One of the largest investment projects in the field of artificial intelligence is facing a serious financial obstacle. The problem is not a lack of capital or investor interest, but a question that could change the way the entire industry is financed: how much will today's most expensive computer chips be worth in five or ten years?

US investment giant Blackstone is facing uncertainty over the implementation of a financing arrangement worth approximately $6.35 billion, intended to develop computing infrastructure that would be used by Anthropic, one of the world's leading developers of artificial intelligence systems. Although this is an industry attracting enormous investment and representing one of the most important directions of technological development, financial institutions are increasingly scrutinising the risks associated with long-term financing of such projects. The core issue is not whether artificial intelligence will continue to develop, but whether the infrastructure being financed today with billions of dollars will retain sufficient economic value throughout the loan repayment period.

The development of advanced artificial intelligence systems requires extremely powerful computing processors, primarily graphics processing units, known as GPUs, which enable the processing of vast amounts of data and the execution of complex computing operations. These processors are among the most expensive components of modern data centres. Individual chips can cost tens of thousands of dollars, while complete computing systems can reach values of several million dollars. However, unlike traditional infrastructure such as office buildings, power plants or industrial facilities, computer processors have a relatively short period of technological competitiveness. A processor that represents the cutting edge of technology today may become significantly less efficient within just a few years compared with a new generation of products offering greater computing power, lower electricity consumption and reduced data-processing costs. This rapid pace of technological change represents one of the most significant risks for financial institutions considering long-term lending for AI infrastructure.

Why Are Credit Rating Agencies Concerned About the Value of Chips?

For a financing arrangement to receive an investment-grade credit rating, credit rating agencies must assess the likelihood that financial obligations will be met, the stability of future cash flows and the quality of the assets serving as collateral. In the case of financing computing infrastructure, one of the main challenges is the rapid economic depreciation of technological equipment. If one billion dollars is invested today in the most advanced computing processors, their market value may be significantly lower in three or five years, even if they remain technically fully functional. The reason is simple: a new generation of processors may be so much more efficient that using older equipment becomes economically less attractive. This creates a serious problem when the equipment is financed through long-term loans, as there is a possibility that, at a certain point, the outstanding debt will exceed the market value of the assets underpinning the financing.

Unlike real estate financing, where banks can assess the long-term value of land and buildings with relative confidence, financing computer processors involves considerably greater uncertainty. An office building can retain a significant portion of its value for several decades, while a computer processor can lose much of its market value in just a few years. If a borrower defaults on its financial obligations, the lender must consider selling the equipment to recover the outstanding debt. However, it is difficult to predict in advance how many buyers there will be for a particular generation of processors after five years and what their market price will be. Credit rating agencies may therefore require higher equity contributions, additional guarantees, shorter repayment periods or more conservative estimates of future revenues. It should be emphasised, however, that a credit rating does not depend solely on the value of the chips, but also on the creditworthiness of the infrastructure user, the agreed lease payments, the reliability of future revenues and the overall financing structure.

Why Is an Investment-Grade Credit Rating So Important to Blackstone?

Blackstone, one of the world's largest investment managers, is seeking to secure capital for projects of this kind from institutional investors, including insurance companies, pension funds and other major financial asset managers. These institutions often operate under strict risk-management rules that limit investments in financial instruments with lower credit ratings. An investment-grade credit rating, which under certain international rating scales begins at BBB−, represents an important dividing line between financial instruments considered relatively safer and those carrying increased credit risk. If the planned arrangement does not receive the expected rating, the number of potential investors could decrease significantly, while the remaining investors may demand higher interest rates and more favourable terms to protect their claims. The consequences could include higher overall financing costs, the need for additional capital or changes to the entire financing structure.

Therefore, Blackstone's problem is not necessarily an inability to find investors interested in artificial intelligence, but rather that certain categories of institutional investors may be unable to participate in the financing under the originally planned terms. This is an important distinction, as it demonstrates that even projects with enormous technological potential must meet traditional creditworthiness and financial sustainability criteria. Investors assess not only the possibility of generating high revenues, but also the likelihood of recovering their invested capital under unfavourable market conditions.

What Happens When Chips Lose Half Their Value?

