After market close on June 11, SpaceX priced the largest initial public offering in history: 555,555,555 shares at $135, raising roughly $75 billionand blowing past Saudi Aramco's 2019 record of $29.4 billion. The stock begins trading today on the Nasdaq under the ticker SPCX. Three days earlier, OpenAI confirmed it had confidentially filed a draft S-1 with the SEC, and Anthropic did the same on June 1. Three of the most-watched private companies on earth are heading for the public market inside the same quarter.
The obvious question — “should I buy?” — isn't one a quant model answers. The more useful question for anyone who scores stocks is narrower and more answerable: can the model even see these names yet? For SpaceX, the honest answer is “in principle, but it's starved of the price history three of its five factors need.” For OpenAI, it's “not yet on the board at all.” That gap — between a name that just listed and a name that has only filed paperwork — is the whole story.
What actually happened — and what didn't
It's worth being precise, because the three deals are at very different stages and only one of them is something you can act on today.
- SpaceX is public. The S-1 filed May 20 and a roadshow priced it at a fixed $135. It trades as SPCX starting today. This is a real, buyable security with a live order book.
- OpenAI has only filed.A confidential draft S-1 (announced June 8) lets a company hand financials to regulators before they're public. There is no ticker, no price, and no share you can buy. Reporting points to a possible listing window of September to November 2026 — and OpenAI has said timing isn't settled.
- Anthropic is in the same confidential-filing stage, having submitted its draft S-1 on June 1.
On the numbers, the contrast is just as sharp. SpaceX's prospectus reported $18 billion in 2025 consolidated revenue against a $4.9 billion net loss, with $6.58 billion of adjusted EBITDA. The consolidated figure folds in two very different businesses: a profitable Starlink segment ($11.4 billion revenue, $4.4 billion operating profit) and a consolidated xAI segment that lost about $6.4 billion at the operating line on $3.2 billion of revenue. Founder voting control sits near 85%.
OpenAI's disclosed figures are run-rate, not GAAP full-year revenue: annual recurring revenue crossed roughly $20 billion in 2025 (up from about $6 billion in 2024), while internal projections widely reported in late 2025 point to a loss on the order of $14 billion in 2026 and cumulative losses well into the tens of billions before any profitability. It was last valued around $852 billion in a March 2026 round; the trillion-dollar figures attached to its IPO are analyst speculation, not a printed price.
Why a fresh IPO is the hardest case for a factor model
The QScoreis a weighted blend of five factor categories. Each one needs a specific kind of input. A company that started trading this morning, or hasn't started at all, simply doesn't supply most of them — and the model is built to say so rather than guess.
| Factor | What it needs | What a day-one IPO gives it |
|---|---|---|
| Momentum | 12-, 3-, and 1-month trailing returns, RSI(14), 50- vs 200-day moving average | One day of prints — every input is undefined for months |
| Risk | Beta from ~5 years of returns vs the market; 60-day realized volatility | No regression history and no 60-day window — undefined |
| Value | P/E, P/B, P/S, EV/EBITDA, z-scored against sector | Computable, but distorted — a net loss makes trailing P/E negative |
| Growth | Year-over-year revenue, EPS, and free-cash-flow growth from public filings | Real and often strong — but clean sector-comparable history is thin |
| Profitability | ROE, ROA, gross/operating/net margin, FCF yield | Computable once filed — and for cash-burning IPOs, scores low |
Momentum is the cleanest example of the problem. It blends 12-month, 3-month, and 1-month trailing returns with RSI(14) and the 50-day versus 200-day moving-average position. SPCX has exactly one session of price data. You cannot compute a 200-day average from one day, and RSI needs a fortnight of up-and-down days before it means anything. This factor is mathematically undefined for a new listing and stays noisy for months after.
Risk is in the same position for the same reason. Beta is the slope of a stock's returns regressed against the market over years of history; 60-day realized volatility needs sixty trading days. On day one there is no regression to run and no window to measure. A factor model that reported a beta for SPCX this week would be inventing one.
Value is computable, but the inputs misbehave. SpaceX's $4.9 billion net loss makes its trailing P/Enegative — a number that's mathematically real and practically useless. QScoring deliberately assigns negative-P/E names a fixed low value score rather than ranking them as “infinitely cheap.” On EV/EBITDA, $6.58 billion of adjusted EBITDA against a multi-hundred-billion-dollar enterprise value is a rich multiple by any sector standard. For OpenAI, a possible $1 trillion valuation on roughly $20 billion of ARR would imply a price-to-sales ratio in territory the public market has rarely sustained.
