Coupang CPNG

Reference table

1. What they actually sell — the segments, with real numbers

Coupang discloses three revenue lines and no segment profitability. Here is the full disaggregation available:

The single most important fact in this table: Third-Party Merchant Services has gone from 10.6% of revenue in FY23 to 20.6% in FY25, and it is compounding at more than double the rate of the retail line. Marketplace commissions and advertising carry structurally higher margins than buying and reselling goods, and this mix shift is exactly why gross margin expanded from 25.4% (FY23) to 29.2% (FY24) to 29.4% (FY25). That is 400 basis points of margin on a $34.5B revenue base — roughly $1.4B of incremental gross profit purely from mix.

And here is the disclosure problem. Coupang reports segments in its own filings as Product Commerce and Developing Offerings (per the company profile: Eats delivery, Play streaming, fintech, and the Farfetch luxury marketplace, plus operations across Taiwan, Singapore, Japan and elsewhere). This data file contains none of that split. There is no segment revenue, no segment operating income, no geographic breakdown whatsoever — seg_geo is completely empty. The income statement shows $0 R&D in every single year and one consolidated SG&A line of $9.67B for FY25.

Our own knowledge base called this exact problem eighteen months ago. business_breakdowns (2024-10-20, conviction 55, neutral): "Coupang runs ~25% gross margins with one lumped opex line; weak disclosure is a real criticism, though it hit EBIT/FCF profitability in 2023." The criticism has aged into the central obstacle to underwriting this name today. We cannot attribute the current losses to a segment because the data does not permit it.

2. The profit reversal — the actual problem, quarter by quarter

This is the reason the verdict is Watch. Read the operating income column downward.

Three deteriorations happening at once:

1. Revenue growth has halved — from +17.8% (Q3'25) to +10.9% to +7.5%. Note that Q4'25 revenue ($8,835M) was below Q3'25 ($9,267M) and Q1'26 fell again to $8,504M. Two consecutive sequential declines.

2. Gross margin has given back its gains — 31.3% (Q4'24) to 29.4% (Q3'25) to 27.0% (Q1'26), the lowest since Q1'24. Since the mix has been improving (third-party growing faster), a falling gross margin means the underlying unit economics are being pressured — most plausibly by price competition, promotional intensity, or the cost of a newly built network operating below capacity. The file cannot distinguish these.

3. Operating expenses have not flexed. SG&A ran $2,536M (Q4'25) and $2,539M (Q1'26) on revenue that fell from $8,835M to $8,504M. Flat costs against falling revenue produces exactly the -$242M operating loss observed.

FY25 in full: revenue $34.53B, operating income $473M (1.4% margin), net income $208M. Consensus FY26: net income -$491M, EPS -$0.286. That is a $700M year-over-year swing into loss on 7.7% revenue growth.

Precedent matters here, and it cuts both ways. Coupang has been through a far worse version of this: FY21 operating income was -$1.494B on $18.4B of revenue (-8.1% margin), and by FY23 it had reached +$473M (+1.9%). Management has demonstrated it can turn losses into profits when it chooses to. But FY23's 1.9% operating margin remains the all-time peak, on a business now 42% larger — Coupang has never demonstrated that this model produces more than about two cents of operating profit per dollar of revenue. Everything in the bull case depends on that ceiling being a choice rather than a limit, and this data cannot tell us which.

3. The balance sheet and the working-capital engine

FY25 year-end: cash $6,318M, no short-term investments, inventory $2,256M, total current assets $9,691M; total current liabilities $9,357M; short-term debt $960M, long-term debt $648M, total debt $4,635M (the gap between $1,608M of funded debt and $4,635M total is approximately $3.0B of capital-lease obligations against the $7,032M fulfilment property base); total assets $17,787M; equity $4,623M.

