Anthropic IPO: A Risk Perspective

On November 10, Anthropic set an expected price range of $70 to $78 per share in an amended S-1 filing for its initial public offering (IPO), valuing the company at roughly $1 trillion. It would be the second mega-cap IPO of 2026, after SpaceX began trading June 12.

IPOs of this size raise practical questions for portfolio and risk managers before a single share has traded. When does a risk model see the new stock, and what does it see? A company with no trading history has no measurable volatility, beta, or momentum. Yet the stock can enter portfolios on its first trading day and some benchmarks within weeks.

In this article, we use SpaceX’s first five months in the FactSet Equity Risk Model — U.S. (Daily) as a guide to what investors could expect after Anthropic’s IPO. The model comes in two versions: Short-Horizon (SH), tuned for one day to two months, and Mid-Horizon (MH), tuned for one to six months. We discuss three questions and their implications for investor portfolios and for Anthropic:

• Is the new stock covered on its first trading day?

• What does the model know about it before it has a trading history?

• How does the model’s picture change as real trading data arrives? 

Will Anthropic Be Covered on Day One?

Our daily equity risk models are fundamental factor models. A stock’s risk is described by its exposures to a market factor, industry factors (and country factors in multi-country models), a set of style factors, and a stock-specific component. To produce a risk profile for a security, the model chiefly needs a small set of inputs:

• Price and a share count, which give its market capitalization and Size exposure

• Industry classification

• Country of risk

Where an exposure cannot yet be computed from the stock’s own data, it is either estimated from industry peers or set to a placeholder value until the stock’s own data arrives. Stock-specific risk is estimated from comparable companies of similar size in the same industry until enough return history exists.

SpaceX’s IPO shows the mechanism at work. SpaceX appeared in both the SH and MH versions of the U.S. Daily model in the update for its first trading day, June 12, and in the FactSet Equity Risk Model — Global (Daily) with very similar exposures. A portfolio holding SpaceX at the close of that first session could be analyzed with factor exposures and a risk forecast from day one.

Coverage is distinct from membership in the model’s estimation universe: the list of stocks whose returns are used to estimate the factor returns. The estimation universe is deliberately selective. It admits only primary listings of common shares and screens them for size, trading frequency, turnover, and extreme returns so that the factor returns are driven by liquid, representative stocks.

In addition, a stock needs a valid value for every exposure. A company of SpaceX’s size clears the size screen immediately. Its entry into the estimation universe is therefore set by trading history: it joins once every exposure can be computed from its own data. For SpaceX, that took until July 15 in our MH version and until September 7 in our SH version, with identical dates in the Global Daily model (Exhibit 1).

The difference between the two horizons is by design. The MH version uses a Momentum factor that skips the most recent month of returns, so it can be computed after roughly one month of trading. The SH version uses a Long Momentum factor that skips the most recent three months and pairs it with a separate Mean Reversal factor for short-term moves. Long Momentum therefore needs about three months of history before it reflects the stock itself.

What Does the Model Know on Day One?

It is tempting to assume a stock with no trading history is a blank slate to a risk model. It is not. A fundamental factor model draws on everything that is known about the company, and a company that files an S-1 ahead of its IPO has already published several years of financial statements.

On June 12, seven of SpaceX’s style exposures were computed from its own market and fundamental data, not borrowed from peers: Size, Profitability, Earnings Yield, Book to Price, Growth, Leverage, and Dividend Yield.

Exhibit 2 shows those day-one style exposures in our U.S. Daily model (SH) alongside two natural reference groups: the Telecommunication Services companies in the estimation universe (SpaceX’s industry under its GICS classification) and the 10 largest U.S. stocks, each as a cap-weighted average.

Two aspects stand out.

• SpaceX looked nothing like a typical telecom. Its industry peers in the estimation universe were smaller, more leveraged, cheaper on earnings and book value, and paid far higher dividends. SpaceX had a Leverage exposure close to zero and negative Earnings Yield and Book to Price.

• SpaceX looked nothing like a typical mega-cap. Its Size exposure was at the model’s top level, alongside the largest U.S. stocks. Yet its Profitability exposure was about two standard deviations below the market average, placing it in the bottom decile of the estimation universe on its first trading day. The largest U.S. companies, by contrast, scored clearly above average on Profitability.

This matters for portfolio construction. An investor who proxied the new holding with an industry ETF or a basket of mega-caps would have assigned it the wrong style exposures on both counts. The factor model captured, from day one, a company that was mega-cap in size, far less profitable than the average stock, and expensive relative to its fundamentals.

