Anthropic reportedly seeks $60 billion in AI chip financing as Blackstone considers a $9 billion investment

Anthropic is reportedly negotiating a $60 billion debt-financing package to support artificial intelligence computing infrastructure, with Blackstone considering $9 billion in junior financing at a yield exceeding 10%. The proposed Project North Star arrangement is associated with approximately 1.5 gigawatts of Google TPU capacity and involves Broadcom as a technology and financing partner.

The reported structure divides financing between $42 billion of senior debt and $18 billion of junior debt. Blackstone's proposed $9 billion is part of the junior tranche, not an additional amount on top of the $60 billion. The structure illustrates how frontier AI developers can secure computing capacity through long-term commitments supported by institutional lenders rather than funding every hardware purchase through equity.

The transaction has not been established as fully closed. Proposed commitments, financing availability, actual drawdowns and completed hardware purchases are different events, and the final agreements will determine which parties bear the associated risks.

··········

PROJECT NORTH STAR: HOW THE $60 BILLION FINANCING IS STRUCTURED.

The proposal separates capital into senior and junior layers with different repayment priorities and risk exposures.

........

Component	Reported amount	Share of total
Senior debt	$42 billion	70%
Junior debt	$18 billion	30%
Blackstone portion of junior debt	$9 billion	15%
Other junior financing	$9 billion	15%
Total financing	$60 billion	100%

........

The Blackstone amount is included within the junior debt and should not be counted twice. Senior lenders generally benefit from priority claims and potentially stronger security, while junior investors accept greater loss exposure in return for higher yields. The exact legal waterfall, collateral package and guarantees are not yet fully established by public transaction documentation.

Broadcom reportedly supports the senior financing. The financial significance of that support depends on whether it consists of guarantees, purchase commitments, other credit enhancements or some combination; it would be inappropriate to treat the full $42 billion as an unconditional Broadcom guarantee without final terms.

··········

BLACKSTONE'S REPORTED DOUBLE-DIGIT YIELD INDICATES THE COST OF JUNIOR RISK.

A yield exceeding 10% on the proposed $9 billion junior investment suggests that lenders are seeking substantial compensation for exposure to AI infrastructure. It does not mean the entire $60 billion package carries the same rate, nor does a quoted yield necessarily equal the annual cash coupon.

At a simplified annual return of 10%, the investment would generate $900 million. A higher yield increases that amount, although discounts, fees and payment-in-kind provisions could change the cash received.

........

Illustrative yield	Annual return on $9B
10.0%	$900M
10.5%	$945M
11.0%	$990M
12.0%	$1.08B

........

These are mathematical illustrations rather than disclosed payment obligations. A debt security issued at a discount may offer a higher effective yield than its coupon, and the financing could include deferred interest or amortization. Investors therefore need final pricing, maturity and repayment terms before translating the reported yield into an annual cash-flow forecast.

··········

SENIOR AND JUNIOR TRANCHES ALLOCATE LOSSES DIFFERENTLY.

The $42 billion senior tranche represents 70% of proposed financing, while $18 billion of junior capital sits beneath it in the capital structure. In a simplified loss-absorption framework, the junior layer can absorb losses before senior lenders are impaired.

........

Hypothetical loss	Junior capital remaining	Senior impairment
$0	$18B	$0
$5B	$13B	$0
$10B	$8B	$0
$18B	$0	$0
$25B	$0	$7B

........

This is an illustrative waterfall, not a representation of signed Project North Star contracts. Recoveries, collateral rights, guarantees and intercreditor arrangements could materially change the actual allocation.

Subordination explains why junior investors demand higher returns even when senior lenders may regard the same project as comparatively well protected. The credit risk is concentrated rather than distributed equally.

··········

BROADCOM'S ROLE CONNECTS HARDWARE DEMAND WITH FINANCIAL EXPOSURE.

Broadcom is associated with the custom AI infrastructure involved in Anthropic's computing expansion and reportedly plays a role in supporting the financing. Such an arrangement can help suppliers secure large orders by making infrastructure purchases easier for customers to finance, but it may also create contingent liabilities or counterparty exposure.

