The artificial intelligence sector may need to produce close to $6 trillion in yearly revenue by 2031 to support the massive infrastructure buildout now underway across chips, data centers, networks and power systems, according to research from Bain & Company. The consultancy's analysis points to a widening gap between the capital being committed to AI capacity and the revenue that AI applications currently generate.
Bain estimates that Microsoft, Google, Amazon, Meta and Oracle could together spend as much as $780 billion on capital expenditure in 2026, roughly five times their combined level three years earlier. Annual spending on AI infrastructure could reach $1.5 trillion by 2031, covering new data center construction and computing capacity as well as upgrades to installed GPUs, memory and networking equipment.
The scale of the buildout is already visible in the size of leading AI data centers. Facilities approaching 1 gigawatt of power capacity today could reach about 2 GW by 2027, while 9 GW campuses could emerge by the end of the decade, Bain said.
Revenue gap of $4.2 trillion
Bain's calculation assumes capital expenditure will represent roughly 25% of industry revenue, a level the consultancy considers ambitious but reasonable based on trends among cloud providers. At that level, sustaining $1.5 trillion in annual infrastructure spending would require an AI market approaching $6 trillion in annual revenue.
Some of that revenue is already taking shape. Consumer AI products, through subscriptions and advertising, could generate between $200 billion and $400 billion annually by 2031, Bain estimates. Enterprise adoption could contribute another $1 trillion to $1.4 trillion as companies deploy the technology across software development, sales, marketing, customer service and IT operations. Together, consumer and enterprise AI could represent between $1.2 trillion and $1.8 trillion in annual revenue.
Even at the upper end of that range, the industry would still face a roughly $4.2 trillion shortfall relative to the nearly $6 trillion market Bain estimates would be required to fund the infrastructure investment. "That revenue must come from new sources of economic value," Bain said, arguing that productivity improvements in existing business processes cannot be the sole economic justification for the AI infrastructure boom.
Search, autonomy and physical AI
Bain identifies several areas that could help close the gap. AI model providers could generate between $100 billion and $200 billion or more by incorporating advertising into chatbot products as consumers increasingly use AI tools in place of traditional internet search. Another potential $400 billion opportunity could come from autonomous vehicles and industrial automation, spanning consumer vehicles, robotaxis and automated logistics.
The consultancy sees an even larger opportunity in what it calls physical AI. Advanced models can create realistic simulations and digital twins of physical processes, allowing companies to test changes, improve productivity and accelerate the deployment of autonomous systems. AI-powered robots, including humanoid machines, could operate in less structured environments, opening applications across manufacturing and other parts of the physical economy. Bain estimates the physical economy could represent a $900 billion opportunity across sectors including automotive, electronics, semiconductors and aerospace and defense, assuming a 10% reduction in research, development and manufacturing costs from higher yields and faster factory ramps.
The fourth category is less defined but potentially broader: entirely new products and services enabled by increasingly capable AI. Bain points to AI-driven drug discovery as one example, where faster and cheaper research could make treatments for rare diseases economically viable. Other opportunities include always-available mental health support, materials science breakthroughs that could enable new batteries and semiconductors, and autonomous scientific research that accelerates progress in areas ranging from neuroscience to fusion energy.
These applications would represent a shift away from using AI primarily to perform existing tasks more efficiently toward creating economic activity that does not currently exist. For investors, the analysis frames the central question of the AI cycle: whether the revenue eventually generated by the technology can justify the capital now being poured into it.
This article is for informational purposes only and does not constitute financial advice.