How Does Adding a Mega-Cap IPO Shift Factor Exposures?

Exposures matter because they add up. A portfolio’s factor exposures are the weighted sums of its holdings’ exposures, so a new position shifts them from the day the position is bought. Exhibit 3 takes the two reference groups from Exhibit 2 as cap-weighted baskets and adds a 5% SpaceX position bought at the close on June 12.

The shifts differed between the two baskets:

• Mega-cap basket: Its Size exposure was unchanged, so the position looked like a natural fit by size. Yet it removed close to a fifth of the basket’s Profitability tilt.

• Telecom basket: The position diluted the basket’s defining tilts. Dividend Yield, Earnings Yield, Book to Price, and Leverage all fell.

These are the questions a risk model enables an investor to answer from the first trading day, and they will apply equally to Anthropic:

• Does the new name add the factor exposure I intended?

• Does it dilute a tilt I am paid to hold?

• How much of my tracking error will come from one stock whose own history is still being written?

How Fast Does the Picture Sharpen?

The exposures that depend on trading history—Volatility, Beta, Liquidity, and Momentum—cannot be measured on day one. Until enough history exists, they carry placeholder or peer-based values, and each switches to the stock’s own data as soon as enough of that data is available (Exhibit 1). Exhibit 4 tracks three of the exposures in the SH version of our model from the day each became data-driven.

The picture tracks the stock’s trading record. The shares swung sharply after the IPO and ended October 33% below their June high.

• Volatility: Once computed from SpaceX’s own returns, the Volatility exposure was above the market average from the outset. It rose further in July and August, a period in which the shares moved more than 5% on 10 of 43 trading days. From July onward, SpaceX ranked among the 10 most volatile of the 50 largest stocks in the model’s estimation universe on all but two days.

• Liquidity: The exposure was persistently positive, reflecting heavy trading in the newly listed shares.

• Long Momentum: The exposure turned increasingly negative after it became data-driven in September. By the end of October it was about two standard deviations below the market average, as the look-back window filled with the stock’s retreat from its June highs.

For a portfolio manager, each of these changes carries through directly to the stock’s factor risk, its beta to the market, and its contribution to a portfolio’s tracking error. Each is visible in the daily model updates as the data arrives.

A fundamental model has a natural limit in the months after an IPO. Until the stock has several quarters of return history, its stock-specific risk is estimated from comparable established companies, and such estimates cannot anticipate the extra turbulence that often accompanies price discovery. Investors holding large positions in a recent IPO should treat early risk estimates as a starting point and monitor them closely as the data builds.

What to Expect After Anthropic’s IPO

Anthropic’s S-1 contains audited financial statements, so the same mechanism should apply from its first day of trading. Subject to the usual data availability, we would expect the following:

• Day-one coverage in the U.S. Daily and Global Daily models, with Size, Profitability, Leverage, and the other style exposures that need no trading history computed from its own market and fundamental data.

• Industry factor exposure determined by the GICS classification assigned at listing. Its S-1 registration lists a software industry code, but GICS assignments are made independently. Where Anthropic lands will shape both its industry factor exposure and the peer group used for its early risk estimates.

• Market-based exposures (Volatility, Beta, Liquidity, Momentum) switching from placeholder and peer-based values to Anthropic’s own trading data over its first weeks and months.

• Entry into the estimation universe after roughly one month in the MH version of our model and after three months in the SH version, once all its exposures can be computed from its own data, provided it passes the size and liquidity screens. Until then, Anthropic’s returns will not influence the estimated factor returns used for every other stock.

Conclusion

Mega-cap IPOs present risk managers with a stock that can move a portfolio before it has a history. SpaceX’s IPO shows how a daily fundamental factor model handles that situation:

• Coverage from the first trading day.

• Day-one style exposures computed from the company’s own market and fundamental data, distinct from both its industry and its size peers.

• A transparent sequence in which market-based exposures switch to the stock’s own trading data and the stock then joins the estimation universe.

No model can know on day one how a newly public stock will trade. But knowing what the model knows, and when, lets investors use it with appropriate confidence as Anthropic and the next generation of mega-cap listings come to market.

Learn More

The FactSet Equity Risk Model — U.S. (Daily) and FactSet Equity Risk Model — Global (Daily) are available in SH and MH versions within FactSet Portfolio Analysis and as data feeds.  

This blog post is for informational purposes only. The information contained in this blog post is not legal, tax, or investment advice. FactSet does not endorse or recommend any investments and assumes no liability for any consequence relating directly or indirectly to any action or inaction taken based on the information contained in this article.