Anthropic is considering setting aside 15% to 20% of the shares in its initial public offering for retail investors, according to people familiar with the matter. JPMorgan is expected to lead the US retail offering. The company has not confirmed the allocation, and the preliminary S-1 prospectus says only that the banks may reserve shares for their online brokerage customers.

The move would mirror recent large listings that have leaned on individual investors to support demand. SpaceX set aside about 20% of its June listing for retail investors after first planning 30%. A retail-heavy book can help banks price a deal higher because individual money is often less sensitive to price than institutional money. But it also raises questions about whether ordinary investors are being asked to absorb risk that professional buyers view more cautiously.

Key facts

• Anthropic is weighing a retail allocation of 15% to 20% of its IPO shares, with JPMorgan expected to lead the US retail offering.
• The S-1 says only that banks may set aside shares for online brokerage customers; SpaceX reserved about 20% for retail in its June listing after first planning 30%.
• Senator Elizabeth Warren sent CEO Dario Amodei a four-page letter with questions due by 3 November, covering existential AI risk, ties to Amazon and Broadcom, and a valuation that could approach $2tn.
• The filing warns that advanced AI could pose catastrophic or existential risks and notes that proposed US laws contemplate nationalisation or mandatory shutdown of frontier systems.
• Anthropic says knowledge work was worth more than $40tn globally in 2025, above SpaceX's $28.5tn market estimate.
• Compute commitments include up to $120.5bn with SpaceXAI, $44.6bn with Nscale, and $41.3bn with Lambda, plus $125bn to $150bn for TPU capacity from 2028 to 2031 outside the commitments table.
• First-half cost of revenue was $2.38bn against $16.3bn of revenue, a gross margin of about 85% before training costs; total compute spending was $8.95bn.
• Cloud marketplace sales were 48% of first-half revenue, up from 11% in 2023; Google has an option to buy up to $10bn of stock at the IPO price.
• Some shares can be sold from the first trading day, but co-founders and holders of more than 0.5% must wait until after third-quarter 2027 results or 365 days.

Retail allocation and the pricing calculation

Retail participation has become a more visible part of the IPO process as online brokerages make it easier for individual investors to place orders. Banks typically build a book by gauging demand from mutual funds, hedge funds, pension funds, and other institutions, then allocate shares before trading begins. A retail tranche gives individual investors a formal route into that process rather than leaving them to buy in the secondary market, often after the first-day pop.

A 15% to 20% set-aside would be large for a technology listing, though not unprecedented. It would give the deal a broader shareholder base and could support aftermarket trading. The trade-off is that retail demand can be fickle. Individual investors may be more likely to sell quickly if the stock falls below its offer price, and they may have less access to the detailed diligence that institutional buyers receive. Bankers also have to decide how much of the retail allocation is genuine long-term demand and how much is short-term speculation.

The final size of any retail tranche will depend on demand and the banks' bookbuilding. JPMorgan's role as lead for the US retail offering suggests that Anthropic wants a well-known Wall Street bank to manage the process rather than relying only on online brokers. The S-1 language leaves room for banks to reserve shares for their own brokerage customers, but it does not commit the company to a specific percentage. That flexibility matters because the allocation can be adjusted as the listing approaches.

Warren's questions and the political spotlight

Senator Elizabeth Warren has told Anthropic to answer four pages of questions about its IPO filing by 3 November. The Massachusetts Democrat sent the letter to chief executive Dario Amodei on Tuesday. She asked how Anthropic can ensure its technology will not become an existential threat to humanity. She also asked about its ties to Amazon and Broadcom, and how it will justify a valuation that could approach $2tn.

The letter said the risks demand greater scrutiny before they become intertwined with the public capital markets. Anthropic did not immediately respond. The deadline gives the company little time to prepare detailed answers while it is also working on the listing. Warren's intervention adds a political dimension to a transaction that is already unusual because of the scale of Anthropic's compute commitments and the safety claims attached to its business.

Her questions touch on three areas that investors will also weigh. The first is governance and safety: whether Anthropic can credibly limit the risks of advanced models while selling them widely. The second is corporate dependence: how much of Anthropic's future is tied to Amazon and Broadcom through cloud capacity, chips, and investment. The third is valuation: whether a company with rapidly changing revenue and huge capital needs can support a valuation near $2tn. Washington cannot set the IPO price, but political scrutiny can affect timing, disclosure, and demand.

Risk factors: existential danger and regulation

The filing notes concerns inside Anthropic and outside it about where advanced AI could lead. It sits alongside the warning that Anthropic could become a political target. Sufficiently advanced AI systems could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures, pose catastrophic or existential risks to humanity, the filing says.

