Anthropic's $63 Billion Deal: How AI Companies Are Competing With the US Government for Borrowing

The AI boom has found a new source of money: Wall Street

For years, the artificial intelligence boom was largely a story about technology companies racing to build better models, attract users and dominate the next generation of computing. Now, it is becoming a story about something less glamorous but considerably more expensive: borrowing money.

The latest example came on October 26, 2026, when Bank of America, Citigroup, Morgan Stanley and Blackstone announced a $63 billion financing arrangement to support the deployment of artificial intelligence computing infrastructure for Anthropic, the company behind Claude.

The financing will support the purchase of Broadcom's AI accelerators and networking equipment, which will subsequently be leased to Anthropic. The arrangement will use a special-purpose vehicle, a separate legal entity created to hold the assets and arrange the financing.

It is an extraordinary transaction, not least because it illustrates how the AI industry is increasingly relying on the financial system to fund its expansion.

Anthropic is not simply raising money from venture capital investors to develop another AI model. Its computing ambitions are now large enough to require financing structures involving some of Wall Street's biggest financial institutions.

And Anthropic is not alone.

Microsoft, Alphabet, Amazon, Meta, Oracle and other companies are spending enormous sums on data centres, advanced chips, electricity infrastructure and computing capacity. The AI race is becoming a capital-intensive industrial expansion, with borrowing increasingly helping to pay the bill.

The important question is whether this investment boom is beginning to compete with the US government for the money available in financial markets.

Why Anthropic needs $63 billion

To understand the significance of the financing, it is worth looking at what artificial intelligence actually requires.

Developing and operating advanced AI models is extraordinarily expensive. Companies need specialised processors, enormous data centres, networking equipment, electricity, cooling systems and the infrastructure required to connect everything.

Unlike a conventional software company, which can often expand its customer base without proportionately increasing its physical infrastructure, an AI company must continually secure computing capacity to train and operate increasingly sophisticated models.

Every new generation of models can require more computing resources. Every additional user creates demand for inference capacity. And every company racing to build more capable systems needs access to the same limited supply of advanced chips and data centre infrastructure.

That is where the financing problem begins.

Anthropic's $63 billion arrangement involves a special-purpose vehicle that will purchase Broadcom equipment and lease it to Anthropic. Bank of America, Citi and Morgan Stanley are arranging the senior financing, while Blackstone is providing the committed second-lien financing. Broadcom is also providing residual-value support for the equipment.

In simpler terms, financial institutions are helping finance the machines that will power Anthropic's AI services, while the equipment is leased to the company that needs it.

The structure is significant because it demonstrates how AI infrastructure can be financed through arrangements involving banks, private equity firms, equipment manufacturers and institutional investors.

The company building the AI model does not necessarily need to purchase every chip and data centre asset outright. It can obtain access through financing and leasing arrangements.

For Wall Street, this creates a new category of infrastructure financing. For AI companies, it provides another way to expand without paying the entire cost upfront.

But it also means that the AI boom is becoming increasingly connected to the debt markets.

America's AI companies are becoming major borrowers

The scale of investment required by the AI industry is changing how technology companies finance their operations.

For much of the previous decade, large technology companies could finance substantial investments through their enormous cash flows. Their businesses generated enough money to fund acquisitions, research and development, and infrastructure expansion without relying heavily on borrowing.

That is changing as AI infrastructure spending accelerates.

According to the Bank for International Settlements (BIS), major technology companies increased their corporate bond issuance substantially in 2025, with gross issuance exceeding $100 billion.

The organisation also identified growing use of private credit and off-balance-sheet financing arrangements to fund data centre development.

The distinction matters because the borrowing is not always visible in the conventional corporate debt figures.

A technology company might finance a data centre through a separate company that owns the building and equipment. The technology company then signs a long-term agreement to lease the infrastructure or purchase its computing capacity.

The debt may sit with the separate entity rather than appearing directly on the technology company's balance sheet.

The BIS describes some of these arrangements as "shadow borrowing" because they create debt-like economic obligations outside the conventional corporate balance sheet.

This does not automatically mean companies are hiding financial problems. Such structures can serve legitimate financing purposes. However, they make it harder for investors to understand the full extent of financial commitments supporting the AI expansion.

In other words, the AI industry's appetite for money may be considerably larger than traditional corporate debt figures suggest.

The US government has its own borrowing problem

Here is where the story becomes particularly interesting.

