The Yardstick Is Still Missing

Banks are preparing to expand financing for artificial intelligence investments. Yet most have not developed separate criteria to measure the risks specific to these investments.

A bank’s costliest mistake may not be a miscalculation. Not knowing what it has failed to measure can create a bigger problem.

There was a detail that attracted little attention in the bank lending survey published by the European Central Bank (ECB) on October 27.

Banks expect growing demand for loans to finance artificial intelligence investments, particularly data centers. Yet most banks responding to the survey say they have not yet established quantitative criteria specific to this area when assessing AI-related loans.

This does not mean banks are lending without controls. Traditional credit analysis continues. But the methods used to assess a new technology’s economic life, capacity utilization risk and future revenue are still developing.

Interestingly, another development that same week highlighted this measurement problem in finance from a different angle.

On October 22, London Stock Exchange Group (LSEG) agreed to acquire credit data and analytics provider Octus for $3.535 billion in cash.

Octus specializes in researching hard-to-access contracts, collateral terms and the financial covenants borrowers must meet, particularly in private credit markets. The company turns this information into structured data that banks and investors can use.

Octus, whose 2026 revenue is expected to exceed $300 million, has more than 900 employees.

At a time when AI can review documents in seconds, paying billions of dollars for a company where people research and verify hard-to-access information may seem surprising at first.

But putting the two developments side by side changes the picture somewhat.

As the financial world rapidly expands its capacity to process data, the need to obtain the information behind decisions and measure new risks accurately does not disappear.

The price of unknown risk

Consider a loan agreement.

You have the borrower’s financial statements. You have calculated its cash flow and assessed its ability to repay debt. But a subsequently amended contractual provision may have changed certain creditors’ priority or their rights over collateral.

If you are unaware of that change, your investment decision may be wrong even if every calculation is mathematically correct.

Incomplete information can sometimes lead to a higher risk premium, sometimes to more stringent collateral requirements, and sometimes to no loan being granted at all.

Nor are risky companies the only ones that pay the price. A financially strong business may also fail to secure the financing terms it believes it deserves if it cannot present its position transparently enough.

There is an important distinction here: More information does not always mean cheaper credit. The facts that emerge may sometimes show that the risk is higher than previously thought.

The value of information lies not in necessarily producing a favorable outcome, but in its ability to prevent risk from being mispriced.

This is also why the specific focus on private credit, illiquid investment vehicles and complex financial products in the U.S. Securities and Exchange Commission’s (SEC) 2027 examination priorities, announced on October 23, is significant.

These developments are not consequences of one another. But they reveal different aspects of the same financial need: As capital grows, the risk it carries needs to be visible.

Which risks are sufficiently visible in Türkiye?

The question facing companies in Türkiye is somewhat different.

On October 22, the Central Bank of the Republic of Türkiye (CBRT) cut its policy rate to 36%. Yet in the week of October 16, the weighted average annual interest rate on Turkish lira commercial loans stood at 48.44%.

September’s manufacturing PMI was in contraction territory at 47.9. Consumer confidence fell to 84.8 in October.

By contrast, data from the Union of Chambers and Commodity Exchanges of Türkiye (TOBB) show that the number of companies established in the first nine months of the year was 5.4% higher than in the same period a year earlier, while the number of company closures was 1.5% lower. It would be wrong to say that the entire economy is under the same degree of strain.

But the reasons behind businesses’ increased demand for credit in the CBRT’s third-quarter lending survey are striking: debt restructuring, inventory building and working capital needs.

Under these conditions, a company’s most important information gap may not concern what a new AI model can do.

The real gaps may be how reliable its existing customers are, when orders will turn into cash, or which payments it may be unable to meet over the next three months.

For example, a manufacturer may have a strong order book. But if customers are taking longer to pay, inventories are growing and the credit used to finance production is becoming more expensive, sales growth can worsen the company’s cash position.

In this situation, before developing a more sophisticated sales forecast, the company needs an accurate view of the entire process from order to collection.

AI can accelerate this process. It can identify changes in customers’ payment behavior, flag the possibility of a cash shortfall earlier and compare different scenarios.

But if customer contracts are out of date, receivables records are inaccurate or the validity of collateral has not been checked, the system will produce a faster but still incomplete assessment.

For businesses in Türkiye, particularly those facing high financing costs, this distinction has direct economic consequences.

Capital tied up in unnecessary inventory, receivables collected late or misjudged customer risk are not merely operational problems. They also affect financing needs and, consequently, a company’s actual profitability.

Whose information underpins the valuation?

The amount LSEG has agreed to pay for Octus is a strong indication of the commercial value of reliable information. But we cannot say that the price is being paid for data alone. The deal also includes existing customers, specialist teams, technology and expected future growth. Moreover, the acquisition has not yet been completed.

From a company valuation perspective, the more interesting question is this:

If AI makes financial analysis increasingly inexpensive, will the competitive advantage of companies producing analysis grow stronger, or will it be that of companies owning reliable, hard-to-access information?

There is no single answer. How current the data is, access rights, the extent to which competitors can obtain it and its ability to generate commercial results will be decisive.

For company management, however, the decision is more concrete.

Before allocating capital to a new technology, management should assess which financial uncertainty it will resolve, its contribution to cash flow and the investment’s payback period. Particularly in lending, investment and acquisition decisions, the risk created by missing information should be assessed alongside the accuracy of the information being used.

AI can accelerate calculations. But deciding which calculations are worth making and which information cannot yet be trusted still requires serious financial judgment.