AI Financing Boom: How to Separate Durable Bank Earnings From Market Hype

A rush of investment into artificial intelligence can create profitable work for banks. Companies need equity, loans, bonds, hedging and advice to fund data centres, chips, power infrastructure and acquisitions.

That does not necessarily mean the underlying investment boom is sustainable. Banks can earn fees when money is raised whether the projects eventually produce attractive returns or not.

For investors, the important distinction is between activity and value creation.

How banks make money from an investment boom

Investment banks act as intermediaries between companies seeking capital and investors willing to provide it. When markets are confident, this can produce several connected sources of revenue.

Underwriting fees

When a company issues shares or bonds, banks may organise the transaction, market it to investors and help determine its price. In return, they receive underwriting fees.

These revenues can have considerable operating leverage. Banks maintain teams of advisers, traders and compliance staff through the cycle, so additional transactions can make a noticeable contribution to profit once fixed costs are covered.

The reverse is also true. If markets weaken, planned deals can be postponed or cancelled quickly. A strong quarter for underwriting should not automatically be treated as a new long-term earnings level.

IPO and follow-on activity

An active initial public offering market usually indicates that investors are willing to accept uncertainty in pursuit of growth. IPOs generate direct fees, but they can also support trading, market-making and future share issues.

Yet flotation volumes are sensitive to sentiment. If newly listed companies perform poorly, investors may become more demanding about valuation. The next group of businesses may then have to reduce their expectations or delay listing altogether.

IPO activity is therefore both a source of bank revenue and a measure of market confidence.

Debt issuance

Large infrastructure programmes often require funding well before they generate cash. Banks can earn fees from arranging bonds, loans and related interest-rate or currency hedges.

Debt issuance can make an investment cycle appear self-reinforcing. Funding allows companies to spend more, suppliers receive new orders, and stronger reported demand encourages further financing.

However, debt introduces fixed obligations. If expected cash flows arrive late, borrowers can face higher interest costs, refinancing pressure or weaker credit ratings. Banks may then move from collecting fees to managing credit losses or difficult financing commitments.

Why capital expenditure matters

AI investment involves more than software. It may require data centres, computing equipment, electricity generation, network capacity and specialist property.

This creates a chain of potential beneficiaries. Technology companies spend money, suppliers expand production, infrastructure operators raise capital, and banks arrange the financing.

The central question is not whether spending is increasing. It is whether the spending earns an adequate return.

Investors can assess this by comparing incremental capital expenditure with changes in revenue, operating profit and free cash flow. Rising expenditure can be productive when it creates valuable assets and pricing power. It is less attractive when a company must keep spending simply to maintain its competitive position.

This distinction also matters when reviewing adjusted profit measures. Strong revenue or adjusted EBITDA does not always translate into cash available to shareholders after capital expenditure, interest, tax and other obligations. The same principle applies when assessing acquisitive businesses, as discussed in this review of Supreme's revenue and adjusted EBITDA growth.

Does strong fee income validate the AI story?

Not by itself.

Banks are paid to facilitate transactions. They can benefit from enthusiastic markets without taking the same long-term commercial risk as the company building the infrastructure.

High valuations can also create a reflexive cycle. Rising share prices make it easier to raise capital. That money funds additional spending, which boosts suppliers' order books and appears to support the original optimism.

If commercial returns disappoint, the process can reverse. Valuations fall, financing becomes more expensive, capital expenditure is reduced and planned transactions disappear from banks' pipelines.

The Bank of England's Financial Policy Committee record highlights why investors should consider how leverage and interconnected financing can broaden the consequences of disappointment in AI-related investment.

Assessing the quality of bank earnings

For shareholders analysing a bank or diversified financial group, it helps to separate recurring income from cyclical transaction revenue.

Underwriting and advisory fees can be highly profitable, but they depend on market conditions. Wealth management, custody, payments and other repeatable fee streams may offer greater stability, although none is risk-free.

Useful questions include:

• How much of the earnings improvement came from underwriting and advisory work?
• Are revenues diversified across business lines and customers?
• Did employee compensation absorb much of the additional revenue?
• Is the bank using its own balance sheet to support transactions?
• Are credit provisions unusually low because economic conditions remain supportive?
• Does the bank have sufficient capital if markets weaken?

Investors should also check whether management is presenting a cyclical rebound as structural growth. An improvement in trading or deal activity deserves a different valuation from contracted or recurring revenue. This distinction is also relevant when examining a financial business reporting growth in assets under management, such as Pollen Street Group's reported performance.

A checklist for AI-related investments

Before considering an AI-linked company, fund or investment trust, write down answers to the following questions.

Cash flow: Is the business producing free cash flow after the investment needed to support growth?

Evidence: Does management disclose measurable revenue, customer or margin benefits, or does the case rely mainly on broad AI language?

Competitive advantage: What prevents customers or competitors from replacing the product?

Capital productivity: Is additional spending generating proportionate growth in profit and cash flow?

Funding: Can the company finance expansion internally, or does it depend on favourable debt and equity markets?

Balance sheet: What are its debt maturities, interest costs, lease commitments and refinancing requirements?

Valuation: What assumptions about growth, margins and future spending are already reflected in the price?

Downside: What evidence would show that the original investment thesis was wrong?

The FCA's guidance on spotting investment hype and managing FOMO is particularly relevant when a compelling technology narrative creates pressure to act quickly.

Follow the cash rather than the excitement

A financing boom can produce real earnings for banks while still creating poor outcomes for some companies and investors. These ideas are not contradictory.

The strongest businesses should eventually demonstrate that investment in AI produces defensible revenue, attractive returns on capital and sustainable free cash flow. Banks should be judged on the diversity and resilience of their earnings, rather than one unusually active period for transactions.

For UK retail investors, the durable approach is straightforward: identify who is supplying the capital, who is carrying the risk and who ultimately captures the economic return. Market excitement may open the financing window, but cash generation determines whether the investment case can survive after that window closes.