Hardware revenue and credit support must be analyzed separately. A supplier can generate revenue from delivering equipment while assuming obligations tied to the financing or customer performance. The scope of any guarantees, maturity profile and accounting treatment would determine whether those obligations affect liquidity, leverage or reported contingent commitments.

For Broadcom investors, the key questions are not merely how much equipment is ordered, but what financial risks accompany the sales and how concentrated those risks become among a small number of large AI customers.

··········

THE PROPOSAL IS LINKED TO APPROXIMATELY 1.5 GIGAWATTS OF GOOGLE TPU CAPACITY.

The proposed financing is associated with roughly 1.5 gigawatts of Google TPU infrastructure, an indication of the scale of the computing buildout. At a hypothetical constant load, 1.5 GW over 8,760 hours would consume 13.14 terawatt-hours per year.

This is not a forecast of electricity consumption. Actual demand depends on deployment timing, utilization, facility efficiency and whether the quoted capacity measures IT equipment or total facility power.

Dividing the reported $60 billion financing by 1,500 megawatts produces $40 million per MW. That is a scale comparison, not a construction-cost estimate: the financing could cover equipment, supporting infrastructure and long-term contractual commitments that are not captured by a simple megawatt figure.

··········

DEBT FINANCING PRESERVES EQUITY BUT CREATES FUTURE CASH OBLIGATIONS.

Infrastructure financing can allow Anthropic to secure additional compute without issuing new shares for an equivalent amount. Equity financing dilutes ownership but does not generally require fixed contractual repayments; debt and financed leases preserve ownership while creating scheduled payments and other obligations.

The economic advantage depends on whether the added capacity generates sufficient operating cash flow to cover those commitments. Strong utilization and demand growth could justify financing a large infrastructure base. Conversely, slower adoption, falling inference prices or delayed deployment could pressure cash flow even if the contractual payments remain unchanged.

A $60 billion financing package should not automatically be treated as $60 billion of immediate cash proceeds on Anthropic's balance sheet. The accounting outcome depends on the final legal structure, drawdown schedule and whether obligations are recorded as debt, leases or other commitments.

··········

SCENARIO ANALYSIS: WHAT COULD THE BLENDED COST OF FINANCING BE?

The reported Blackstone yield concerns part of the junior financing. Without final rates on the $42 billion senior tranche and the full $18 billion junior tranche, the package's weighted financing cost is unknown.

A scenario calculation illustrates sensitivity. At an assumed 6% senior rate and 10.5% junior rate, annualized financing cost would be $42B × 6% + $18B × 10.5% = $4.41 billion, equivalent to a 7.35% blended rate.

........

Illustrative scenario	Senior rate	Junior rate	Annual cost	Blended rate
Lower-cost	5%	10%	$3.90B	6.50%
Base illustration	6%	10.5%	$4.41B	7.35%
Higher-cost	7%	11%	$4.92B	8.20%
Stress	8%	12%	$5.52B	9.20%

........

These figures are hypothetical, not reported financing terms. They exclude fees, guarantees, deferred interest, principal amortization and staggered drawdowns.

A one-percentage-point increase in the blended annual rate on a fully outstanding $60 billion principal would represent $600 million of additional annualized cost. The scale underscores why financing structure matters as much as the nominal size of the transaction.

··········

ANTHROPIC'S REVENUE GROWTH DOES NOT AUTOMATICALLY ESTABLISH DEBT CAPACITY.

Anthropic reportedly generated approximately $16.8 billion in third-quarter 2026 revenue, up from $11.6 billion in the preceding quarter, equivalent to roughly 44.8% quarterly growth. Expanding demand helps explain the pursuit of additional computing capacity, but revenue is not the same as free cash flow available for debt service.

For scale only, the illustrative $4.41 billion annual financing cost corresponds to about $1.10 billion per quarter, or approximately 6.6% of reported third-quarter revenue. This is not a forecast of actual interest expense, because neither the assumed rates nor full utilization of the proposed financing has been established.

A credible debt-capacity analysis must incorporate infrastructure operating expenses, model training, personnel, capital expenditures, other obligations and the timing of contractual payments. Rapid top-line growth can coexist with substantial cash requirements.

