What the Offering Size Leaves Out: Reading Q2 2026 US IPOs From the Prospectus

An FMP Signals Lab Case Study

Methodology version 1.0 | Quarter: Q2 2026 | Data snapshot: 19 August 2026 | Benchmark table version 1.0

Offering size is one of the first figures attached to an IPO, but it does not show how the offering was constructed. It does not tell an analyst how much the underwriters charged, how much of the company was sold, whether existing shareholders participated, or which share class carried the economic and voting rights.

The IPO Offering Structure methodology standardizes those terms by defining the universe, measures, retrieval process, and validation requirements used in each quarterly study. This case study applies version 1.0 to Q2 2026 and reports only the findings that the available prospectuses support.

Executive Summary

Q2 2026 included offerings ranging from Space Exploration Technologies at roughly $75 billion to deals of less than $10 million. Cerebras Systems raised $5.55 billion, Fervo Energy raised $1.89 billion in its base offering, and dozens of smaller companies also came to market. Applying the IPO Offering Structure methodology to the quarter produces three findings.

First, the validated offerings were overwhelmingly companies raising capital rather than existing holders selling. Across the offerings whose documents could be retrieved and validated, exactly one included shares sold by existing holders in the base offering, and that component was 2.12% of the deal.

Second, automated screening was not sufficient to define the universe. Six of the 31 retrieved prospectuses did not describe conventional IPOs, and each was identified by reading a single sentence on the cover page. That is roughly one in five of the documents reviewed.

Third, the structure of an offering was frequently different from what its headline figures suggested. Fervo Energy's reported base offering was 13% below the amount the company ultimately raised because the underwriters exercised their option in full. Its founders also held approximately 2.75% of the capital stock while controlling approximately 53.03% of the voting power. Neither fact was visible in the headline offering size or share count.

This study reports what the filings disclose. It does not rank the offerings, score them, or comment on how any of these securities has performed or will perform.

What the Quarter Contained

The FMP IPOs Calendar API listed 54 priced offerings on major US exchanges during Q2 2026 that passed the methodology's initial screens for operating companies with common-stock tickers. They ranged from a single offering of roughly $75 billion to deals of less than $10 million.

The recognizable names spanned that range. Cerebras Systems, the AI chip designer, raised $5.55 billion. Fervo Energy, a geothermal developer, raised $1.89 billion in its base offering. X-Energy, Quantinuum, Pershing Square, HawkEye 360, Seaport Therapeutics, and Kardigan all listed in the same three months, as did Space Exploration Technologies. Not all of them could be examined for this study, and the coverage section below sets out which offerings could be measured and which could not.

Scale alone says little about how these offerings were constructed. The structure is disclosed in the prospectus rather than in the headline, so the analysis began by establishing which of the 54 offerings were conventional IPOs at all.

Establishing What Counted

The methodology defines its universe by rule: offerings marked priced in the calendar, listed on a major US exchange, issued by an operating company rather than a blank-check vehicle or fund, and carrying a common-stock ticker rather than a unit, warrant, or rights tranche. Two of those screens use structured fields. Two use documented heuristics applied to the company name and ticker form.

Applying those screens to Q2 2026 admitted 54 offerings. Reading the covers of the 31 retrieved prospectuses disqualified six of them. Each was identified by a single sentence.

Offerings admitted by the screens and disqualified on reading

Offering	What the cover states	Why it fails the rule
DUKE Robotics	Relates to the resale by the Selling Stockholders of up to 170,954 shares	A resale registration. No company is raising capital.
AIAI Holdings	The Registered Stockholders are offering to sell their Shares	A resale registration by existing holders.
AMASS Brands	The Registered Stockholders are offering to sell shares of their Common Stock	A resale registration by existing holders.
Cantor Equity Partners VII	The underwriters are offering the Class A ordinary shares on a firm commitment basis	A blank-check vehicle. The name carries no term the screen looks for.
Exyn Technologies	We are offering 2,500,000 units, each unit consisting of one share	A unit offering, which the rule excludes.
Eloxx Pharmaceuticals	Eloxx Pharmaceuticals, Inc. is offering 2,975,000 shares of common stock	A registered offering by a company already listed, not an IPO.

Six of 31 is close to one in five. The failures are not random. A resale registration and an IPO can both appear in a calendar as priced offerings of common stock on a major exchange, and the difference between them may be a single sentence. A blank-check vehicle named Cantor Equity Partners contains none of the words a name-based screen looks for, while a unit offering can still carry a four-letter ticker.

Another 23 offerings could not be examined because no final prospectus could be matched to them. One of those companies has a profile that classifies it as a shell company and is therefore a probable further exclusion. Without a prospectus, however, it remains in the eligible universe. Counting every unexamined offering as a conventional IPO produces the most conservative denominator, and that is the basis used throughout this study.

Coverage

The methodology requires each study to state what it could measure. For Q2 2026, coverage was as follows:

Coverage for Q2 2026

	Offerings
Admitted by the universe screens	54
Disqualified on reading the cover	6
Eligible on the most conservative basis	48
Final prospectus retrieved and validated	25
Structured offering economics available	7

The eligible figure counts every offering whose document could not be retrieved as a conventional IPO. Any further exclusions would reduce the denominator and raise coverage.