• Our computed net cash: $6,318M − $4,635M = $1,683M net cash. The vendor's enterprise value of $28.57B implies net cash of only $905M; our EV is $27.79B and all EV multiples here use it. If you exclude capital leases and count only funded debt, net cash is $4,710M — a materially different picture. Both framings are defensible; we use the conservative one (leases as debt), because a fulfilment network's lease obligations are as fixed as its mortgage.
• Current ratio 0.97 — working capital is negative ($9,691M current assets against $9,357M current liabilities gives +$334M at year end, but the trailing-twelve-month figure in the file is -$260M). This is by design, not distress.
• Cash conversion cycle: -53.7 days. Inventory turns in 29.7 days, receivables collect in 3.6 days, and payables are stretched to 87.1 days. Coupang collects from customers seven weeks before it pays suppliers. This is the Amazon model and it is a genuine structural advantage — the business is financed by its own suppliers. It also means growth generates cash and contraction consumes it, which is the risk embedded in the current deceleration.
• Leverage: total debt $4,635M against a TTM EBITDA that our own arithmetic puts at roughly $740M (Q2'25 $345M + Q3'25 $320M + Q4'25 $174M + Q1'26 -$99M). That is 6.3x gross debt to EBITDA, or net cash on a net basis. The EBITDA collapse is the story: $345M to $320M to $174M to -$99M across four quarters.

4. Cash flow — the second deterioration

Free cash flow has fallen 70% in two years — from $1.76B (FY23) to $522M (FY25) — driven by a 42% increase in capital expenditure in FY25 alone. Capex is now 3.6% of revenue and 70.6% of operating cash flow.

This is the most important ambiguity in the entire file, and it is genuinely two-sided. Rising capex plus falling margins is either (a) a company investing heavily into new capacity and new markets whose returns arrive later — the moat-building interpretation, consistent with the knowledge base — or (b) a company whose maintenance capital intensity is structurally higher than the bulls assume while competition compresses returns. The data does not adjudicate. Note also that stock-based compensation of $475M is now approaching the $522M free cash flow figure — on an SBC-adjusted basis FY25 free cash flow is close to zero.

5. Valuation — priced in or room?

At $16.42 (market cap $29.48B, our EV ~$27.79B), TTM revenue is $35.13B (Q2'25 through Q1'26).

Two coherence problems, flagged rather than smoothed: FY29E revenue growth of +3.0% following +11.2% is implausible as a forecast and reflects a shrinking analyst panel (14 to 9 contributors). FY29E EPS of $0.622 sitting BELOW FY28E's $0.747 is not a coherent earnings path. **We use FY27E and FY28E only, and treat FY29E as noise.

6a. What today's price assumes (the inversion)

At $16.42 — 0.79x TTM EV/sales and 22.0x FY28E consensus EPS — the price embeds roughly:

• Revenue reaccelerates from the current +7.5% to +14.4% in FY27, adding $7.4B of revenue between the TTM $35.1B and the FY27E $42.6B. (Source: consensus, 12 analysts.) Falsifier: two more quarters below +10% YoY. Note that Q1'26 already printed +7.5% and revenue has now declined sequentially twice.
• The company swings from a -$491M net loss (FY26E) to a +$730M net profit (FY27E) — a $1.22B turn in twelve months — and then to +$1.30B in FY28. (Consensus.) Falsifier: today's Q2 print showing an operating loss wider than Q1'26's -$242M.
• Operating margin recovers to roughly 3% to support the FY28E EPS of $0.747 on $47.3B of revenue — above the all-time peak of 1.9% achieved in FY23, on a revenue base 94% larger. (Our arithmetic from consensus net income.) Falsifier: any full year in which operating margin fails to exceed 1.9%.
• THE MOST FRAGILE ASSUMPTION: that the FY26 losses are investment rather than erosion. The 0.79x sales multiple already reflects deep scepticism — it prices Coupang roughly in line with a low-margin retailer, not a marketplace. But even that multiple requires the loss to be temporary. If the 27.0% Q1'26 gross margin is the new normal rather than a trough, the FY27-28 earnings path does not exist at any multiple. This is the assumption the entire equity rests on, and this file cannot test it.

6b. The return bridge (why the multiple moves)

Expected return ≈ EPS growth + multiple drift + shareholder yield.

Because Coupang is currently loss-making, the base case cannot be built on earnings growth from a positive base — the return has to come from the market re-rating a sales multiple as profitability returns. Our base case of $21 is constructed two ways that converge:

1. Sales-multiple route: 0.95x EV/sales on FY27E revenue of $42.56B = $40.4B EV, plus $1.68B net cash = $42.1B equity value ÷ 1,855M diluted shares = $22.70.

