percent of global electricity consumption.

AI-focused accelerated servers are expected to be the fastest-growing component. The IEA projects electricity consumption from accelerated servers, mainly driven by AI adoption, to grow by 30 percent annually in its base case. It also warns that while data centres may still account for a relatively limited share of global electricity demand, they tend to cluster in specific locations, creating local grid stress.

This is why the AI bubble debate also revolves around about power plants, cooling systems, transmission networks, land, water, chips, fibre connectivity and long-term infrastructure financing. If AI demand continues to rise, these assets may become essential. And if demand disappoints, some infrastructure could be underutilised.

So, is it a bubble?

The most optimistic answer is that AI is not a pure bubble, but parts of the AI trade show bubble-like characteristics. The fundamental case is strong. AI is already generating massive revenues for chipmakers, cloud providers and parts of the enterprise software ecosystem.

Consumer adoption has also been unusually fast. Stanford’s 2026 AI Index states that generative AI reached 53 percent population adoption within three years, faster than the PC or the internet, with especially high adoption in some countries such as the UAE.

At the same time, there are clear signs of financial exuberance. The market is depending heavily on a narrow group of companies, capex is rising at unprecedented speed, private capital is being pulled into long-duration infrastructure, and many enterprises have yet to prove meaningful profit impact from AI.

A June 2026 academic paper titled “Boom, Bubble, or Buildout?” concluded that AI is best understood as a real technological revolution with “localized bubble dynamics” rather than either a pure speculative mania or a bubble-free productivity miracle.

That may be the most useful framing: AI is real, but not every AI valuation is realistic.

What could burst the AI bubble?

The first trigger would be a slowdown in hyperscaler spending. If Microsoft, Alphabet, Meta, Amazon or other major cloud players cut capex expectations, the shock would move quickly through semiconductors, data centre developers, power suppliers, cooling companies and AI infrastructure financiers.

The second risk is weak enterprise monetisation. If companies continue using AI but fail to generate measurable productivity, revenue or margin gains, software and cloud spending could be repriced. The MIT and McKinsey findings suggest that the market still needs more evidence of broad-based enterprise returns.

The third risk is depreciation. AI chips can become economically outdated faster than traditional infrastructure. Goldman Sachs notes that the useful life of AI silicon is one of the most important variables in determining the scale of cumulative AI infrastructure investment, because chips have much shorter replacement cycles than buildings or power infrastructure.

The fourth risk is energy. If electricity, grid connections, cooling systems and permitting become bottlenecks, AI infrastructure may become more expensive and slower to deploy. That would affect returns, especially for companies assuming rapid capacity expansion.

The fifth risk is competition. Today’s AI leaders benefit from scarcity… scarcity of chips, models, talent and compute. But if more competitors enter the market, prices for AI services may fall while capital costs remain high.

Why this matters for Oman and the larger GCC

For Oman and the wider GCC, the AI bubble debate should not be viewed as a warning against AI investment. It should be viewed as a warning against undisciplined AI investment.

Oman is already positioning digital transformation as part of its economic diversification agenda. The Ministry of Transport, Communications and Information Technology has stated that Oman’s digital economy contribution reached approximately OMR800 million in 2023, while the next phase of the National Digital Economy Program will focus on digitising promising economic sectors and exporting digital economy services.

Oman’s AI and Digital Future Program includes initiatives such as a National Open Data Platform, a National Centre for AI Research and Development, an AI Studio, an Omani language model, and support for AI-focused startups and new AI data centres. The programme aims to raise the digital economy’s contribution to GDP to 10 percent by 2040, compared with 2 percent in 2021.

Across the GCC, the AI infrastructure race is accelerating. Reuters reported that the UAE’s Stargate AI data centre project is expected to begin operations in 2026, with the broader Abu Dhabi site eventually planned to host 5 gigawatts of data centre capacity. The first 200 megawatts are expected to go live in 2026, using advanced Nvidia systems.

Saudi Arabia-backed AI company Humain is also seeking financing for data centres and GPU chips, with plans for 2 gigawatts of capacity, around one-third of its target by 2034. Reuters reported that the financing package could be worth at least 20 billion riyals, or USD5.33 billion.

These developments show that AI infrastructure is becoming part of the region’s diversification strategy. But the lesson from the bubble debate is clear: countries and companies should prioritise use cases with measurable economic value. In Sultanate of Oman, that means AI should be directed toward logistics, energy efficiency, banking, public services, manufacturing, tourism, food security, healthcare and Arabic/Omani-language digital services – areas where productivity gains can be measured and linked to national development priorities.

Disclaimer: This article is for general information and analysis only. It does not constitute investment advice, financial recommendation or endorsement of any stock, company or asset class.