Two consequences follow, and the methodology's own rules determine both. A gross spread was validated against the filing for four offerings. That is far below the level at which a quarter-wide comparison may be reported, so this study makes no claim about spreads across Q2 2026 as a whole. The four observations are reported individually, and nothing is aggregated from them.

The split between company shares and existing-shareholder shares was validated for 25 of 48 offerings, or 52.1%. That clears the methodology's 50% coverage threshold, so a quarter-level statement about that measure can be reported. The figure is a floor rather than an estimate because it assumes every unexamined offering is eligible.

Space Exploration Technologies: Reading Scale and Structure Together

The final prospectus for Space Exploration Technologies reports an offering of 555,555,555 Class A shares at $135.00 per share, producing gross proceeds of $74,999,999,925. Every share in the base offering was sold by the company.

Measure	Value
Shares offered	555,555,555 Class A
Offering price	$135.00
Gross proceeds	$74,999,999,925
Underwriting discount	$0.90 per share
Gross spread	0.67%
Retained fraction	99.33%
Secondary component	none
Offered-share ratio, Class A denominator	7.53% of 7,380,196,910 shares
Offered-share ratio, all-share denominator	4.25% of 13,075,865,175 shares
Underwriters named on the cover	22

The economics are straightforward, but their interpretation is not. The 0.67% gross spread was the lowest among the four spreads validated for this study. It was also attached to an offering of extraordinary scale. A spread that low is notable, but it cannot be treated as a score for execution because the benchmark table did not contain enough comparable offerings in the relevant size band to establish a valid deviation.

The share structure shows why a float figure means little until its denominator is named. The shares offered represented 7.53% of the Class A shares outstanding after the offering but only 4.25% of all shares outstanding. Neither ratio is more correct than the other. One measures the offering against the listed class, while the other measures it against the company's full economic share base.

The underwriting syndicate provides a separate validation lesson. The prospectus cover names 22 underwriters, while the automated name scan recognized 20. The scan did not identify two institutions that were absent from its reference list. That difference is why the methodology treats automated extraction as a way to produce candidates and takes the final syndicate count from the filing itself.

Space Exploration Technologies is the clearest illustration in the quarter of how spread, offering size, share class, and syndicate structure have to be read together. The headline amount describes the scale of the transaction. It does not describe how much of the company was sold or how the offering was organized.

Where the Proceeds Went

Across the 25 validated offerings, exactly one included shares sold by existing holders in the base offering. Lincoln International sold 20,604,046 shares on its own account alongside 445,942 shares sold by selling stockholders, a secondary component of 2.12%. In every other validated offering, the company was selling all of the shares.

The result requires one clarification because the language in the filings invites a mistake. Selling-shareholder wording appears on many covers where no existing holder sells anything in the base offering. Existing holders frequently grant the underwriters the over-allotment option without participating in the base deal, which produces a secondary ratio of zero under this methodology and a prospectus that nonetheless discusses selling stockholders at length. Detecting the phrase is not the same as measuring the split, which is why the measure is read from the offering-share table.

The finding is narrow and should be read precisely. In the offerings this study could validate, the proceeds went overwhelmingly to the issuers. It says nothing about how many shares existing holders may sell once lock-up periods lapse, nothing about the 23 offerings that could not be examined, and nothing about quarters before or after this one.

Cerebras Systems: One Offering, Two Very Different Denominators

According to Cerebras Systems' final prospectus, the company priced its IPO on May 14, 2026, at $185.00 per share and offered 30,000,000 shares of Class A common stock for gross proceeds of $5.55 billion.

Cerebras Systems, offering terms

Measure	Value
Shares offered	30,000,000 Class A
Offering price	$185.00
Gross proceeds	$5,550,000,000
Underwriting discount	$4.3475 per share, $130,425,000
Gross spread	2.35%
Retained fraction	97.65%
Secondary component	none
Offered-share ratio, Class A denominator	100.00% of 30,000,000 shares
Offered-share ratio, all-share denominator	13.95% of 215,110,345 shares

The share structure is where the methodology's denominator rule earns its place. Cerebras has three classes of common stock. Class A carries one vote per share, Class B carries 20 votes per share and converts into Class A, and Class N is non-voting and also converts into Class A. No Class N shares were outstanding immediately after the base offering, but the Class A and Class B counts still produced two very different views of the transaction.

The 30 million shares offered represented 100% of the listed Class A shares outstanding immediately after the offering. Measured against all 215,110,345 Class A, Class B, and Class N shares outstanding, the same offering represented 13.95%. The prospectus also states that Class B holders would control approximately 99.2% of the voting power immediately following the offering, based on beneficial ownership as of March 31, 2026.

An offered-share ratio for a company like this therefore has more than one defensible denominator, and each answers a different question. Reporting only the 100% figure would make the public offering appear to encompass the whole company. Reporting only 13.95% would conceal that the entire listed class initially came from the offering. The useful answer is to report both and name the denominator behind each.

