Bank of America has raised questions about the financial sustainability of the current artificial intelligence investment cycle, pointing to a substantial increase in capital expenditure forecasts and declining prices for AI services.
In a Friday research note, investment strategist Sebastian Raedler compared the expansion of AI infrastructure with previous periods of concentrated investment in emerging technologies.
The bank’s analysis focused on the relationship between capital spending, industry competition, service pricing and the ability of AI providers to generate revenue from increasing usage.
Hyperscaler Investment Expectations Have Tripled
Bank of America said consensus forecasts for hyperscaler capital expenditure over the next 12 months have risen from $330 billion in early 2025 to $1 trillion.
The projected spending represents approximately 3% of US GDP, compared with 2.2% during the railroad investment expansion of the 1870s and 1.1% during the technology investment cycle of the 1990s.
Raedler noted that earlier capital cycles often resulted in lower financial returns than investors initially anticipated, despite the continued adoption and development of the underlying technologies.
According to the bank, elevated profit margins during the initial stages of investment cycles can attract additional participants and encourage capacity expansion.
As supply increases and competition develops, companies may experience reduced pricing power and pressure on profitability.
AI Model Competition Differs From Earlier Cloud Computing Expansion
Bank of America contrasted the current AI market with the development of cloud computing during the 2010s.
The bank said the cloud industry benefited from substantial barriers to entry, which restricted the number of competitors and helped companies maintain pricing power.
AI model development has followed a different competitive pattern, with additional providers entering the market, including Chinese companies offering open-weight models.
Raedler reported that average token prices have fallen by around 50% since the summer.
The decline has occurred despite continued increases in the amount of AI processing activity, creating a distinction between rising service consumption and the revenue generated from that activity.
Lower Token Prices Have Yet to Produce Comparable Spending Growth
One argument supporting continued AI investment is that declining service costs could encourage wider adoption and greater overall consumption.
This interpretation draws on Jevons Paradox, which suggests that improvements in efficiency can increase total demand for a resource or service.
Bank of America acknowledged this argument but pointed to spending data that present a more mixed picture.
Figures from Ramp AI indicate that US business expenditure on artificial intelligence has edged lower since August, even as token volumes have continued to increase substantially.
The bank identified this difference as a potential challenge for companies seeking to convert rising AI usage into corresponding revenue growth.
Raedler said the combination of increased competition and pressure on monetisation resembles developments observed during earlier capital investment cycles.
Technology-Related Equity Indices Decline
The research note also highlighted recent movements in markets exposed to semiconductor and technology investment.
South Korea’s KOSPI has declined 20% since early October, while the Philadelphia Semiconductor Index is trading 13% below its recent peak, according to Bank of America.
The bank cited these declines alongside its assessment of the AI investment cycle and the potential financial effects of increasing industry capacity.
BofA Forecasts 10% Decline in Stoxx 600
Bank of America maintained its negative assessment of European equity markets, forecasting that the Stoxx 600 will fall 10% to 580 by the second quarter.
The bank also retained its preference for defensive sectors over cyclical industries.
Its positioning remains overweight consumer staples and pharmaceuticals, while banks, capital goods and semiconductor companies remain underweight.
The bank’s analysis suggests that the financial outcome of the AI investment cycle will depend partly on whether growing demand can offset declining service prices and the effects of additional competition.
For now, Bank of America continues to identify the scale of planned capital expenditure and the divergence between AI usage and spending growth as factors in its market outlook.
Leave a Reply