The most important document in technology this week has no cover page. Anthropic will hand updated financials to a group of institutional investors gathering at its San Francisco headquarters, according to people familiar with the matter, a step that comes before the formal registration statement the company must file to list.
CNBC reported the meeting on October 13. The gathering is set for October 14 and functions as the first measurement of whether the largest private company in artificial intelligence can price an offering that bankers have discussed in terms of trillions.
The numbers investors already have are the context. Anthropic told investors in August that its annualized revenue run rate reached $65 billion at the end of July, roughly seven times the level a year earlier. Revenue grew 142% from the first quarter to the second. The company has projected 2028 revenue of between $190 billion and $200 billion, according to two people familiar with the figures.
What this week’s update adds is granularity. The Financial Times reported that Anthropic has told investors it will be profitable for a second consecutive quarter on gross margins above 80%, while having yet to post a quarterly profit on net income. Both statements can be true at once, and the distance between them is the question a public-market investor will spend the most time on.
A gross margin above 80% is a software number attached to a business that rents accelerators by the gigawatt. The figure depends heavily on how inference costs get categorized, on whether depreciation on committed compute sits above or below the line, and on how much of the revenue base comes from high-margin API calls rather than from enterprise contracts with service attachments.
Valuation talk has ranged from $1.8 trillion to $2 trillion, with some investors discussing figures as high as $2.3 trillion. Bloomberg has reported that bankers could open the roadshow in the week of November 9 with the aim of listing before Thanksgiving. None of that is possible until the audit closes and the filing is public, and companies have slipped schedules before.
The reason the financials matter beyond one company is that AI laboratory economics have become the pricing variable for the technology trade. For three years investors valued the sector on capacity announcements and model benchmarks. Both are inputs. Revenue, margin and cash consumption are results, and a public market discounts results.
The comparison that will frame the offering is with OpenAI, which has said it now targets 2027 for a listing. That leaves Anthropic as the first frontier laboratory to test public demand for the sector, and the first to publish audited financials for a business whose largest cost is compute. However the pricing lands, it sets a reference the next filing will be measured against.
A registration statement forces four things into the open that private rounds allowed the company to summarize: revenue concentration, the terms of its compute commitments, the cost of serving a token, and the cash it burns against what it earns. Anthropic has signed cloud and data center agreements that analysts have valued in the hundreds of billions of dollars, and the obligations attached to them move onto a document that carries legal weight.
Anthropic has spent the past week adding contracts that make its revenue look less like a consumer trend and more like enterprise software. On October 12 the company signed a global premier partnership with HCLTech, under which the Indian services firm will build a dedicated Anthropic business unit and a center of excellence, and its software arm will embed Claude and Claude Code into products including BigFix and AppScan. The company also put Claude inside Workday’s human resources and finance workflows.
Both deals point at the same asset. Enterprise distribution is what turns a fast-growing revenue line into a durable one, because seat-based contracts renew while consumer chat usage is spiky in ways that are hard to forecast a quarter ahead. A services partner that deploys the model into customer systems also absorbs the integration cost that slows adoption.
The risk sits in the concentration that comes with the strategy. A growing share of revenue from a small number of large enterprise and cloud relationships is a disclosure item a filing will have to quantify, along with the compute commitments Anthropic has signed and the obligations those create.
That is the part of the story the private market has been able to skip. Investors buying into the last rounds were told a run rate and a growth multiple. Buyers of a public offering will be told how the money gets spent, what it costs to earn, and what happens if demand slows.
A company that has reached a $65 billion run rate in under four years has an argument the skeptics have to answer. It also has a cost structure, a set of long-term leases and a customer list that no private round required it to explain. The filing will supply all three at once.
For now the only number on offer is the one handed out this week behind closed doors, to investors who will decide what it implies about a company that still has to file.