2. Earnings route: FY28E EPS $0.747 × 26x = $19.42.

Averaging to ~$21. This explicitly assumes multiple EXPANSION from 0.79x to 0.95x EV/sales — roughly a 20% re-rating — and the driver named is straightforward: a retailer that returns to profit gets paid as a marketplace rather than as a reseller. That expansion is the fragile leg of the bridge. If the margin does not recover, the multiple does not expand, and the stock is worth what it is worth today. There is no dividend; buybacks were $243M in FY25 (0.8% of market capitalisation), a rounding error against a $491M loss. Shareholder yield contributes essentially nothing.

Note carefully: at an unchanged price the EV/sales multiple mechanically declines from 0.79x (TTM) to 0.59x (FY28E) simply because revenue grows. That rolldown is not a de-rating and should not be read as one.

6c. Variant perception (where we differ, what would surprise)

• We are more negative than the street on the near term and roughly aligned on the asset. Consensus targets average $26 (+58%) with 14 of 16 analysts at buy and none at sell, and the lowest target on the street is $23 — 40% above spot. A price target band of $23-$31 on a stock at $16.42, in a name that just lost $266M in a quarter, is a very confident sell side. Our base of $21 is deliberately below the street's low target. That is our variant perception: we think the recovery is less certain and later than the modelled consensus, and the 0.79x multiple is telling you the market agrees with us rather than with them.
• Where we are MORE constructive than the market: the third-party mix shift. A line growing +27.5% to $7.1B, now 20.6% of revenue and rising 500 basis points a year, is a structurally higher-margin business being built inside a lower-margin one. A -51% drawdown does not obviously price that. This is the reason the verdict is Watch rather than Avoid.
• Positive surprise that would force a repricing: today's print showing the operating loss narrowing materially versus -$242M, with gross margin recovering above 28% and third-party revenue holding 25%+ growth. That combination would date the trough at Q1'26 and make the FY27 consensus credible. Watchable at: 2026-08-04, today.
• Negative surprise: an operating loss wider than -$242M, or gross margin below 27%, or revenue growth below +7%. Any of these would suggest structural erosion rather than investment, and the bear case at $11.50 becomes the relevant scenario. Watchable at: 2026-08-04, today.

Synthos fair values

• Bear ~$11.50 (-30%). Losses prove structural: gross margin stabilises near 27%, revenue growth settles at high single digits, and the market prices Coupang as a low-margin retailer at 0.55x EV/sales on FY26E revenue ($37.21B × 0.55 = $20.5B EV, plus $1.68B net cash, ÷ 1,855M shares = $11.95). Rounds to $11.50 with the 52-week low of $14.92 already breached on the way down.
• Base ~$21 (+28%). The mid-point of the two routes above; assumes profitability returns roughly on the consensus schedule and the multiple expands from 0.79x to ~0.95x sales.
• Bull ~$31 (+89%). The FY23 playbook repeats — operating margin recovers past 3%, third-party crosses 25% of revenue, and Coupang is re-rated as a marketplace at ~1.3x EV/sales on FY27E revenue. This matches the street's high target of $31 and is still 7% below the 52-week high of $33.53.

6. Knowledge base — real conviction, but every claim predates the problem

6 claims across 2 tracked voices. The lane is unambiguously positive on the business and completely silent on the current situation, and that gap is the most important thing about it.

business_breakdowns (2024-10-20, five claims from a single episode):

• Conviction 70, bullish: "Coupang's owned first-party logistics built trust/reliability rivals' third-party marketplaces can't match, creating a durable e-commerce moat in Korea."
• Conviction 65, bullish: "Coupang has long runway: ~25% e-commerce share but high-single-digit overall retail penetration, and cohorts keep increasing spend over time."
• Conviction 60, neutral: "Naver is Coupang's biggest competitor, followed by SSG; third-party marketplaces undercut on price but lose on reliability and returns."
• Conviction 55, neutral: "Coupang runs ~25% gross margins with one lumped opex line; weak disclosure is a real criticism, though it hit EBIT/FCF profitability in 2023."
• Conviction 45, neutral: "Coupang's ~$500M Farfetch buy (~$4B GMV) looks cheap but strategic rationale is unclear, possibly to bolster weak apparel/beauty categories."

invest_like_the_best (2025-04-15, conviction 95, bullish — the highest-conviction claim on any name in this batch):

• "Coupang's Rocket one-P delivery is a jaw-dropping customer experience; still buying shares after 15 years on the board."

