What If the AI Boom Slows? The Numbers Behind an AI-Dependent Global Economy

Global trade is doing better than one might think, despite wars and trade tensions.

Part of this resilience rests, however, on a very particular engine: massive investments in servers, chips, and data centers necessary for artificial intelligence.

If these expenditures were to slow, the repercussions could extend far beyond Silicon Valley.

The AI boom no longer only affects software. It mobilizes chip manufacturers, construction companies, electricity providers, cabling groups, and carriers. With every new data center, billions of dollars cross borders before a single user asks a chatbot a question.

This dynamic explains why some economists today question the robustness of global growth. The issue is not to predict an imminent collapse of AI, but to measure what would happen if the pace of investments stopped accelerating.

42% of the rise in world merchandise trade linked to AI technologies

In its trade and development report published in October 2026, the United Nations Conference on Trade and Development (UNCTAD) estimates that AI-related technology products, including semiconductors, servers and data-center equipment, accounted for about one-sixth of the value of global merchandise trade in 2025.

More striking still, this same category of products contributed to 42% of the growth in the value of global merchandise trade in 2025. This does not mean that AI produces 42% of global wealth, nor that 42% of jobs depend on chatbots. It is a contribution to the rise in goods trade in a given year, in a statistical category that also includes computer hardware used for other purposes.

This distinction matters: a category can explain a large portion of annual variation while representing only a fraction of total trade. Global trade remains far broader, with energy, agricultural products, automobiles, textiles, medicines, or industrial equipment.

Asia at the heart of the movement: the 80% figure does not apply to the whole world

In a note dated November 9, 2026, the economic research firm Capital Economics estimates that AI-related goods generated about 80% of the recent rise in Asian exports. The expansion of orders for chips and servers has particularly supported certain producers on the continent.

This percentage thus applies to Asia and to a recent period studied by the firm. It should not be presented as “80% of world exports” nor as “80% of global GDP growth.” They are three entirely different notions.

The manufacturing chain explains this concentration. Advanced chips are designed by a small number of companies, manufactured mainly in certain cutting-edge factories in Asia, paired with specialized memories and then integrated into servers assembled in several places around the world. Countries that do not produce these devices can therefore participate far less in the current rise in exports.

Nearly $202 billion invested in three months by four groups

The scale of the movement also appears in the accounts of the American giants. In the third calendar quarter of 2026, investment and equipment purchases reported by Amazon, Alphabet, Microsoft, and Meta total about $201.9 billion, up from $97.3 billion in Q3 2025. The increase is roughly 108% year over year.

These amounts correspond to total investment expenditures drawn from the companies’ financial documents: they include data centers and AI-related equipment, but also other infrastructure. It would be inaccurate to claim that the $201.9 billion was spent exclusively on AI. The individually rounded values also explain the small discrepancy with the total derived from the non-rounded amounts.

To understand the scale, these four groups allocated in three months more cash to their equipment than they had spent in the same quarter a year earlier, with a gap of about $104.6 billion. That money becomes orders for their suppliers, but also fixed costs and assets that will need to be amortized.

Why a drop in orders would have cascading effects

Imagine a major IT services provider delaying the construction of several data centers. It orders fewer processors and less memory. Manufacturers scale back certain investments. Companies responsible for electricity, cooling, cables, and buildings revise their schedules. Part of export growth and industrial activity slows.

This is not necessarily a broad crisis. Investments could stabilize at a high level and AI use could continue to advance. But when an economy is used to orders rising rapidly, a simple deceleration can be enough to slow its growth rate.

Capital Economics expects the rise in IT equipment to continue supporting exports into early 2027, but foresees a shift in trend later in the year. This is the firm’s scenario, not a guaranteed slowdown timetable.

Financial markets can amplify the movement

The risk concerns not only factories. Investors price certain companies based on expectations of strong AI-related revenues. If these revenues grow less fast than hoped, their stock prices can correct. A prolonged downturn in valuations could raise the cost of financing and weigh on highly indebted firms building such infrastructure.

UNCTAD warns that a significant stock-market retreat could trigger asset sales and weigh on credit beyond the tech sector. But one should not confuse a valuation correction, a decrease in investment, and a global recession: the mechanisms are different, and none automatically follows from the other.

Who really benefits from server manufacturing?

The surge in equipment sales is not distributed evenly. In the study of an advanced server system such as Nvidia GB200 NVL72, UNCTAD notes that a large portion of the value captured accrues to a very small number of players, notably to the firms that hold the most sought-after technologies and intellectual property.

The challenge for emerging economies is therefore twofold: participate in the industrial chain and capture enough value in the form of skills, qualified subcontracting, research, or services. Assembling equipment does not yield the same margins as designing its most advanced components.

What possible effects for Tunisia?

Tunisia is not a major producer of processors for global data centers. It can nevertheless be affected in several ways: by European demand for equipment and electronic components, by IT services exports, by training needs, and by foreign partner investment decisions.

A slowdown in global demand could weaken certain international orders, but a drop in equipment prices could also facilitate some local projects. There is no precise measure in the data reviewed that would attribute a definite share of Tunisian growth to AI. Any national numerical forecast would be premature.

Is AI a bubble? The real question lies elsewhere

The International Monetary Fund projects global growth of 3.1% in 2026 and 3.4% in 2027, in its October edition of the World Economic Outlook. These are projections for the whole economy, not a scenario calculating the effects of a halt in AI-related investments.

A sector can experience financial excesses while producing durable innovations. The central question is therefore how quickly firms will translate their investments into productivity gains and measurable revenues. The next phase of the AI boom could depend less on the number of servers purchased than on what they will actually enable to produce.