AI Needs $6 Tn Annual Revenue by 2031 to Justify Compute Spending: Bain Report

The global AI industry will need to generate about $6 trillion in annual revenue by 2031 to support the infrastructure investment required to meet growing compute demand, according to Bain & Company’s seventh annual Global Technology Report.

Bain estimates that existing AI applications across consumer and enterprise markets could generate $1.2 trillion to $1.8 trillion in annual revenue. The remaining $4.2 trillion will need to come from new AI-driven products and markets.

“The debate today is fixated on employee productivity. The economics of AI infrastructure demand trillions in new revenue beyond productivity gains,” David Crawford, Chairman of Bain’s Global Technology practice, said in a statement.

Bain said the additional revenue could come from four areas: AI models replacing search engines and integrating advertising, autonomous systems such as cars, trucks and drones, physical AI including robotics and digital twins, and new applications such as drug discovery, mental health, and energy generation.

“What the industry needs is a wave of innovation that will dwarf what mobile and cloud unlocked,” Crawford said. He added that funding the infrastructure being built ahead of demand would require roughly 1% of additional annual global GDP growth.

AI Demand Lifts Hardware

The rise in AI compute demand has also changed the performance of hardware and semiconductor companies. Their market capitalisation grew at a 24% compound annual rate between 2020 and 2026, compared with 6% for software companies, Bain noted.

High-bandwidth memory, advanced packaging, and custom silicon are among the fastest-growing segments. Application-specific integrated circuits are also scaling as hyperscalers and AI companies develop chips for specific workloads.

Bain said custom chips are taking a larger share of data centre compute as training, inference, and agentic workloads reach volumes that can justify custom designs.

Supply risks are also changing procurement strategies. Companies are securing capacity across multiple locations and suppliers as natural disasters, geopolitical disruptions, and export controls create risks for semiconductor supply.

AI Compresses Cyberattack Timelines

AI is also changing the cybersecurity landscape. Bain estimates that AI has reduced the time required for a typical cyberattack from about four weeks to 18 hours.

The growing use of AI agents adds another layer of risk, while existing approaches to monitoring software and AI vendors struggle to keep pace with frequent changes.

Bain’s CISO survey found that companies are increasing remediation budgets by double-digit percentages. Some are also redirecting 20–25% of cybersecurity staff from other work to address alerts generated by AI-powered scans.

The report identifies legacy platforms, network layers, and software-as-a-service vendors among the areas where companies are focusing remediation efforts.

AI Absorption 

Bain also identifies AI absorption speed, or how quickly companies can put AI into production, as a new competitive factor.

Leading AI labs are investing up to $9.75 billion in forward-deployed engineering to help enterprises integrate AI into their operations. Vendors are also developing application and infrastructure layers that connect AI models with enterprise systems.

Despite falling token prices, Bain does not expect large language models to follow a traditional commoditisation path.

“We see the more likely scenario as a continuum of frontier and mature models,” Crawford said. “New and unproven cases will initially favour frontier models, then likely migrate to lower-cost alternatives as the cases mature.”

Bain expects frontier models to remain focused on tasks where higher intelligence matters, while lower-cost and specialised models handle more established enterprise workloads.

The report also found that technology leaders expect AI to deliver a 148% improvement in release-cycle speed and a 95% increase in software developer productivity over the next one to two years. Current productivity gains range from 20% to 27%.

Bain said the gains from faster coding could create new bottlenecks in software review, coordination, quality, and governance, making changes to the wider software development lifecycle necessary.