London: The rapid expansion of artificial intelligence infrastructure is creating a major test for the technology industry as companies commit enormous sums to data centres, computing systems and chips while the economic returns from those investments remain uncertain.
A PwC study projects $31.6 trillion in global data centre capital expenditure through 2050 across 46 countries and territories. The figure is a long term projection, rather than money that has already been spent or committed.
PwC estimates that annual data centre capital spending could increase from about $800 billion in 2026 to $1.8 trillion by 2050. The investment is being driven largely by the rapid growth of artificial intelligence and the computing power needed to operate increasingly advanced systems.
The scale of the spending has raised questions about whether AI companies will be able to generate enough revenue and productivity gains to support the investment.
Bain and Company estimates that the AI industry would need about $6 trillion in annual revenue by 2031 to support the level of infrastructure investment examined in its analysis. Bain estimates that existing consumer and business AI applications could generate between $1.2 trillion and $1.8 trillion, meaning about $4.2 trillion in additional annual revenue could need to come from new markets and applications.
Those new areas could include autonomous vehicles, robotics, industrial systems, drug discovery, energy and other uses that are still developing.
Anthropic provides one example of the financial scale involved. The artificial intelligence company reported about $4.6 billion in revenue in 2025, according to its IPO filing. Its revenue increased about 12 times during the year.
Anthropic also reported a net loss of about $42 billion, although about $34 billion of that amount came from an accounting charge connected with financing instruments that could eventually be converted into shares. Its operating loss was about $8.06 billion.
The company spent about $7.33 billion on computing and infrastructure in 2025 and had about $20.28 billion in cash, cash equivalents and short term investments at the end of that year.
The company has also disclosed about $518 billion in future cloud, computing and infrastructure commitments. The figure represents future commitments and should not be interpreted as money that Anthropic has already spent.
The financing of AI infrastructure is itself becoming an important part of the story. On October 1, Broadcom agreed to lend Anthropic up to $42 billion to help finance infrastructure spending. The facility could cover about one third of Anthropic's $125.2 billion five year commitment for tensor processing unit computing capacity.
The arrangement shows how closely linked AI companies, chip suppliers, cloud providers and financing are becoming as the industry expands.
Another major question is whether the investment will produce enough productivity growth across the wider economy.
JPMorgan has argued that substantial productivity growth would be needed to justify some of the valuations associated with major AI companies. However, broad economy wide productivity gains from AI remain difficult to measure.
Research from Stanford's Digital Economy Lab has found signs that AI exposure is already affecting parts of the labour market. Its study found that employment among US workers aged 22 to 25 in highly AI exposed occupations was about 19 percent below the level it would have reached if it had continued to grow at the pace of less exposed occupations.
The study did not find evidence of widespread economy wide job displacement. The effect appeared mainly through reduced hiring of younger workers rather than large increases in job losses.
The physical demands of the AI boom are also becoming clearer. The International Energy Agency estimates that global data centre electricity consumption could almost double from about 485 terawatt hours in 2025 to around 950 terawatt hours in 2030. Electricity use by AI focused data centres is expected to grow even faster.
The expansion is already encountering bottlenecks involving electricity supplies, grid connections, transformers, advanced chips and other equipment needed to build data centres.
These pressures mean that the future of AI investment will depend not only on technological progress but also on the ability of companies to secure financing, electricity and infrastructure while generating enough revenue from their systems.
A financial correction in the AI industry would not necessarily mean that the infrastructure being built today would lose its long term value. Earlier technology and infrastructure booms, including the railway expansion of the 19th century and the internet investment boom, suffered major financial corrections while leaving infrastructure that continued to have economic value.
The central question for the AI industry is therefore becoming increasingly clear. Companies are investing enormous sums today in the expectation that AI will create much larger markets and productivity gains in the future. Whether those gains arrive quickly enough to justify the scale of investment remains uncertain.