Anthropic Could Go Public in Weeks. Here Are the 5 Numbers That Will Actually Matter.

Anthropic, which confidentially filed for an IPO on June 1, could publicly release its S-1 within weeks, potentially supporting a valuation near $2 trillion. The filing is expected to clarify reported annualized revenue growth from $47 billion in mid-May to $65 billion at July-end, including roughly $11.5 billion of Q2 booked revenue, as well as gross-margin trends. Investors will also scrutinize more than $500 billion of reported decade-long cloud, chip and data-center commitments, related-party revenue from strategic backers, and export-control risks after a June suspension of its two most powerful models.

Analysis

The filing is primarily a read-through on whether hyperscaler AI capex is supported by arm’s-length utilization or by capital recycling among strategic investors. If a material portion of revenue is sourced from AMZN, GOOG, MSFT, or NVDA-linked arrangements, the market should discount it as lower-durability demand and reassess how much current cloud and accelerator spend reflects incremental end-customer workloads. Conversely, non-cancellable commitments with credible third-party customer concentration would improve revenue visibility for AMZN and MSFT, while reducing the perceived risk that their AI infrastructure build is speculative.

The key underwriting ratio is not reported revenue alone but contracted compute obligations relative to gross profit and cash generation. Large fixed minimums can produce impressive supplier backlog while creating severe operating leverage for the model provider: a slowdown in enterprise inference demand would force price cuts or capacity resale, pressuring both gross margin and the value of unused GPU capacity. For NVDA, this is a mixed signal—commitments validate near-term demand, but disclosures showing cloud spend is largely pass-through would reinforce the market’s concern that accelerator orders have been pulled forward rather than supported by profitable inference economics.

Near-term, the event is unlikely to move diversified strategic holders on ownership marks; the more tradable catalyst is a disclosure that changes expectations for 2027-28 hyperscaler capex returns. Export-control exposure is a separate downside tail: model-access interruptions can quickly impair enterprise retention and make long-duration capacity contracts less financeable. A bullish cloud-supplier thesis is falsified if contractual obligations are cancellable, related-party revenue dominates, or gross margin is declining despite scale.

Key Decisions for Investors

• Do not initiate a pre-filing directional trade in AMZN, GOOG, MSFT, or NVDA; the public S-1 is an information event, but diversified parents have insufficient direct earnings sensitivity for clean event risk/reward.
• Set a filing-day alert for related-party revenue above 50% of total revenue or a disclosed concentration with strategic investors. Either outcome supports reducing AI-infrastructure beta, especially NVDA, because it would weaken the independent-demand argument behind supplier backlog.
• Conditional 1-3 month pair: long AMZN and MSFT equally / short NVDA only if the filing shows substantial non-cancellable cloud obligations, stable or improving gross margin, and predominantly third-party revenue. The thesis is that contracted cloud consumption improves visibility for infrastructure providers while NVDA retains greater pull-forward and valuation risk; exit if obligations are materially cancellable or gross margin deteriorates.
• Track contractual commitments divided by annualized gross profit once disclosed. A rapidly rising ratio without a matching increase in operating cash flow is a warning of capacity overbuild and should be treated as a negative read-through for AI capex beneficiaries rather than a validation of demand.