Everything About AI
NextFin News - Anthropic, the maker of the Claude AI assistant, has told investors it may need to seek additional financing sooner than it previously anticipated, while offering little detail on how it plans to reach an investment-grade credit rating - the threshold that would unlock cheaper, deeper pools of capital. The warning, buried in the company's IPO prospectus made public on Monday, pairs four-digit revenue growth with a debt load and infrastructure bill that few companies in history have ever shouldered.
The Filing: Growth on One Side, Obligations on the Other
Anthropic's numbers are a study in extremes. In the six months ended June 30, 2026, the company reported revenue of about $16.3 billion and a net loss of $80.7 billion. The loss is dominated by accounting rather than cash: roughly $72.4 billion of it stems from a non-cash charge for the change in fair value of the convertible notes, derivative liabilities and SAFEs the company issued to fund its buildout. Strip those paper losses out, along with stock-based compensation and other non-cash items, and Anthropic posted adjusted net income of $2.7 billion over the same period.
The growth, however, is real and vertiginous. Quarterly revenue climbed from $376 million in the first three months of 2025 to $11.6 billion in the second quarter of 2026 - a 1,368% jump from a year earlier. But the prospectus also lays bare the bill for that growth: roughly $518 billion in planned cloud, compute and infrastructure obligations, split among Google (about $111 billion, with a separate agreement committing Anthropic to spend at least $114.5 billion with the Alphabet unit between April 2026 and July 2033 regardless of actual usage), Amazon (about $110 billion), Microsoft (about $31 billion) and Broadcom (about $161 billion). Around 80% of those commitments are binding and non-cancelable.
That is the tension at the heart of the filing. Anthropic is racing toward a potential $2 trillion valuation - its most recent private round, a $65 billion Series H-1 closed in May 2026, valued the company at about $965 billion - yet it is already locked into infrastructure payments that exceed its 2025 revenue of roughly $4.6 billion by more than a hundredfold. For every dollar it earned last year, it has committed roughly $113 to future compute.
Why Investment Grade Matters - and Why Anthropic Is Not There Yet
Reaching investment grade is not a vanity metric for a company in Anthropic's position. Insurance companies and pension funds - the largest buyers of corporate bonds - are generally restricted to investment-grade paper. A BBB-minus rating or better would let Anthropic borrow more cheaply and tap a far wider investor base at precisely the moment its funding needs are ballooning.
Anthropic is close, but not there. In June, Kroll Bond Rating Agency assigned the company an issuer rating of BB+, the highest speculative-grade score and one notch below investment grade. The rating came with a $35 billion chip-financing package led by Apollo Global Management, in which about $4.4 billion of junior debt - exposed directly to Anthropic's standalone credit - was sold at a yield of roughly 8.5%, while the Broadcom-guaranteed senior tranche carried an interest rate of about 5.75%. Given the limited disclosures available at the time, Kroll produced a risk score rather than a traditional full financial model.
Even that is not the end of the borrowing. Bankers have begun assembling a roughly $60 billion debt package, dubbed Project North Star, to finance Anthropic's acquisition of custom AI chips co-developed by Google and Broadcom. The structure splits into $42 billion of senior debt backed by Broadcom and $18 billion of junior debt backed only by Anthropic's own lease payments on the chips. Blackstone has been tapped for about $9 billion of the junior piece, and the company has offered lenders a double-digit yield - above 10% - on the chip debt, which would buy roughly 1.5 gigawatts of Google's custom TPU capacity.
"We may be required to seek additional financing sooner than we currently anticipate," Anthropic said in the prospectus. "Even if we determine that additional financing is required, we may not be able to raise it on terms acceptable to us or at all. Any such financing, if completed, may be costly, dilutive, or lead to restrictions on how we operate our business."
The filing says little about the path to investment grade. That silence is itself a signal: rating agencies are being asked to underwrite a business model with no public-market history, GAAP losses, and obligations that dwarf the revenue base. Bankers for Anthropic and rival OpenAI have lobbied the largest rating agencies to grant the AI labs investment-grade ratings once they go public, according to people familiar with the discussions.
The Mechanism: Convertible Notes, Fair-Value Charges, and the Cash-Flow Squeeze
To understand why Anthropic keeps coming back to the capital markets, follow the financing plumbing. The company has funded its compute buildout largely through convertible notes, SAFEs and derivative instruments rather than straight equity. Those instruments carry a quirk: as Anthropic's valuation rises, the fair value of the conversion option rises too, and the company must book a non-cash loss. That is why a company growing revenue at four-digit percentages can still report an $80.7 billion net loss.
The accounting loss does not burn cash directly. What burns cash is the infrastructure itself. Anthropic's adjusted net income of $2.7 billion in the first half of 2026 sounds healthy until set against the $518 billion commitment stack and the interest on tens of billions of debt. The senior debt is cheap only because a chipmaker - Broadcom - guarantees it; the moment a lender takes Anthropic's standalone credit, the yield jumps from 5.75% to 8.5%, and for the newest junior paper, above 10%. Each notch of additional risk has a price, and that price is rising even as revenue compounds.
This is the transmission channel behind the financing warning. Revenue growth must stay ahead of committed compute, infrastructure and personnel costs, or the company's cash-flow profile turns negative and it must return to investors - sooner than expected, on worse terms, or not at all. The prospectus frames it plainly: it is "critical that we maintain strong revenue growth, grow revenue efficiently relative to our committed compute, infrastructure, and personnel costs, and manage our cash flows effectively."
