Anthropic eyes $64 billion IPO, but order book diverges on its AI edge

- Reading
- 5 min
- Tables
- 1
Anthropic's roadshow set a record for institutional demand, outpacing even the early days of SpaceX's offering. But behind the oversubscribed book, banks and investors still don't agree on how much the company's AI lead is worth.
Anthropic has entered the final stretch of its IPO. According to a Reuters report published by Economic Times on November 14, 2026, the company's roadshow runs through next Wednesday, when pricing for the $64 billion offering is expected to be set. The company and its banks (Goldman Sachs and Morgan Stanley among them) are meeting investors in New York, Boston, and San Francisco to gauge demand.
Institutional interest, according to two sources who spoke with Reuters, is outpacing the pace SpaceX saw in the early days of its own roadshow, used as one of the market benchmarks for offerings of this size.
The book grew fast, but that doesn't mean consensus
On the first day of the roadshow (Tuesday), Anthropic's order book was already oversubscribed two times over. By the end of Thursday, the roadshow's third day, demand had jumped to six times the number of available shares, according to one of Reuters' sources. For comparison, SpaceX closed the second day of its roadshow with the book only twice oversubscribed.
Bankers and investors tell Reuters that a single-digit subscription would be considered modest for a highly anticipated IPO, but is seen as strong given the size of Anthropic's offering. Two of the sources expect the book to grow substantially over the coming days.
That doesn't amount to certainty about the valuation. High demand on paper and conviction about the long-term thesis are different things, and that's exactly where the report shows the deepest fissure among investors putting money into the offering.
The two questions the roadshow didn't answer
According to Reuters, investors attending the meetings are grappling with two structural doubts about the investment thesis:
- Whether Anthropic can maintain its technical lead as competitors like Google (Alphabet) move quickly in developing advanced models.
- Whether Claude, currently strong mainly in coding, can sustain an edge in a broader market: knowledge workers in general.
These questions sum up the problem of valuing a company whose revenue is growing fast but whose long-term competitive position remains uncertain. Investors themselves, according to the report, still don't know whether the AI sector will end up dominated by one or two companies, or whether it will support several leaders coexisting side by side, with premium pricing becoming harder to sustain if rivals offer comparable models.
How to price a company with no direct comparable
The core metric banks use to arrive at the valuation is Anthropic's projected 2028 revenue, estimated at between $270 billion and $280 billion, applied as an enterprise-value-to-revenue multiple. The problem is that sell-side and buy-side don't even agree on the basis for that calculation:
| Projection | 2026 Revenue | 2027 Revenue | 2028 Revenue |
|---|---|---|---|
| Sell-side (banks) | ~$52-53B | ~$135-137B | ~$270-280B |
| Optimistic buy-side | 10% to 15% above sell-side | 10% to 15% above sell-side | not detailed |
| Conservative buy-side | not detailed | not detailed | ~30% below sell-side |
An investor who conducted due diligence on the offering and decided not to participate has a much more conservative view for 2028, about 30% below the sell-side estimate, according to Reuters. It's common for banks to track the company's own projections more closely, and the company in turn has an incentive to aim low ahead of the IPO, leaving room to beat expectations after going public.
The difficulty of finding direct peers shows up even in the choice of comparables: valuation models use SpaceX and cybersecurity company CrowdStrike as benchmarks, according to one of the report's sources. That's a sign there is currently no category of "publicly listed AI model company" with enough track record to serve as a yardstick.
Spotty disclosure during the roadshow
One of the investors who spoke with Reuters said Anthropic didn't give detailed answers to some questions raised during a roadshow meeting. The company didn't comment on the claim.
Another point drew attention: one of the group meetings didn't have the traditional conference call that lets institutional investors dial in and ask questions in real time, a break from the usual roadshow standard, according to someone who attended the meeting. At that meeting, Goldman Sachs and Morgan Stanley representatives limited themselves to reading prepared remarks instead of opening the floor to questions. Morgan Stanley didn't comment; Goldman Sachs didn't respond to Reuters' request.
Not every company can (or wants to) answer every investor question at a roadshow. But combined with the valuation disagreements, this level of limited disclosure is described by Reuters as a reason for caution among some investors facing a richly priced offering, at a moment of broader shakeout in the AI sector.
What this teaches anyone raising an AI round in Brazil
The Anthropic episode isn't just Wall Street news. It shows, on a massive scale, the same dilemma any AI startup founder in Brazil faces today at the negotiating table with a fund: how to prove that a current technical edge will last long enough to justify the multiple being asked.
If even sophisticated banks and investors, with direct access to Anthropic's management, can't reach a consensus on whether Claude's lead holds up against Google's competition beyond the coding niche, it's reasonable to expect Brazilian investors to apply the same skepticism to local theses built on top of wrappers or integrations with third-party models.
Three points from the report are worth reading as practical takeaways for anyone raising money here:
- Multiple on future revenue, not current margin is the yardstick the market is using to price AI right now; expecting the same in smaller rounds in Brazil isn't a stretch.
- Lack of direct comparables (to the point of using SpaceX and CrowdStrike as benchmarks for a language model company) is a reminder that the category is still being invented, including its valuation criteria.
- Disclosure and governance matter as much as the growth number: an institutional investor noticed, and publicly noted, the absence of a standard Q&A call.
The final offering price, expected next Wednesday, will be the first real test of which of these revenue views the public market is willing to pay for. Until then, Anthropic remains the most expensive and most closely watched case in the current crop of AI IPOs, and the closest barometer the Brazilian market has for calibrating valuation expectations in generative AI.
Translated from the Brazilian Portuguese original · Read the original
PyPI redesigns project pages to expose security signals for each package
The official Python package repository completed, in November, the overhaul announced in July: a dedicated security tab shows provenance, attestations, and warnings of possible compromise for each release.