The problem can be illustrated with a simple financial example. Suppose an investor purchases computing equipment worth one billion dollars, financing $800 million through a loan and contributing the remaining $200 million in equity. At the beginning of the financing period, the debt-to-value ratio is 80%, meaning there is a certain equity cushion protecting creditors against a potential decline in the value of the assets. However, if the market value of the equipment falls to $400 million after three years, while the outstanding loan remains at $600 million, a significant imbalance emerges between the value of the assets and the amount of debt. In this situation, the loan-to-value ratio reaches 150%, meaning that selling all the equipment at its estimated market value would not be sufficient to repay the creditor in full.

Of course, a decline in the value of equipment does not automatically mean that the borrower will stop repaying the loan. If the data centre generates stable revenues and the infrastructure user continues to meet its contractual obligations, the financing may remain sustainable despite the decline in the market value of the processors. However, credit rating agencies must also analyse adverse scenarios, including falling demand, lower prices for computing services, the termination of long-term contracts or financial difficulties experienced by the infrastructure user. A particular risk arises when revenues and the value of collateral decline simultaneously, as creditors may then be left without sufficient cash flow to service the loan and without assets whose sale would enable them to recover their claims in full.

Is Artificial Intelligence Creating a New Category of Financial Risk?

Traditional infrastructure projects, such as power plants, telecommunications networks, roads and commercial buildings, generally have relatively long economic lifespans and can generate revenue over several decades. The situation is considerably more complex with data centres designed for artificial intelligence, because the physical infrastructure itself may have a long useful life, while its most expensive technological components may need to be replaced much sooner. This creates a potential mismatch between long-term financial obligations and the economic lifespan of the assets that generate revenue. Additional uncertainty comes from future electricity costs, changes in computing capacity prices, the development of new technologies and the possibility that some AI models may eventually operate using significantly fewer computing resources than they do today.

This does not mean that investment in artificial intelligence infrastructure is unjustified. Rather, it requires a different approach to risk assessment. Financial institutions must carefully analyse the period during which equipment remains economically viable, the possibility of replacing it, the stability of long-term contracts and the user's ability to generate sufficient revenue. One possible solution is to shorten loan repayment periods so that financial obligations are better aligned with the expected economic lifespan of the chips. Another option is to increase investors' equity contributions, thereby reducing creditors' exposure. Long-term contracts with financially strong users of computing capacity may also provide an important layer of protection for investors, particularly when they include appropriate guarantees and clearly defined payment obligations.

What Does This Case Tell Us About the Future of Global Investment?

The development of artificial intelligence has attracted enormous amounts of capital in recent years, with the world's largest technology companies announcing investments worth tens and hundreds of billions of dollars in data centres, processors and energy infrastructure. However, the pace of technological development is simultaneously increasing uncertainty about the long-term profitability of certain investments. The greatest risk is not necessarily that artificial intelligence will fail to deliver the expected technological progress, but that a particular generation of computing infrastructure may lose its economic competitiveness before investors manage to recover their invested capital. This is precisely why financial markets increasingly need to distinguish between the prospects of an industry and the creditworthiness of individual projects operating within it.

The Blackstone and Anthropic case is an important indication that the global artificial intelligence industry is gradually confronting traditional financing rules. Technological innovation can create enormous economic value, but it does not in itself guarantee stable cash flows or sufficient collateral value. Credit rating agencies and institutional investors will therefore likely pay increasing attention to financing structures, the quality of contractual relationships, equipment depreciation periods and projects' ability to withstand adverse market conditions. In the long term, the ability to align technological development with financial discipline could become one of the most important factors determining success in the artificial intelligence industry.

Conclusion

The most important lesson from this case is that even the most advanced technology cannot escape the fundamental principles of the financial system. A credit rating is not determined by how revolutionary a particular technology is or how much attention it attracts in the market, but by the borrower's ability to meet its obligations on time and the likelihood that creditors will recover their claims. Computer chips that form the foundation of artificial intelligence development today can simultaneously represent a significant financial risk precisely because they become technologically obsolete so quickly. As a result, financing AI infrastructure will require more careful alignment of loan maturities, expected revenues and the economic lifespan of equipment.

In a world where a computer chip can lose a substantial portion of its market value in just a few years, financing such equipment over a ten-year period cannot be treated in the same way as financing real estate, a power plant or other traditional infrastructure. This difference lies at the heart of the financial challenge facing Blackstone, Anthropic and a growing number of investors in the global artificial intelligence industry.