The Interest on AI You Do Not Use

Nothing has to go wrong inside a factory for its value to fall.

Its machinery, orders and operating profit may all be unchanged. Yet a purchase price considered reasonable six months ago may look expensive today.

Nothing at the factory has changed. What has changed is the price of the money needed to buy it.

Two central bank publications released last week introduced an unexpected actor into this story: Artificial intelligence.

The Fed’s minutes, published on 7 October, noted that US Treasury yields with maturities of two to ten years had risen by around 35 basis points over the period under review. Strong economic data, interest-rate expectations, geopolitical developments and uncertainty over Treasury operations were among the factors behind the rise.

But market assessments also mentioned another factor: large-scale private-sector borrowing to finance AI infrastructure, increasing competition for capital.

Six days later, European Central Bank Executive Board member Philip Lane highlighted a similar point in a speech in Dublin. He said it was plausible that the global wave of AI investment and expectations of future productivity gains were contributing to higher long-term interest rates.

Europe’s AI investment is smaller than that of the United States. Even so, the effects transmitted through global bond markets can tighten financing conditions in Europe.

Neither statement says AI is the sole cause of higher interest rates. It would be wrong to draw that conclusion.

But central banks are now considering this question:

Could a technology expected to make the economy more productive in the future raise other companies’ cost of capital before that future arrives?

That is what makes the question interesting.

AI investment has an extraordinary appetite for finance. Data centres, processors, energy infrastructure, cooling systems, communications networks… None can be built on a bright technological idea alone. They require real capital today.

Moreover, the timing of the investment’s financing needs does not match the timing of its economic benefits.

Figures shared by the ECB on 13 October are striking in this respect. Euro area companies are preparing to allocate around 10% of their total investment to AI this year. AI-related lending accounts for roughly a quarter of recent growth in corporate credit.

Yet no clear AI effect on aggregate productivity has been identified so far.

That does not mean the investment is being wasted. In technological transformations, it is unsurprising for infrastructure spending to precede productivity gains.

But it creates a substantial financial issue:

The timing of the expected return is uncertain. The debt’s maturity is not.

This is where the AI story meets the financial world most tangibly.

Suppose an investment group wants to buy an industrial company with stable annual operating profit and cash-generating capacity.

The acquisition will use €40 million of debt. If the annual cost of that debt is 7.5% rather than 6%, all other things being equal, the annual interest burden rises by €600,000.

This is a hypothetical example. It is not a real transaction or a measured effect of AI investment.

But the financial outcome is clear.

The target company’s sales have not changed. Its EBITDA has not changed. Its management has made no mistake.

Nevertheless, less cash may remain for equity holders after the acquisition. Debt-service capacity has narrowed. The purchase price the buyer can accept may also fall.

Nor is this outcome limited to M&A. The same logic applies to long-term machinery investments, energy projects and debt-financed capacity expansion.

What is interesting is that a company can become less attractive to investors even when its own performance has not changed.

This is not new in financial markets. What is new is that the enormous demand for capital associated with AI has become part of the debate.

The IMF’s financial stability assessment of 13 October also highlights the other side of the optimism. If AI investments fail to meet profitability expectations, financial assets could undergo a correction. High bond yields and growing borrowing needs are a separate source of vulnerability.

There is something almost ironic about this.

If AI proves more successful than expected, expectations of higher future productivity could put upward pressure on real interest rates.

If it proves less successful, deteriorating investor confidence and the repricing of financial assets could create costs of a different kind.

These are not two inevitable outcomes, of course. They are channels through which risks could emerge under different conditions.

But the message for companies matters: even very strong expectations for the future do not make today’s investment decisions risk-free.

In Türkiye, however, the same debate cannot simply be copied across.

It would be economically misleading to treat US data-centre investment as the main reason for high commercial lending rates in Türkiye.

Monetary policy, inflation, the country risk premium, banks’ funding costs and lending regulations are far more decisive here.

The CBRT held its policy rate at 37% at its meeting on 10 September. According to the 4 September data included in its latest meeting summary, TRY commercial lending rates, excluding overdrafts and credit cards, stood at 49.3%.

This shows that a significant share of Turkish companies already operate with expensive capital.

It does not, by itself, mean that the entire economy is under the same pressure.

According to data released by TurkStat on 9 October, industrial production fell 1.9% year on year in August, while manufacturing output declined 2.5%. The Istanbul Chamber of Industry’s September manufacturing PMI remained in contraction territory at 47.9.

On the other hand, the seasonally adjusted capacity utilisation rate rose to 74.1% in September. Exports increased by 15.4% year on year that month to approximately $26 billion.

The same divergence is visible in the domestic market. Retail sales volume rose 10.1% year on year in August, while wholesale trade volume fell 1.9%.

Companies differ in their resilience. Some have strong export orders; others face weak domestic demand. Some collect their receivables quickly; others need more borrowing to finance growth.

TOBB’s latest published figures for August also resist a one-directional bankruptcy narrative: although the number of companies closing that month rose 11.3% year on year, total closures in the first eight months were 1.1% below the same period a year earlier. Understanding the risks of court-supervised restructuring and bankruptcy additionally requires examining the types of court decisions and their distribution across sectors.

These differences matter because the cost of capital does not affect everyone in the same way.

For example, a manufacturer with strong export revenue, low debt and regular customer collections may be well placed to invest in new technology.

Another business in the same sector may struggle to fund working capital despite being profitable on paper.

For the second company, the most valuable AI investment may not be an impressive automation system. It may be a more modest application that reduces inventory forecasting errors, identifies collection risks early or aligns production plans with actual orders.

The success of such an investment can be measured: how much cash was released? How far did financing needs fall? Which costs were avoided?

Now consider the other side.

For some Turkish companies, global AI investment creates more than financing risk. It could also generate new orders in energy equipment, power electronics, cooling, engineering, infrastructure and certain manufacturing activities.

But winning a large order and doing good business are not always the same thing.

When a company expands capacity to supply equipment for data-centre investment, it may tie up cash for months, from raw materials through production to delivery. If advance-payment terms are weak, collection periods long and financing costs high, expected growth can put pressure on cash.

At that point, the discussion needs to focus on the cost of financing the opportunity, rather than just its size.

Perhaps that is why the most interesting financial consequence of the AI era will not be the difference between companies that use the technology and those that do not.

In a world where capital is being repriced, the quality of their balance sheets may determine which of two companies doing the same work stands on firmer ground.

One may have more technology. The other may have a shorter cash-conversion cycle, more manageable debt and the discipline to avoid poor investments.

The second should not be underestimated.

Because the value technology companies place on the future and the interest an industrial business must pay today ultimately meet in the same capital market.

In the period ahead, relying solely on a base case will not be enough, particularly for debt-financed acquisitions and capacity investments. Debt-service capacity, free cash flow and investment returns should be recalculated under different financing costs. AI investment should not be exempt from that assessment: a promise of productivity that does not show the money the company will actually earn is not, on its own, an investment rationale.