The US government is already one of the world's largest borrowers. Washington finances spending through taxation, borrowing and other government revenues. When expenditure exceeds revenue, the Treasury issues securities to cover the difference.

Investors purchase Treasury bills, notes and bonds in exchange for interest payments.

These securities are central to the global financial system. Banks, pension funds, insurance companies, investment managers, foreign governments and central banks hold them as investments, reserves or collateral.

But government borrowing has continued to expand.

According to the Joint Economic Committee's October 2026 debt update, the gross US national debt reached approximately $40.25 trillion on October 5, 2026. Debt held by the public accounted for approximately $32.43 trillion.

The government also faces a growing interest bill. The average interest rate on marketable federal debt reached 3.518% in September 2026, according to the same update.

The implications are substantial. As existing government securities mature, the Treasury must refinance them. If new borrowing takes place at higher interest rates, the government can face increasing debt-servicing costs.

Washington therefore has a continuing need to attract investors willing to purchase its securities.

Now consider what happens when the private sector suddenly needs hundreds of billions of dollars to finance another major investment cycle.

The US government is not the only institution asking financial markets for money anymore.

Are Anthropic and Big Tech competing with Washington for the same money?

Technically, yes, but the relationship requires some explanation.

The US Treasury and technology companies do not borrow in precisely the same way. The Treasury issues government securities, while companies raise money through corporate bonds, bank loans, private credit, leasing arrangements and structured financing.

Nevertheless, these financing activities draw on overlapping pools of capital.

Pension funds, insurance companies, investment managers, banks and other financial institutions have finite resources and investment mandates. They must decide how much money to allocate across different assets.

Treasury securities are generally considered among the safest investments in global financial markets because they are backed by the US government.

Corporate debt offers investors a different proposition. Companies generally need to provide higher returns to compensate investors for additional credit risk.

If an AI infrastructure project offers attractive returns, investors may allocate more money towards corporate debt and related financing opportunities.

That could increase the cost of borrowing for other businesses and, under certain conditions, put upward pressure on government borrowing costs.

Economists describe this potential mechanism as crowding out.

It occurs when increased borrowing or investment demand from one part of the economy makes financing more expensive or less available to another.

The Federal Reserve Bank of Dallas examined this issue in February 2026, identifying three important channels through which AI infrastructure financing could affect interest-rate markets: long-term corporate bond issuance, interest-rate swaps associated with private-credit borrowing and potential crowding out of other financial issuers.

Its analysis estimated that AI-related investment-grade corporate bond issuance could reach approximately $300 billion in 2026.

That would represent a substantial new source of long-term debt supply.

However, there is an important qualification. More corporate borrowing does not automatically mean the US government will pay higher interest rates. Treasury yields are influenced by many factors, including inflation expectations, Federal Reserve policy, economic growth, government borrowing requirements and global demand for safe assets.

AI borrowing is becoming an important consideration in the bond market, but it has not been established as the principal driver of Treasury yields.

The $1 trillion AI infrastructure race

The financing requirements could become even larger.

According to the Bank of England's July 2026 Financial Stability Report, market participants' estimates of AI hyperscaler capital expenditure for 2028 had increased to more than $1 trillion, compared with less than $600 billion in estimates available at the end of 2025.

The report also highlighted the increasing use of external financing, including corporate bonds, private credit and structured financing.

This is a remarkable development.

The companies leading the AI industry are among the most valuable businesses in the world, yet even their enormous cash reserves are not necessarily sufficient to finance every planned infrastructure project.

The construction of data centres requires enormous upfront investment. Electricity infrastructure, advanced cooling systems, networking equipment and specialised processors all add to the bill.

And the spending does not end when the building is completed. Data centres require maintenance, electricity, hardware upgrades and continuous investment to remain competitive.

A data centre built today could become technologically outdated much sooner than the building itself.

That creates a difficult financing question: how long will the infrastructure generate enough revenue to justify the money spent constructing it?

For investors, the answer matters because debt must eventually be repaid regardless of whether the latest AI product becomes commercially successful.

Technology companies can delay a product launch. Creditors generally cannot delay an interest payment simply because the chatbot is having a difficult quarter.

Private credit is becoming an important part of the AI boom

One of the most important developments is the growing role of private credit.

Private credit refers to loans provided by non-bank financial institutions, including private investment funds, rather than through conventional public bond markets.

These funds can finance projects that might not fit traditional bank lending arrangements. They can also structure complicated transactions involving data centres, equipment, infrastructure and long-term contracts.