Fervo Energy: The Option That Changed the Number

The Fervo Energy final prospectus reports a May 13, 2026, offering price of $27.00 per share and a base offering of 70,000,000 Class A shares for $1.89 billion.

Fervo Energy, offering terms

Measure	Value
Shares offered, base offering	70,000,000 Class A
Offering price	$27.00
Gross proceeds, base offering	$1,890,000,000
Underwriting discount	$1.62 per share, $113,400,000
Gross spread	6.00%
Retained fraction	94.00%
Over-allotment option	10,500,000 shares, exercised in full
Gross proceeds including the option	$2,173,500,000
Secondary component	none

The methodology reports the base offering and excludes the over-allotment option because, when the prospectus is filed, the option is contingent. Including it would mean counting shares that might never be sold. Applied consistently, the base-offering convention makes offerings more comparable with one another.

It also has a cost, and Fervo shows it plainly. The underwriters exercised the option in full, so the company sold 80,500,000 shares and raised approximately $2.174 billion, as confirmed in Fervo's filing after the offering closed. The $1.89 billion figure reported under the methodology is 13% below the amount ultimately raised. The convention is not wrong, and neither is the larger number. They answer different questions. A reader who treats the base offering as the final amount raised would be wrong by 13% on this deal.

Fervo also illustrates why an offered-share ratio says nothing about control. Measured against Class A shares outstanding after the offering, the ratio is 25.38%. Measured against all shares outstanding, it is 24.69%, a relatively small difference because Class B is a small class. But Class B carries 40 votes per share, and the company's two co-founders held approximately 2.75% of the capital stock while controlling approximately 53.03% of the voting power. The offered-share ratio measures how much stock was sold. It does not measure who decides.

Lincoln International: The Quarter's Only Validated Secondary Component

The Lincoln International final prospectus reports an offering priced on May 20, 2026, at $20.00 per share. It was the only validated offering in the quarter in which existing holders sold shares in the base offering.

Lincoln International, offering terms

Measure	Value
Shares offered by the company	20,604,046
Shares offered by selling stockholders	445,942
Total shares offered	21,049,988
Offering price	$20.00
Gross proceeds	$420,999,760
Underwriting discount	$1.40 per share
Gross spread	7.00%
Secondary component	2.12%

The cover states the position without ambiguity, and the proceeds table carries a separate line for amounts payable to the selling stockholders. That separation allows the measure to be read rather than inferred, and it is the reason the methodology takes the split from the offering-share table.

No direction is attached to the result. A secondary component of 2.12% is neither favorable nor unfavorable on its own, and the sellers in an offering of this kind may be founders, employees, funds, or other existing holders. What the measure records is that some of the proceeds from this offering, unlike the other 24 validated offerings in the quarter, went to existing holders rather than to the company.

The 7.00% spread is worth noting alongside it without over-reading it. It is the rate most commonly associated with smaller US offerings, applied here to a deal of roughly $421 million. The methodology's size benchmark cannot determine whether that spread was unusual because the applicable size band did not contain enough observations.

What This Study Does Not Establish

The benchmark table used for this study, version 1.0, benchmarks only one of its six size bands. The band covering offerings below $50 million holds the minimum of 20 offerings with a matched structured spread that the methodology requires. Every larger band falls short.

Most offerings in this study, including all four with validated spreads, therefore carry an absolute gross spread and no benchmark deviation. The rule is working as designed. A benchmark computed from only a handful of observations would produce a number with the appearance of authority and none of the substance.

Space Exploration Technologies anchors the study because its scale and multi-class structure show why spread and offered-share ratio have to be read together. Cerebras and Fervo were selected because their disclosures expose limitations in the methodology itself: the denominator question in one case and the over-allotment convention in the other. Lincoln International is included because it is the single observation behind the quarter-level finding on secondary shares. None of these offerings was selected to imply that it was better or worse than another deal.

Another 23 offerings could not be examined at all. They include several of the quarter's largest offerings, and nothing in this study describes their structure. The quarter-level statement about company and existing-shareholder shares covers the 25 offerings that were validated and no others.

Finally, nothing here is a view on any security. The measures describe how offerings were constructed, not how they have traded or will trade. A low spread, a large syndicate, or a small secondary component is a structural fact rather than a judgment about an investment.

Method and Reproduction

The initial universe was assembled from the IPOs Calendar API in month-sized windows. Structured offering economics were attached where available through the FMP IPOs Prospectus API, while the SEC Filings By Form Type API was used to locate final prospectuses filed under Forms 424B4, 424B1, and 424B3.

Every figure reported in this study was confirmed against the relevant final prospectus. Where a structured record also existed, it served as a cross-check rather than the controlling source. The universe rule, measures, retrieval process, validation requirements, and reference implementation are defined in the IPO Offering Structure methodology.

This study applies methodology version 1.0 to the quarter ended June 30, 2026, using data retrieved on August 19, 2026, and benchmark table version 1.0. Those identifiers fix the rules, data snapshot, and comparison basis behind the results.