Forty-five years of US IPOs, from Jay Ritter's dataset. The question was never whether the 2019–2024 IPO cohorts would burst — they already have, and the results are final. Here's who lost, how much, and why the biggest first-day fireworks reliably precede the worst three-year returns. In 2026 the framework finally gets live subjects: the two largest listings in history — SpaceX and SK hynix — six weeks apart, and they don't belong in the same bin.
$250BLeft on the table, 1980-2025
−78.6%2020 cohort, 3-yr vs. market
60%Of all IPOs lose money in 3 yrs
$1.2TRaised across 9,343 IPOs
−0.38Pop-vs-return correlation
!
Not investment advice. All content reflects personal analytical views by DamithInvest and is based on publicly available information, including Jay Ritter's (University of Florida) published IPO statistics and public market data. This is not a buy, sell, or hold recommendation on any security mentioned, including SPCX or SKHY. US stocks are not regulated by the SEC of Sri Lanka. Do your own research and consult a licensed financial advisor before investing. Past IPO performance does not predict future results.
2020 COHORT
3-yr vs. market
−78.6%
Worst cohort in 45 years
41.6% avg day-1 pop
2021 COHORT
3-yr vs. market
−68.6%
311 IPOs — the volume flood
32.1% avg day-1 pop
MEDIAN IPO
3-yr return, 1975-2021
−25.7%
Averages hide this
60% of all IPOs lose money
SKHY
Day-1 close · Jul 10 2026
+12.8%
Below the 19% avg pop
Fwd P/S ≈ 4.5×
SPCX
Since $135 IPO · Jul 10
≈−35%
From Jun 16 peak +67%
P/S at IPO ≈ 94×
CORRELATION
Pop vs. 3-yr return
−0.38
Across 45 cohort-years
Top-right quadrant: empty
◆ Our Analysis — July 10, 2026
"Post-2017 IPOs are bad" is too blunt a claim. 2017 and 2018 were genuinely strong vintages that beat the market. The real wipeout is the COVID/SPAC window — 2020 and 2021 especially — and it's already realized, not pending. Across 45 cohort-years, the correlation between a cohort's average first-day pop and its 3-year return versus the market is −0.38: hotter debuts predict worse multi-year outcomes. Splitting the data by quality (offer price, profitability, revenue, price-to-sales) shows the average return hides a bimodal reality — real, profitable, sensibly-priced companies roughly track the market, while junk-tier IPOs are where 60% of all listings post a negative 3-year return. 2026 puts this framework on two live subjects at once: SK hynix, priced at a forward P/S of ~4.5× with real profit, and SpaceX, priced at ~94× sales with no path to profitability disclosed in its own filing. Same "mega-IPO" headline. Opposite bucket.
01
The Cohorts — 45 Years of IPO Vintages
"Post-2017" is too blunt. The burst is 2019-2024.
2017 and 2018 were genuinely strong vintages that beat the market. The wipeout is the COVID/SPAC window — 2020 and 2021 especially — and it's already realized, not pending. The chart below plots the number of IPOs per year (bars, colored by era) against each cohort's 3-year return versus the market (gold line), using Ritter's Table 1 and Table 19 data, 1980-2024.
Every dot below is one cohort-year: horizontal position is the average first-day pop, vertical position is that cohort's 3-year return versus the market. Dot size scales with the number of IPOs that year. The cloud slopes down: hot markets misprice the most, flippers capture the pop, and the buyers left holding underperform for years. Correlation across the 45 cohorts: −0.38.
Top-right is empty. No cohort ever paired a huge pop with a strong three-year return. The fireworks and the payoff never co-occur.
Bubbles cluster low. The gold (2020-22) and red (1999-2000) dots — the loudest markets — sit at the bottom of the chart, exactly where the losses are.
03
Junk vs. Quality — "IPOs" Is Two Different Animals
3-yr return vs. market, by quality split
The average hides a bimodal reality. Real-revenue, profitable, sensibly-priced companies roughly track the market. The junk tier — penny stocks, no revenue, no profits, absurd price-to-sales — is where the wealth destruction lives.