For AI companies, private credit offers another source of financing.

For investors, it creates opportunities to earn returns from the infrastructure supporting artificial intelligence.

However, private credit is generally less transparent than publicly traded corporate bonds. Investors may have limited visibility into the underlying loans, the financial health of borrowers and the full extent of interconnected exposures.

The BIS has warned that financing structures involving special-purpose vehicles, private credit funds, insurers and banks can create new channels through which financial shocks spread.

Consider a situation in which demand for AI computing grows more slowly than expected.

A data centre operator may find that it cannot lease its entire capacity. The revenue expected to service its loans could fall short. Its lenders may then face refinancing difficulties, while the technology company leasing the infrastructure could reconsider its commitments.

The risks do not necessarily remain within one company.

This is why the financing of AI infrastructure is attracting attention from financial regulators, not just technology investors.

What happens if the AI investment boom disappoints?

There is a difference between investing in a technology with enormous potential and investing in every company building infrastructure around it.

Artificial intelligence could transform productivity, medicine, education, manufacturing and financial services. But that does not guarantee that every data centre, chip investment or AI business will generate sufficient returns.

The industry is making enormous investments based on expectations of future demand.

If AI adoption grows more slowly than companies anticipate, some infrastructure projects could become less profitable. If model development becomes cheaper or computing requirements change, certain investments could lose value.

This is particularly important for debt-financed projects.

Equity investors can lose money when a company performs poorly. Debt obligations, however, remain contractual commitments.

A company cannot simply tell its lenders that the AI revolution is taking longer than expected and ask them to accept payment in enthusiasm.

The Bank of England has highlighted the potential financial stability implications of increasing AI-related borrowing, particularly if an adverse shock affects companies' ability to service their debts.

That does not mean an AI debt crisis is inevitable. It means that the growing scale and complexity of financing deserve closer attention.

The financial system is becoming more exposed to the assumption that demand for AI computing will continue expanding.

Could AI borrowing make America's national debt more expensive?

This is the central economic question.

The United States already faces substantial fiscal pressures. Its national debt has grown, interest payments consume an increasing share of government resources, and the Treasury must continually refinance maturing securities.

The AI industry is adding another major source of financing demand.

If AI companies continue issuing large amounts of long-term debt, financial markets must absorb additional securities. If private-credit funds finance more infrastructure, they may also need to hedge their exposure through interest-rate markets.

These activities can influence the supply and demand dynamics that determine borrowing costs across the economy.

However, the relationship is not entirely negative for government financing.

AI companies that raise money do not necessarily spend every dollar immediately. Some financing proceeds may remain in cash or money-market funds, which can themselves hold Treasury securities.

Furthermore, successful AI investment could increase productivity, economic growth and future tax revenues. If AI makes American businesses more productive, the government could ultimately benefit from a larger economy.

There is also a possibility that AI-driven productivity improvements could strengthen the long-term economic outlook and support demand for US financial assets.

The outcome therefore depends on whether the economic returns from AI infrastructure justify the enormous capital being deployed.

For now, the more immediate concern is the scale of financing required and the increasing connections between technology companies and financial institutions.

The bigger story: Wall Street is financing America's AI ambitions

The AI boom is increasingly becoming a financial story as much as a technological one.

Anthropic's $63 billion financing arrangement demonstrates how deeply the industry's expansion is becoming connected to major banks, private investment firms and equipment manufacturers.

Microsoft, Alphabet, Amazon, Meta and other companies are also committing enormous resources to computing infrastructure.

Meanwhile, the US government continues to borrow trillions of dollars to finance its own obligations.

Both the public and private sectors need access to capital, and both are increasingly dependent on the financial markets that connect global investors with borrowers.

This does not mean Anthropic is directly taking money away from the US Treasury, nor does it establish that AI borrowing is responsible for rising government interest costs. But it does mean that the AI infrastructure boom is becoming large enough to influence the broader financing environment.

For years, the defining question of the AI race was who would build the most powerful model.

Now another question is becoming just as important: who will finance the machines, electricity and infrastructure required to run it?

The companies building artificial intelligence are no longer just competing for customers, chips and talent. They are becoming major participants in the global debt market.

And Washington, which has spent decades borrowing on an enormous scale, now has another group of very ambitious customers sitting across the table from Wall Street.

The AI revolution may eventually make the economy more productive. But before that happens, somebody has to pay for the infrastructure. And increasingly, that somebody is the credit market.