Read, carefully. This is a breadth-2, positive lane with one exceptionally high-conviction claim, and it deserves weight on the asset question — an investor with fifteen years on the board buying shares is about as informed a signal as exists. But it is dated 2025-04-15, and the stock was materially higher then; the loss swing began three quarters later.

Five of the six claims are from a single 2024-10-20 episode — that is one source-day carrying 83% of the lane. Breadth of two is thin, and concentration in one date is thinner still. Two of the five 2024 claims are explicitly cautionary: the disclosure criticism (conviction 55) and the Farfetch scepticism (conviction 45, "strategic rationale is unclear"). The panel's own reservations from eighteen months ago — opacity and questionable capital allocation — are precisely the two issues that make today's losses un-diagnosable.

No tracked voice has commented on Coupang since 2025-04-15. The lane is stale relative to the facts. We record it as positive-but-stale and decline to convert it into a buy signal.

7. Technicals — a long, deep, unrecovered decline

• Price $16.42 (+0.43% today), -51.0% from the 52-week high of $33.53, +8.6% from the 52-week low of $15.12. Maximum drawdown from peak: -67.5% — meaning the stock's all-time peak sits well above the 52-week high.
• Below the 50-DMA ($16.89) by 2.8% and far below the 200-DMA ($21.02) by 21.9%. The 50-DMA is well beneath the 200-DMA — a fully broken trend structure in which the intermediate average has been below the long-term average for an extended period.
• RSI 35.96 — soft but not washed out; there is room to fall further before reaching a genuine oversold extreme. MACD -0.44.
• Relative performance is the standout: -44.2% over 12 months against SPY +19.9% and QQQ +23.9%. That is 64 points of underperformance against the S&P in one year, and it did not happen in a straight line — 3-month return -20.3%, 6-month -18.6%. The decline has been persistent rather than event-driven, which is consistent with a slow-motion earnings deterioration rather than a single shock.
• Beta 1.109 — market-like, notably lower than the growth names in this batch.

There is no constructive technical setup here. The stock is below both averages, in a confirmed downtrend, with an RSI that has not reached capitulation. Technically, this is a falling knife, and that reinforces the fundamental case for waiting.

8. Moat & competitive position

The moat is real and our panel is right about its nature. Coupang's advantage is owned end-to-end fulfilment in a dense, urbanised market — $7.03B of property, plant and equipment, 108,000 employees, and a delivery promise that third-party-dependent marketplaces cannot match on reliability. business_breakdowns states the mechanism precisely: rivals "undercut on price but lose on reliability and returns." Vertical integration in logistics is expensive to build, slow to replicate, and improves with density — every incremental order in an existing zone raises route efficiency. That is a genuine, compounding structural advantage.

Three qualifications, all supported by the data:

1. It is a single-country moat. Korea has roughly 52 million people. The knowledge base frames the runway against total retail penetration rather than e-commerce share — a real and long runway, but a bounded one. The file's seg_geo block is entirely empty, so we cannot size the international businesses at all.

2. Naver is a serious competitor (named by the panel as the primary rival, followed by SSG), and a search-and-commerce incumbent with its own network effects is a different class of opponent than a pure marketplace.

3. The moat has never produced more than a 1.9% operating margin. A moat that cannot be converted into durable profit is a moat around a low-return business. The single most important open question about Coupang is whether the ceiling is strategic choice or structural limit — and no data available to us answers it.

Capital allocation is the weaker flank. The Farfetch acquisition — roughly $500M for a ~$4B GMV luxury marketplace — was flagged by our own panel as strategically unclear at the time of purchase. The acquisitionsNet line shows $0 in FY25 and +$68M in FY24, so no further large deals appear in the file. Buybacks total $421M across FY24-25, meaningful but not decisive.