By offer price
2001-2024 · Ritter Table 19c
Below $5.00
−82.6%
Exactly $5.00
−91.3%
Above $5.00
−11.8%
By profitability
1980-2024 · Ritter Table 16b
Unprofitable
−30.7%
Profitable
−13.0%
By trailing revenue
1980-2024 · Ritter Table 16
$0-10M
−52.3%
$10-20M
−35.8%
$20-50M
−30.1%
$50-100M
−15.3%
$100-500M
−2.6%
$500M+
−4.4%
By price-to-sales (rev > $100M)
1980-2024 · Ritter Table 18d Panel A
P/S < 5
−1.3%
5-10
−8.0%
10-20
−7.7%
20-40
−16.3%
40+
−58.5%
◆ Why the Count Exploded but Quality Didn't
Much of the 2021-2025 surge in the raw IPO number is sub-$5 foreign microcaps — Ritter counts 109 penny-stock offerings in 2025 alone. A real-revenue AI listing and an $8M shell are both "IPOs" in a headline, and about 83% apart three years later.
04
The Money — Two Separate Losses
Don't conflate them
There are two distinct pools of money here. Money left on the table is wealth handed from the issuing company to day-1 flippers — the issuer under-priced. Separately, secondary buyers underperform for years. The dot-com and COVID bubbles light up on both.
Aggregate proceeds vs. money left on the table ($B)
Ritter Table 1 · 1980-2025
Proceeds raisedMoney left on the table
Where every IPO cohort landed after 3 years
Share of IPOs by 3-yr return · from first close · 1975-2021 · Ritter Table 16e
60%
of all IPOs since 1975 posted a negative three-year return if bought at the first close. The median IPO lost −25.7%. Averages get rescued by a handful of monster winners — which is exactly why "the average IPO returned X%" is a misleading number.
05
The 2026 Test — SpaceX vs. SK hynix
Two record IPOs, six weeks apart, opposite risk profiles
2026 finally puts the framework on live subjects. SpaceX (Jun 12) and SK hynix (Jul 10) are the two largest listings in history. Same surface — mega-size, 2-5% float, an AI story, heavy retail demand. Underneath, the data sorts them into different bins. The pop is not the tell; the price against sales and profit is.
SK hynix
SKHY
Nasdaq · listed Jul 10 2026
Offer price$149
First-day close$168.01
Day-1 pop (close)+12.8%
Intraday high$174.50
Raised$26.5B
Market cap~$1.0T
Fwd P/S · 2026E~4.5×
Public float2.5%
Real revenue $65B · deeply profitable · HBM #1 at 56%
SpaceX
SPCX
Nasdaq · listed Jun 12 2026
Offer price$135
Peak · Jun 16$225.64
Now · Jul 10~$145
From peak~−35%
Raised~$75B
Market cap~$1.8T
P/S at IPO~94×
Public float5%
Cash-burning AI + launch units · ~94× sales at IPO
First-day pop — the live IPOs vs. history
Day-1 close vs. offer price · bubble cohorts shown for scale
SK hynix · 2026
+12.8%
45-yr average
+19.0%
SpaceX · 2026
+19.3%
2020 cohort
+41.6%
2022 cohort
+48.9%
1999 cohort
+71.2%
SK hynix popped +12.8% — below the 45-year average of ~19%. That is the opposite of a bubble signature. The cohorts that later cratered opened 40-70% higher. SpaceX closed +19% but spiked +67% intraday before fading. A calm debut is a feature, not a warning.
Where each lands on the metric that actually predicts returns
3-yr return vs. market by price-to-sales · rev > $100M · 1980-2024 · Table 18d + 2026 overlay
This is the punchline. On the ratio that best predicts three-year IPO returns, SpaceX sits in the death zone (P/S > 40 → −58.5% historically), while SK hynix's forward multiple sits in the safe zone — its explosive 2026 revenue drags the ratio down to ~4.5×. Same "mega-IPO" headline; opposite bucket.