The second-order implication is what the market has not fully priced. The conventional read is that an IPO at a $2 trillion valuation solves Anthropic's funding problem. The less comfortable read is that the IPO converts private, patient capital into public, quarterly-judged capital - and that a public Anthropic will face the same growth-or-die math that has punished other capital-intensive tech stories. If revenue growth slows even as the $518 billion bill comes due, the equity may not be the escape valve investors expect; it may be the moment the leverage becomes visible to everyone at once.
The cross-asset channel runs through the broader AI trade. Growing investor unease with the debt-fueled AI investment boom has already triggered a global selloff in AI and technology stocks, with higher borrowing costs and rising bond yields adding to caution. Meta Platforms is aiming to raise up to $95 billion in its largest-ever equity fundraising, and SpaceX has inked a $40 billion deal with Apollo to lease AI chips and server hardware. Anthropic's filing is not an isolated credit story - it is a stress test of whether public markets will underwrite enormous, long-dated infrastructure commitments to back frontier-AI growth.
Cyclical or Structural: A Buildout Phase or a Permanent Capital Regime?
The central question for investors is whether Anthropic's debt dependence is a cyclical feature of a capital-intensive buildout phase - one that recedes once the infrastructure is deployed and monetized - or a structural feature of the frontier-AI business model that will not self-correct. On the evidence in the filing, this looks structural, for three reasons.
First, the obligations are non-cancelable and indexed to a compute arms race with no defined endpoint. Anthropic is not building a factory it can depreciate over a decade and then coast on; each new model generation demands more chips, more power, more data centers. The Google agreement alone runs to July 2033, with a true-up clause requiring Anthropic to pay the difference if actual spend falls short.
Second, the financing is becoming more expensive, not less, as the company scales. Junior yields have moved from 8.5% in June to above 10% for the newest paper, even as revenue compounds at a rate few companies have ever matched. In a normal cyclical buildout, credit spreads tighten as the asset base matures and cash flows stabilize. Here they are widening.
Third, the capital structure itself has become a permanent fixture. Convertible instruments now make up so much of the balance sheet that their fair-value swings dominate reported earnings - $72.4 billion of the $80.7 billion loss in the first half came from that single line item. That is not a one-time transition cost; it is the ongoing architecture of a company that funds itself with optionality rather than equity.
The cyclical counter-argument has merit, and it is the view underpinning the 8.5% to 10%+ yields lenders are accepting. Cloud hyperscalers ran heavy capital-expenditure cycles before margins expanded, and Anthropic's revenue is compounding fast enough to make today's obligations look manageable tomorrow. At a 1,368% year-over-year growth rate, a $16.3 billion half-year becomes a much larger base quickly, and enterprise adoption of Claude for coding and agentic workflows suggests demand is real, not promotional. If adjusted net income is already positive at $2.7 billion, operating leverage could turn the cash-flow story within a year or two. In this view, the debt is not a warning sign - it is the fuel that buys the market share a future public company will be valued on.
But the comparison to the hyperscalers breaks down on one dimension: Microsoft, Amazon and Google owned their infrastructure and their distribution. Anthropic is, in effect, a tenant in a digital landscape owned by the very entities that provide its capital and its chips. The counter-thesis rests on one assumption the prospectus itself undermines: that growth will not decelerate. Anthropic warned that its revenue growth rate "could decline over time and fluctuate from time to time."
The falsifying signal is concrete. Watch the ratio of annualized revenue to the scheduled principal and interest on the junior debt - roughly $18 billion carrying 8.5% to 10%+, plus the non-cancelable lease payments coming due - and to the true-up obligations under the Google and Broadcom agreements. If annualized revenue fails to stay ahead of those scheduled outflows for two consecutive quarters, the structural-debt thesis is confirmed and the equity case weakens materially. A second signal: if Kroll or another agency downgrades the BB+ issuer rating rather than upgrading it toward investment grade within 12 months of the IPO, the path to cheap capital has closed.
What Comes Next
In the short term, the IPO itself is the catalyst. A successful listing above $2 trillion would buy Anthropic time and credibility; a tepid reception, or a delay, would force the company back to the private markets it has been leaning on. Rival OpenAI is reportedly rushing its own funding round, which could value it above $1.4 trillion, timed to land before Anthropic's listing - a sign that scrutiny, and the race for capital, are both intensifying.
In the medium term, watch the quarterly revenue growth rate against the cadence of debt maturities and lease payments. That ratio, not the headline net loss, is the real health metric. In the long term, the question is whether frontier AI becomes a winner-take-most market where Anthropic's early lead compounds, or a commoditized utility where the infrastructure providers - Google, Amazon, Broadcom - capture the economics and the model makers become tenants paying rent.
Base case: the IPO clears, the $60 billion package closes, and Anthropic grows into its obligations over three to five years, with the BB+ rating gradually approaching investment grade. Upside case: Claude becomes the default enterprise AI layer, revenue compounds past the commitment stack, and the rating graduates to investment grade within two years. Downside case: growth decelerates before the infrastructure bill peaks, the junior debt trades distressed, and the company is forced into dilutive raises, asset sales or spending cuts.
The market is pricing Anthropic as if the debt is a temporary bridge to a profitable monopoly. The filing says it may be something more permanent: the rent on a future the company has already promised to pay for.
Explore more exclusive insights at nextfin.ai.
