Anthropic filed its S-1 on October 26, 2026. The headline numbers are remarkable: $16.3 billion in first-half 2026 revenue, up approximately 1,300% from the same period a year earlier. Bulls will cite the 900% net revenue retention. Bears will go straight to page two of the risk factors.

The liability stack is what demands attention now, before the IPO roadshow resets the conversation.

The Debt That Doesn’t Show Up in the Headline

As of June 30, 2026, Anthropic carried $6.2 billion in total indebtedness, mainly tied to convertible promissory notes issued to Amazon (through AWS) starting in 2023. That figure looks manageable against trillion-dollar IPO speculation. It is not the number that matters.

Two separate buckets matter more for sizing the fixed-cost glide path. First, Anthropic has disclosed approximately $417 billion of aggregate future minimum payment obligations under new and amended compute and infrastructure agreements entered into after December 31, 2025, with contracts commencing between 2026 and 2027 and terms ranging from less than one year to 20 years. Second, it has disclosed equipment lease arrangements with total rent payments of approximately $161.2 billion excluding interest, with obligations typically commencing when equipment becomes “ready for service.”

In order to operate and grow its business, including to fund its substantial compute commitments, Anthropic has incurred a substantial amount of debt requiring significant interest and principal payments. The S-1 goes further: the business may not be able to generate cash flows from operations sufficient to service its debt and make necessary capital expenditures. That is the company’s own language, not a short-seller’s.

Where the Growth Friction Begins

Compute and infrastructure expenses are Anthropic’s number one operating expense, rising to $5.6 billion for the three months ended June 30, 2026. At that pace, annualized compute and infrastructure expense runs to roughly $22 billion before any capacity expansion. The company explicitly frames these expenses as near-term fixed because it typically procures compute 12 to 18 months in advance of availability.

That rigidity is the structural problem. Compute commitments are soaring while competition is picking up from lower-cost open-weight models. If pricing compression accelerates, Anthropic cannot easily walk away from multi-year, largely noncancelable obligations. The fixed-cost base expands regardless of what happens to revenue per query.

The pricing pressure is already visible in Anthropic’s own product decisions. Anthropic’s 40% price cut on Opus 5.5 shows how aggressively the company is competing on unit economics even as its fixed cost base expands.

The Credit Signal Traders Should Watch

Public, widely followed bond ratings from S&P and Moody’s are not yet the clean signal here, because the company is not a seasoned public issuer with long-cycling debt markets transparency. More actionable for traders is to treat any disclosed private-credit terms as the real-time credit thermometer. In that context, the Financial Times has reported Anthropic was offering Blackstone a double-digit yield on a $60 billion AI chip financing. A double-digit yield at that size is not the financing profile of a company the market considers risk-free.

Three Scenarios for the IPO Window

Bull Case: The third-quarter revenue growth rate of 45% holds into Q4. Operating leverage shows up as fixed compute and infrastructure expense is spread across a larger revenue base. A $2 trillion IPO prices cleanly, refinances the near-term debt stack, and the compute commitments become the moat rather than the liability.

Base Case: Investors settle on a $1.5 trillion IPO valuation rather than $2 trillion. That still funds operations but leaves a narrower buffer against any slowdown in enterprise Claude adoption, given the scale of fixed and minimum payment obligations.

Bear Case: Revenue growth decelerates toward 20% quarter over quarter as open-weight competition intensifies. At that pace, the $80.7 billion net loss in the first half of 2026 becomes structurally harder to absorb. Debt service consumes an increasing share of operating cash flow, constraining the R&D spend required to stay ahead of the model curve.

Valuation compression is not unique to Anthropic’s bear case. OpenAI’s pre-money ask dropped roughly $150 billion in weeks as investors pushed back on AI private-market valuations.

What Active Traders Need to Monitor

The IPO pricing window is the first hard catalyst. Watch the yield Anthropic ultimately pays on any large private-credit tranche: any move above 10.5% signals that institutional credit desks are pricing more risk than the equity roadshow acknowledges. Monitor the spread between Anthropic’s implied valuation at IPO and the May 28, 2026 Series H post-money valuation of $965 billion. Compression is the tell that the market is discounting the liability overhang. Position sizing into any post-IPO lock-up expiry should reflect the reality that compute commitments, not model performance, set the floor on cash consumption.

Revenue growth is real. So is the debt. The trade is in understanding which one the market is actually pricing at any given moment.