SpaceX after the pop — the fade the framework predicts
% vs. $135 IPO price · approximate daily path · Jun 12 - Jul 10 2026
+67% in four sessions, then a ~35% round-trip back toward the offer price as the bond sale and looming insider unlock spooked buyers. SK hynix is one session in at +13% — the arc hasn't run yet, but its fundamentals (real profit, low forward P/S) make a SpaceX-style valuation unwind far less likely. Its risk lives elsewhere — the verdict below.
06
SK hynix Verdict — Right Debut, Demanding Price
Base rates + through-cycle DCF
Pulling the IPO base rates and a through-cycle DCF together. The framework clears SK hynix on quality — it is not the animal that craters. The caution sits one layer down, in a risk the IPO data never captured: memory at a cyclical peak.
Through-cycle DCF vs. where it trades · per ADS
Normalized-FCF model · 3 scenarios · net cash ~$21B · ~730M shares · WACC 9-12%
Bear
$26
$26
−85% vs. close
Base
$51
$51
−70% vs. close
Bull
$124
$124
−26% vs. close
Close / IPO
IPO $149 → Close $168
$168
Current trading
Even the bull case — HBM permanently de-commoditizes, capex eases, a US-listing re-rating — lands near $124, below the $168 close. At today's price the market is capitalizing peak-cycle economics as if boom-bust is over. That is the entire bet.
Not a bubble IPO
Real revenue ($65B), deeply profitable, a +13% pop below the 45-yr average, forward P/S ~4.5×. On Ritter's base rates that profile roughly tracks the market — it sits with the survivors, not the −58% P/S>40 tier.
The risk is cyclical, not IPO-quality
Memory swung from a $9B loss (2023) to a $28B profit (2025); Q1'26 net was inflated by ~$11B of non-operating finance income. The calm debut says nothing about where the cycle turns. Ritter's dataset never measured this.
SpaceX is the textbook case
94× sales, 5% float, retail-driven +67% spike, ~−35% fade. That is exactly the pattern 45 years of data flags. Don't map its trajectory onto SK hynix — different failure mode entirely.
Opposite technical setups
SKHY: 2.5% float + likely December Nasdaq-100 inclusion = passive bid incoming. SPCX: large insider unlock by September = supply flood. The float mechanics point in opposite directions.
◆ Key Findings & Positioning
1. Separate the two failure modes. SpaceX is a valuation bubble — the framework screams. SK hynix is a cyclical peak — the framework is quiet, but the DCF is not. Same headline, different physics.
2. The debut was healthy. A calm +13% is a good sign, not a red flag. The red flag is price vs. normalized earnings: $168 sits above the through-cycle value ($51 base, $124 bull).
3. The whole question is cyclicality. If HBM has genuinely ended boom-bust (management's claim that "demand permanently changed"), the price is defensible. If memory still mean-reverts, the gap to fair value is the drawdown.
4. Base-rate discipline. Large, profitable, sensibly-priced IPOs tend to track the market; peak-cycle names carry real first-year drawdown — the 15 biggest US IPOs averaged −2% at one year, −23% max drawdown. Size for the cycle, not the story.
07
Sources & References
[1] Jay R. Ritter — "Initial Public Offerings: Updated Statistics," University of Florida, Jul 7 2026 · site.warrington.ufl.edu/ritter/ipo-data
[2] Ritter Table 1 — Money left on the table & aggregate proceeds, 1980-2025
[3] Ritter Table 16 / 16b / 16e — Returns by revenue, profitability, and distribution of 3-yr BHR
[4] Ritter Table 18d — Returns by price-to-sales, companies with revenue > $100M
[5] Ritter Table 19 / 19c — 3-yr buy-and-hold returns from first close, by cohort year and offer price
[6] Public market price data for SPCX and SKHY, through Jul 10 2026
Note: All returns are equally weighted, measured from the first closing price (what a day-1 retail buyer earns). Market-adjusted = IPO return minus the Morningstar U.S. total-market index over the same window. This is an independent analysis by DamithInvest and is not affiliated with SpaceX, SK hynix, or any financial institution.