9. Data integrity — what we rejected and why

Vendor corruption checking is standard practice for these dives. For CPNG:

• REJECTED: the entire consensus EBITDA and EBIT series. The est block reports negative EBIT in every forecast year — -$138.3M (FY26E), -$158.2M (FY27E), -$176.0M (FY28E), -$181.2M (FY29E) — while reporting positive net income of +$729.7M (FY27E) and +$1,301M (FY28E) in the same rows. A company with negative EBIT and $1.3B of net income would require $1.5B of non-operating income, which is not remotely plausible here. Separately, the FY25 estimate row shows EBITDA of $307.1M against reported FY25 EBITDA of $1,200M — off by a factor of four. This series fails internal consistency and is excluded entirely. Only revenue, net income and EPS from est are used.
• REJECTED: FY29E as a forecast. EPS of $0.622 sits below FY28E's $0.747 and revenue growth of +3.0% follows +11.2%, both driven by a shrinking analyst panel (14 EPS contributors for FY28E falling to 8 for FY29E). Treated as noise.
• CORRECTED: enterprise value. The vendor reports $28.57B, implying $905M of net cash. From the FY25 balance sheet, cash of $6,318M less total debt of $4,635M gives $1,683M of net cash and an EV of $27.79B. All EV multiples here use our figure. We also disclose the alternative: excluding ~$3.0B of capital leases, net cash is $4,710M and EV is $24.77B (0.71x sales). We use the conservative treatment.
• FLAGGED: ratios_ttm reports a P/E of -170.0x and km_ttm an incomeQualityTTM of -9.75. Both are arithmetic artifacts of a small negative TTM net income and carry no information. Ignored rather than reported.
• FLAGGED: researchAndDevelopmentExpenses = $0 in every year. This is a genuine reporting choice (Coupang does not break out R&D), not a data error, but it means the entire cost base sits in one $9.67B SG&A line with no visibility into technology investment.
• MAJOR GAP: seg_geo is completely empty. No geographic revenue split at all — we cannot size Korea versus Taiwan versus the rest, which is precisely where the Developing Offerings losses would show.
• MAJOR GAP: no segment operating income. Product Commerce versus Developing Offerings profitability is the single number that would explain the loss swing, and it is absent.
• MAJOR GAP: no Q2'26 income statement. The most recent quarter in inc_q is dated 2026-03-31. Q2 reports today, 2026-08-04, with a consensus estimate of -$0.270 EPS on $8.92B of revenue. This dive is written without the most important available fact about the company, and the verdict reflects that.
• NOTE: insider activity is entirely director share awards (six directors granted 17,400-21,300 shares each on 2026-06-11 at $0, plus an officer award of 213,900 shares on 2026-04-01). There are no open-market purchases and no sales in the file. No signal either way.

10. Verdict, kill-criteria & flip conditions

Watch. Coupang owns a genuine logistics moat that our knowledge base independently underwrites, is growing a high-margin third-party marketplace line at +27.5%, holds $6.3B of cash with suppliers financing its working capital, and trades at 0.79x enterprise value to sales after a 51% drawdown. On the asset, the bull case is sound and our base fair value of $21 sits 28% above spot.

But the earnings direction has reversed and we cannot explain why. Operating income went from +$312M to -$242M across five quarters; consensus models a -$491M FY26 net loss; gross margin has fallen 430 basis points from its peak; free cash flow is down 70% in two years while capex is up 42%; the file contains no segment profitability, no geographic split, and no Q2 data; and every substantive knowledge-base claim predates the problem by eighteen months.

Q2'26 reports today. Buying into a binary print, on a name whose loss driver is undiagnosable from available data, using conviction that is a year and a half stale, is not a Stage-In. It is a coin flip with a story attached. Wait for the number.

Re-engagement protocol (this replaces a staged entry, because the gate is informational rather than price-based):

1. Read today's print for three things in order: (a) operating income versus Q1'26's -$242M, (b) gross margin versus 27.0%, (c) Third-Party Merchant Services growth versus +27.5%.

2. If all three improve — loss narrowing, margin above 28%, third-party holding 25%+ — open a ≤1.5% starter and treat Q1'26 as the trough. Base case $21 becomes actionable.

3. If the loss widens or gross margin breaks below 27%, stand down entirely and re-underwrite against the $11.50 bear case. Do not average into a structural margin problem.

4. Regardless of the print, size small. Disclosure quality alone caps this at a starter position — our own panel identified opacity as a real criticism in 2024 and it is now the binding constraint on underwriting.