Growth is the one factor where these names look strong, not starved. Starlink is scaling fast; OpenAI's ARR more than tripled in a year. But the factor needs year-over-year comparisons from public filings that are sector-normalized, and a company whose first detailed financials arrive with its IPO doesn't hand you a clean, comparable multi-year series on day one. The narrative is excellent; the model-ready history is thin.
Profitability is computable the moment financials are filed — and it's exactly where the cash-burn shows. SpaceX's consolidated net loss (the xAI segment's ~$6.4 billion operating loss swamping Starlink's $4.4 billion operating profit) and OpenAI's spend — burning a large share of revenue with multi-billion losses projected — both map to weak profitability scores. This factor works fine; it just doesn't flatter a company still buying growth with losses.
Two of the five factors are mathematically undefined for a brand-new listing. The other three are computable but either distorted, comparable-starved, or unflattering. That is a low-confidence score by construction — and the model is supposed to say so.
That last point is the honest hook. QScoring attaches a confidence ratingto every score precisely because data completeness varies. A name missing its two price-based factors and carrying distorted value inputs is the textbook case for LOW confidence. The responsible output for SPCX this week isn't a crisp composite — it's “not enough data yet.”
The traps that have nothing to do with the model
Even setting the factors aside, freshly-public mega-caps carry structural quirks a single score won't capture:
- Lockups.SpaceX's prospectus points to 90- to 180-day lockups. When they lift — plausibly around December 2026 — early employees, early investors, and the bank syndicate can all sell at once. A wave of supply hitting at a known date is a price event no fundamental factor predicts.
- Governance.Founder voting control near 85% means public shareholders are buying economic exposure with little say. Markets often apply a discount for that, and it isn't in any of the five factors.
- Index-inclusion lag.The forced buying from S&P 500 or Russell inclusion arrives aftera company clears eligibility seasoning, not on listing day. The “momentum” from index demand shows up months later — another reason day-one price action is a poor signal.
- Float and concentration. A small public float relative to a huge valuation makes early prints volatile and easy to misread as trend.
What the disciplined quant actually does
The temptation around a $75 billion headline is to treat the size of the deal as if it were a signal. It isn't. The disciplined move is to wait for each factor to earn its input, and to know roughly when that happens:
- Value and profitability turn on first— at the first public quarter, once filed financials exist. Read them knowing the value inputs are distorted by losses and the profitability inputs are honest but unflattering.
- Realized volatility needs ~60 trading days; a stable beta needs years. Until then the risk factor is a placeholder, not a read.
- Momentum needs 3 to 12 monthsof trading before its blend of trailing returns and RSI says anything trustworthy — and the first few months are contaminated by lockups and index mechanics anyway.
- Let confidence gate the verdict. Incomplete data means LOW confidence, and a LOW-confidence score is a reason to wait, not to act. We'd rather print “not yet” than a composite that's mostly narrative.
None of this is a view on whether SpaceX or OpenAI are good investments. It's the opposite: it's the model admitting what it can't see. A factor score is only as good as the history feeding it, and history is the one thing an IPO can't fast-forward. When SPCX has a few quarters of filings and a couple hundred trading days behind it, the score will mean something. Today it would mostly be a guess wearing a number — and the whole point of scoring stocks instead of arguing about them is to not do that.
Sources
- NPR. SpaceX blasts off with a record-breaking $75 billion IPO — pricing of 555,555,555 shares at $135, the largest IPO on record, ahead of Aramco's 2019 listing.
- Fortune. SpaceX finally files IPO prospectus, reveals revenue is up — but losses are too — $18B 2025 consolidated revenue, $4.9B net loss, and the Starlink / xAI segment split.
- CNBC. OpenAI confidentially files for IPO, prepping Wall Street for mega AI debut — confidential draft S-1, the Sept–Nov 2026 window, and the unsettled timing.
- Fortune. OpenAI plans to report stunning annual losses through 2028 — ARR scale and the multi-year loss projections behind the cash-burn picture.
- Anthropic. Anthropic confidentially submits draft S-1 to the SEC — the June 1, 2026 confidential filing, third of the three AI listings.
Related reads
- How to read a QScore— what each of the five factors actually measures under the hood
- Beta explained— why a stable beta needs years of returns, not days
- P/E ratio explained— why a negative P/E isn't “cheap”
- QScoring methodology— how the composite is built and how confidence is assigned
This article is for informational purposes only and is not investment advice. QScoring does not have a position in, and does not cover, any of the securities mentioned.
Discussion
Comments are powered by GitHub Discussions. Sign in with GitHub to join the conversation.