AI Spending Surges: Big Tech Bets Hundreds of Billions on a Market Still Finding Its ProfitsAI Spending Surges: Big Tech Bets Hundreds of Billions on a Market Still Finding Its Profits

Gartner forecasts global AI spending will reach KSh351 trillion ($2.7 trillion) in 2026, up 49.5% from 2025. That is a forecast of spending across the industry, not a measure of profits or the amount of new revenue already generated, as set out in Gartner’s September forecast.

The investment wave is visible in cloud earnings, chip demand and data-centre construction. Yet the business case is uneven: companies supplying the infrastructure are collecting large orders now, while many customers are still testing whether AI tools produce enough savings or additional sales to justify their costs.

For Kenya and other African markets, this creates both an opening for productivity gains and a risk of paying for technology without capturing much of its value.

Big Tech is building first and measuring returns as it goes

The latest quarterly results from Microsoft illustrate the current balance. In results reported on 28 October, the company said revenue rose 20% year on year to KSh12.0 trillion ($92.28 billion), while its Azure cloud business grew 48% to KSh4.31 trillion ($33.17 billion).

Capital expenditure reached KSh5.88 trillion ($45.2 billion) for the quarter, as the company continued expanding the computing capacity used for AI and other cloud workloads.

The results suggest that demand is translating into meaningful cloud revenue, although the company has not disclosed that all of Azure’s growth comes from AI alone. See Microsoft’s quarterly results.

The wider spending cycle is larger still. Bain & Company estimates that capital spending by five major technology companies, including Microsoft, Alphabet, Amazon, Meta and Oracle, could reach about KSh101.4 trillion ($780 billion) in 2026. This is an estimate for a group of large companies, not a complete tally of every AI investment worldwide. Much of the money is going towards data centres, advanced chips, networking equipment, electricity supply and cooling systems.

The market is not just about the companies selling AI chatbots. It includes chip designers and manufacturers, memory suppliers, cloud providers, data-centre operators, electricity companies, software developers and firms integrating AI into existing products.

McKinsey’s September analysis raised its base-case projection for the global semiconductor market to KSh299 trillion ($2.3 trillion) by 2030, citing higher prices for advanced chips and memory as AI data-centre demand grows. That is a forecast, not a guaranteed outcome, and depends on continued investment and demand.

The money is flowing unevenly through the AI industry

The clearest near-term beneficiaries are often the businesses selling scarce inputs: advanced processors, high-bandwidth memory, cloud computing and data-centre capacity. When demand grows faster than supply, these suppliers can command higher prices. But their advantage depends on customers continuing to spend and on new capacity not arriving faster than demand can absorb it.

AI model developers and software companies face a different test. They must convert technical performance into products people will pay for, while covering computing costs, product development, distribution and customer support. A tool may attract millions of users and still struggle to make a profit if each task is expensive to run or users are unwilling to pay enough.

Bain’s October technology report estimated that the AI industry could need around KSh780 trillion ($6 trillion) in annual revenue by 2031 to support the scale of infrastructure investment being planned.

This is a calculation of the revenue the industry might need under Bain’s assumptions, not a prediction that it will certainly reach that level. The estimate highlights the gap between money being committed to build capacity and the revenue that must eventually support it. Read the report summary from Consulting.us.

The distinction matters to investors. Rising demand for chips and cloud services does not automatically mean every AI company is a good investment. Valuations can assume years of growth before those earnings arrive.

If customer adoption slows, computing prices fall, or infrastructure is built faster than it can be used, some businesses could face weaker margins or write-downs. If adoption spreads into more everyday business processes, current investment could instead create new sources of recurring revenue.

The real test is whether customers get measurable value

Large companies are increasingly embedding AI into products they already sell, including office software, customer service systems, advertising tools and developer platforms. This may be a more dependable path to revenue than relying only on standalone chatbots, because existing suppliers already have customers, distribution and subscription relationships.

Even so, businesses need to look beyond demonstrations. Useful measures include the time saved on a specific task, the error rate, the cost per completed job, customer satisfaction and whether staff can use the tool safely. An AI assistant that drafts a report faster may have value, but the gain can disappear if employees spend the saved time checking unreliable output or correcting mistakes.

For smaller firms, the sensible approach is to start with a defined problem rather than buy AI tools simply because competitors are discussing them. A retailer could test whether automated support reduces unanswered customer enquiries.

An accounting firm could measure whether document extraction cuts processing time without increasing errors. A media company could test transcription or research support while keeping editorial verification with human journalists.

The result should be measured against the old process, including subscription, integration, training and review costs.

What the AI investment race means for Africa

African companies are unlikely to capture the largest share of the hardware spending directly. Advanced chips and much of the cloud infrastructure are produced or controlled by companies based outside the continent. But local businesses can still benefit if AI helps them deliver services at lower cost, reach more customers, translate information into local languages or make better use of limited staff.

The barriers are practical. Reliable electricity, broadband quality, foreign-currency costs, data protection, access to suitable training data and a shortage of specialised skills can determine whether an AI service is affordable and useful. Dependence on overseas platforms also means local firms may be exposed to price changes, service restrictions and rules set elsewhere.

There are opportunities in areas such as financial services, agriculture, logistics, education, healthcare administration and customer support, but each application needs evidence of accuracy and economic benefit. A system that works well in one language or country may perform poorly in another. Sensitive decisions involving credit, hiring or health need stronger oversight than low-risk tasks such as summarising public documents.

For Kenyan entrepreneurs and investors, the market offers a choice between building AI products, supplying the businesses that adopt them, or using existing tools to improve productivity. The best opportunity will not necessarily be the company with the most advanced model. It may be the business that solves a specific local problem reliably, at a price customers can afford, and can show that the result is better than the alternative.

The next phase of the AI market will be judged less by spending announcements alone and more by revenue quality, customer retention, operating costs and demonstrated productivity. Infrastructure demand is strong, but investors should distinguish money being spent today from profits that companies still need to earn tomorrow.

For a related look at how technology competition is reshaping a Kenyan market, read Mwenendo’s analysis of Kenya’s home internet battle between Safaricom and Starlink.