··········

COLLATERAL DEPRECIATION AND TECHNOLOGICAL OBSOLESCENCE COMPLICATE CREDIT RISK.

AI accelerators can lose economic value as newer hardware delivers more compute per unit of electricity or capital. That creates potential tension between long-term financing commitments and the resale value of the equipment securing them.

If Anthropic meets contractual obligations, lenders may primarily depend on the payment stream. If payments are interrupted, recovery could depend on specialized hardware, guarantees and restructuring rights. The realizable value of chips depends on their age, configuration, secondary-market demand and compatibility with existing facilities.

Credit analysis should therefore examine contract enforceability, customer concentration and guarantees alongside hardware values. A nominal equipment purchase price does not necessarily equal the collateral's recoverable value several years later.

··········

PRIVATE CREDIT INVESTORS ARE FINANCING THE AI INFRASTRUCTURE EXPANSION.

Institutional private credit provides an alternative source of capital for large infrastructure projects supported by contractual payments. For Blackstone, junior debt could offer exposure to AI computing demand without acquiring equity in Anthropic, but with limited upside beyond agreed financing returns and potentially significant downside if the project underperforms.

A yield above 10% is consistent with investors pricing uncertainty around utilization, obsolescence, refinancing and counterparty obligations. It does not by itself demonstrate distress, nor does the availability of capital prove the project is low risk.

The proposed transaction follows an earlier financing arrangement associated with approximately $35 billion of additional Anthropic computing capacity. Comparing the two nominal package sizes produces an increase of about 71.4%, although different transaction structures mean this should not be read as an equivalent increase in debt outstanding.

··········

THE MAIN RISKS GO BEYOND INTEREST EXPENSE.

The financial outcome depends on several interacting variables rather than any single quoted interest rate.

........

Risk	Potential consequence
Weaker AI demand	Lower utilization and cash generation
Falling inference prices	Pressure on revenue per unit of compute
Hardware obsolescence	Reduced collateral recovery
Higher borrowing rates	Increased financing cost
Deployment delays	Payments preceding expected revenue
Counterparty concentration	Dependence on a small number of partners
Refinancing constraints	Difficulty replacing maturing obligations
Complex guarantees	Uncertainty over ultimate loss allocation

........

Risks can reinforce one another. For example, newer chips may simultaneously pressure the market value of existing hardware and the price customers are willing to pay for inference, reducing collateral values and operating margins at the same time.

Strong demand, predictable utilization and declining operating costs could have the opposite effect, supporting contractual payments and improving the credit outlook.

··········

WHAT MUST BE VERIFIED BEFORE THE PACKAGE IS TREATED AS COMPLETED.

Reported financing negotiations are not equivalent to executed commitments or funded debt. Investors should seek final agreements covering tranche pricing, maturity, amortization, security interests, guarantees, drawdown conditions and repayment entities.

They should also distinguish Anthropic's own obligations from financing undertaken by special-purpose entities or commercial partners. Depending on the final structure, the relevant accounting disclosures could involve debt, leases, guarantees or other contractual commitments.

Blackstone's proposed $9 billion junior debt participation is not a $9 billion equity investment in Anthropic. Treating it as such would misstate ownership dilution, capital structure and the company's financing economics.

··········

PROJECT NORTH STAR SHOWS HOW CAPITAL-INTENSIVE FRONTIER AI HAS BECOME.

The proposed $60 billion financing combines a large senior tranche, a substantial junior layer and a reported $9 billion Blackstone participation at a double-digit yield. It could enable Anthropic to expand computing capacity without raising an equivalent amount of new equity, while shifting part of the infrastructure financing burden to institutional creditors and commercial partners.

Its sustainability depends on future cash generation, hardware economics, the enforceability of guarantees and the final allocation of losses. The nominal size of the package alone cannot answer those questions.

The most revealing number is the reported yield above 10% on junior financing: institutional capital is willing to fund frontier AI infrastructure, but it is increasingly demanding explicit compensation for the risks of doing so.

··········

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org