The next phase of the artificial intelligence investment cycle is likely to be driven less by semiconductor makers and hyperscalers alone and increasingly by companies using AI to improve productivity, margins and capital efficiency, according to Franklin Templeton.
Grant Bowers, Portfolio Manager at Franklin Templeton, said the recent rise in US interest rates and debate over the pace of frontier AI development have created the first meaningful stress test for the AI trade in years.
However, he argued that the underlying infrastructure investment cycle remains intact because the industry’s constraints are increasingly physical rather than purely financial.
“Compute was AI’s first bottleneck. Power, cooling, transmission and connectivity appear to be next,” Bowers said.
Unlike semiconductor capacity, these constraints cannot be resolved quickly. Power generation and transmission can take years to permit and build, while cooling and connectivity requirements increase alongside new data centre development.
Bowers said higher interest rates could compress valuations across the AI infrastructure supply chain, but would not solve bottlenecks in electricity supply, transmission networks or permitting.
He added that many of the largest hyperscalers are funding investment from substantial operating cash flow, making them less sensitive to higher borrowing costs than more leveraged businesses.
AI Productivity Gains Could Drive 2027 Earnings Surprises
Franklin Templeton also expects AI-related investment opportunities to broaden beyond companies traditionally regarded as part of the technology sector.
Industrials, financial services, healthcare and consumer companies are already using AI to improve pricing, scheduling, inventory management and customer service, according to the report.
Bowers said companies able to expand without increasing labour, overhead or working capital at historical rates could see their earnings profiles improve even without stronger revenue growth.
Higher capital costs may even accelerate adoption by forcing businesses to place greater emphasis on productivity and cost control.
Franklin Templeton believes these benefits may not yet be fully reflected in earnings expectations, creating scope for positive earnings surprises in 2027 as AI-driven efficiency gains become more visible in corporate results.
Scarce Assets Gain Value As AI Becomes More Accessible
The firm also expects the value of proprietary data, trusted distribution networks and deeply embedded workflows to increase as AI capabilities become more widely available.
Bowers argued that while AI can process information and improve business processes, it cannot easily replicate decades of proprietary data, customer trust, regulatory standing or established distribution networks.
“As intelligence becomes cheaper, scarcity becomes more valuable,” he said.
Higher interest rates reinforce this dynamic by making investors less willing to pay for distant earnings and placing greater emphasis on companies with established competitive advantages and strong cash generation.
Franklin Templeton therefore remains constructive on US equities despite elevated valuations and market concentration in some areas.
Bowers said the first phase of the AI trade rewarded the companies building the infrastructure required to make AI possible, particularly chipmakers, hyperscalers and data centre-related businesses.
The next phase, however, could favour a broader group of companies that successfully put AI to work within their existing operations.
That shift, he said, expands the AI investment opportunity well beyond semiconductors and cloud computing and towards companies capable of converting AI adoption into tangible earnings growth.