It notes that proposed US laws contemplate the nationalisation or mandatory shutdown of frontier AI systems. The rules Anthropic itself argues for could slow its growth, it adds. Government contracts make up about 1% of its revenue. The disclosure is unusual for an IPO filing because it treats policy risk and safety risk as core business risks rather than distant regulatory footnotes. It tells investors that the company's own safety agenda could impose costs, limit product releases, or create conflicts with customers and governments.

The political risk is not only about regulation. Anthropic's models are used by enterprises and governments, and any high-profile failure or misuse could invite emergency action. The filing's reference to nationalisation or mandatory shutdown is a reminder that frontier AI is being treated in some policy circles as a strategic technology rather than an ordinary software product. That can mean more scrutiny of compute contracts, model access, and overseas customers.

The market-size claim

Anthropic says knowledge work was worth more than $40tn globally in 2025, which tops SpaceX's $28.5tn market estimate. The figure is a market-sizing claim rather than a revenue forecast. It is meant to show the size of the opportunity if AI systems take over more tasks in law, finance, software, research, and administration. Investors will test that claim against adoption rates, pricing pressure, competition, and the cost of compute. A large addressable market does not guarantee that a company can capture it at attractive margins.

The comparison with SpaceX is also a signal about how Anthropic wants to be valued. SpaceX is treated by investors as a company with a huge future market in launch and satellite services. Anthropic is arguing that knowledge work is an even larger pool. The risk is that market-size estimates can become inflated when they count all spending on a category rather than the share that could realistically move to AI tools. The filing gives investors a headline number, but the more important details are revenue growth, retention, and gross margin.

Compute deals and the cost of ambition

Anthropic has agreed to spend up to $120.5bn with SpaceXAI, the SpaceX unit formerly known as xAI. That reaches about $3.1bn a month from December. After early 2027, either side can cancel on 90 days' notice. It also agreed to spend up to $44.6bn with Nscale and about $41.3bn with Lambda. It expects to pay $125bn to $150bn for more TPU chip capacity from 2028 to 2031, outside its commitments table. Since 30 June, it has signed deals with about $245bn in minimum payments. The total is still blank. Reports have suggested Anthropic could spend $750bn on compute.

The numbers show how central compute is to the business. Training and running large models requires enormous amounts of processing power, and access to chips has become a competitive bottleneck. Long-term contracts can secure capacity, but they also create fixed obligations. The cancellation terms matter because they determine how much of the spending is truly committed. The fact that the total is blank leaves investors without a single figure for the full scale of Anthropic's future obligations.

The mix of suppliers also matters. SpaceXAI, Nscale, Lambda, Google, Amazon, Microsoft, and Broadcom all appear in the company's infrastructure story. That diversity can reduce dependence on one vendor, but it can also complicate operations and bargaining power. If Anthropic's revenue grows more slowly than expected, the compute commitments could become a drag on cash flow and margins.

Margins, revenue quality, and cloud dependence

Cost of revenue, mostly compute that serves customers, was $2.38bn in the first half, against $16.3bn of revenue. That is a gross margin of about 85%, a figure that leaves out training. Total compute spending was $8.95bn. The filing says: We do not expect our revenue will continue to grow at comparable rates, and we cannot assure you that it will continue to grow at all. Margins may also shrink as new compute arrives before the revenue it supports.

Models released in the six months to 30 September made up over 80% of third-quarter revenue. That shows how quickly Anthropic's product mix is changing, and it also means that older models may lose value quickly. Sales through Amazon, Google, and Microsoft's cloud marketplaces were 48% of first-half revenue, up from 11% in 2023. The cloud marketplaces give Anthropic distribution, but they also put important partners in a position of influence over pricing and customer relationships.

Google has an option to buy up to $10bn of stock at the IPO price. Anthropic values the convertible notes Amazon holds at $108.6bn; Amazon puts them at $97.9bn. The gap between those valuations is a reminder that the company's capital structure is complex and that strategic investors have terms that ordinary shareholders may not fully see. For IPO investors, the key question is whether the 85% gross margin is a stable measure of the business or a snapshot that excludes the cost of staying at the frontier.

Lockups and insider selling

Some shares can be sold from the first trading day. Co-founders and holders of more than 0.5% must wait until after third-quarter 2027 results, or 365 days. Most others are free after first-quarter 2027 results, or 180 days. Lockup terms matter because they determine how much stock could hit the market after the listing. A short lockup for some holders can create early selling pressure, while long lockups for founders and large holders can signal confidence but also concentrate future overhang.

Critics have said the company described in the S-1 no longer exists, pointing to the blank third-quarter figures. Some research notes have called Anthropic a much, much better business than expected but said they would not yet buy. The blank third-quarter figures and the approaching lockup expirations are likely to remain central to how investors judge the deal.

Source:TNW | Anthropic News