Pre-registered KILL / avoid criteria:

• Gross margin below 27% in any quarter — the FY23-25 mix-driven margin expansion reversing means competition, not investment.
• Two more quarters below +10% revenue growth, which would break the FY27 consensus reacceleration.
• Operating margin failing to exceed 1.9% (the FY23 record) in any full year through FY28 — this would settle the "choice versus ceiling" question in the wrong direction.
• Free cash flow turning negative with capex still rising.
• Any further large, strategically unexplained acquisition in the Farfetch mould.

Pre-registered FLIP TO STAGE-IN:

• Today's print showing operating income back above zero, or a loss materially narrower than -$242M with margin recovery.
• Third-Party Merchant Services crossing 25% of total revenue — the marketplace transition becoming the dominant story.
• Any improvement in segment disclosure — Product Commerce versus Developing Offerings operating income would remove the single largest obstacle to underwriting this name.
• A fresh knowledge-base claim from a tracked voice engaging with the post-2025 situation.

Where CPNG fits in the Synthos Framework Portfolio. The international e-commerce / owned-logistics sleeve, currently at 0% weight with a ≤1.5% starter authorised only post-print. Note the correlation profile: beta 1.109 is market-like, and the -44% twelve-month return against a +20% index means this has been an idiosyncratic decline rather than a beta event — which is both the opportunity and the warning. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $16.42.

Single biggest risk: an unexplained profit reversal in an opaque reporting structure. Coupang lost $266M in Q1'26 after earning $208M in FY25, and this data file — with one lumped operating expense line, $0 disclosed R&D, no segment profitability and an entirely empty geographic block — makes it impossible to determine whether the losses come from deliberate investment in Developing Offerings and Taiwan, or from margin erosion in the core Korean business. The most fragile assumption in today's price is that the Q1'26 gross margin of 27.0% is a trough rather than a new level. If it is a new level, the consensus path from a -$491M loss to a +$1.30B profit does not exist, and no sales multiple rescues the equity.

Provenance & disclosures

• Traceability: 6 KB claims across 2 tracked voices (breadth 2, net conviction positive-but-stale). Five claims from business_breakdowns on a single date, 2024-10-20 (convictions 70, 65, 60, 55, 45), and one from invest_like_the_best on 2025-04-15 (conviction 95, a board member of fifteen years still purchasing). No tracked voice has commented on Coupang since 2025-04-15, i.e. before the profit reversal began. Two of the six claims are explicitly cautionary (disclosure quality; Farfetch rationale). The lane underwrites the business and is silent on the current problem.
• Data as-of: fundamentals 2026-03-31 (Q1'26, filed 2026-05-05 — the most recent available; Q2'26 reports TODAY and is not in this file) · estimates 2026-08-04 · prices 2026-08-04 ($16.42; 50-DMA $16.89, 200-DMA $21.02, RSI 36.0, beta 1.109) · KB claims 2026-08-04. Sole data source: scripts/deepdive/vti_data/CPNG_data.json.
• Rejected data: the entire consensus EBITDA/EBIT series (negative EBIT in all forecast years alongside positive net income; FY25 estimate off 4x versus actual); FY29E treated as noise (non-monotonic EPS, collapsing analyst count). Vendor enterprise value corrected from $28.57B to $27.79B.
• Disclosed gaps: seg_geo completely empty (no geographic revenue split); no segment operating income (Product Commerce versus Developing Offerings); $0 R&D reported in every year with one $9.67B consolidated SG&A line; no Q2'26 income statement; no per-analyst price-target detail; no user, order-volume or active-customer metrics.
• Fair-value caveat: the $11.50 / $21 / $31 anchors are scenario multiples on consensus revenue and EPS, labelled at each step. The base case requires EV/sales multiple expansion from 0.79x to ~0.95x and that expansion is named as the fragile leg. This is scenario arithmetic, not a discounted-cash-flow model.
• Debt treatment note: total debt of $4,635M includes approximately $3.0B of capital-lease obligations beyond the $1,608M of funded debt. We treat leases as debt (net cash $1,683M, EV $27.79B). The alternative treatment gives net cash of $4,710M and EV of $24.77B (0.71x sales). Both are disclosed.
• Not investment advice. Independent research, educational and informational only, never personalized. Nothing here is a recommendation to buy or sell any security.
• Version: 2026-08-04-full.