Lunit Stock Has Burned More Cash Than It Has Ever Booked

Shares closed at KRW 8,560 on September 3, 2026 (Thu), putting the whole company at KRW 637.4 billion, or roughly USD 469 million at the KRW 1,359.3 per dollar Seoul close that same day. There are 74,458,180 shares out. I hold none of them and have no order working. What pulled me in was arithmetic I do not often get to do this cleanly.

Contents18 min read

Lunit stock trades on a burn that outruns its revenue

Two running totals, five reporting periods

Korean filings give me five periods with complete figures: fiscal years 2022 through 2025 and the first half of 2026. Rather than reading each year on its own, I carried both columns forward and watched the difference.

All figures in billions of Korean won, consolidated, as filed with Korea’s electronic disclosure system. The 1H 2026 row comes from the semiannual report, receipt number 20260811000657. Every cumulative column and every gap is my own running sum and is derived, not displayed by any screen.

The ratio at the bottom row is 1.046, so the burn exceeds the revenue by 4.58%. At the September 3, 2026 (Thu) rate that is roughly USD 171 million out against USD 163 million in. I have not seen many companies where the entire revenue history is smaller than the entire operating cash consumption over the same window, and it is worth knowing before looking at anything else.

The gap has a shape, and the shape matters more than the level

Reading down the last column, the shortfall widened in each of the first three years, from KRW 39.15 billion to KRW 50.55 billion to KRW 62.07 billion. Then it turned. It fell to KRW 34.11 billion by the end of 2025 and to KRW 10.18 billion after the first half of 2026.

The reduction over that last half is KRW 23.93 billion (derived from the two printed gaps). Computing it the other way, from the half’s own revenue of KRW 45.77 billion less its cash consumption of KRW 21.83 billion, gives KRW 23.94 billion. The one hundredth of a billion between them is rounding in the figures I am printing, and the two agree at full precision. I mention it because I would rather show the seam than paper over it.

What that shape means is that the question in front of me has a date attached. On the pace of the half just reported, cumulative revenue overtakes cumulative operating cash consumption during the second half of 2026. That is close enough to test, and I built one of my two falsifiers on it.

Why the cash statement carries this argument

An operating loss can be moved around by timing. Revenue recognition schedules, capitalized development spending, and accrual choices all shift where a loss lands without changing what leaves the bank account. The cash flow statement is harder to arrange.

There is a second reason specific to this name. Its net result and its operating result disagree often enough that I stopped using the net line entirely. A KRW 171.5 billion convertible bond issued in May 2024 is my working explanation, since a separated conversion feature gets marked down as the share price falls and that markdown lands in profit. I was unable to obtain the note-level breakdown that would confirm it, so I am flagging that as my hypothesis and nothing firmer. Either way, a line that can swing on a valuation model is not a line I want under a five-period sum.

Lunit stock and the five periods that made that burn

Revenue compounding at 81.67% a year

From KRW 13.87 billion in 2022 to KRW 83.13 billion in 2025 is a three-year compound annual growth rate of 81.67%, which reproduces exactly against the figure my data screen carries. Very few companies of any size grow a top line that fast for three consecutive years, and that growth is why the burn has a case behind it at all. A company consuming cash to build a base compounding at this rate is doing something different from a company consuming cash to stand still.

It also explains the shape of the gap. Revenue is compounding as cash consumption stays roughly flat in absolute terms, running between KRW 36 billion and KRW 66 billion a year across the four full years. A growing number chasing a stationary one closes eventually. The only question is whether the stationary one stays stationary.

The burn itself has not fallen in a straight line

Cash consumption went KRW 53.02 billion in 2022, KRW 36.47 billion in 2023, KRW 65.70 billion in 2024, and KRW 55.18 billion in 2025. The worst year in the set is 2024, which is also the year after the largest arrival of outside capital, a coincidence I come back to further down.

So anyone expecting the burn to fall smoothly alongside the revenue growth will be disappointed by the record. What has fallen smoothly is the relationship between the two, and that is a different claim requiring different evidence. I try to keep those two claims separate in my own head, because conflating them is how a reader ends up with a tidier picture than the filings support.

What management has said, and why I did not build a condition on it

In February 2026 the company guided to full-year revenue growth of 40% to 50% along with an operating expense reduction of roughly 20% and a headcount cut of about 15%. Chief executive Seo Beom-seok reaffirmed in the August 11, 2026 (Tue) results release that the company still targets a full-year break-even on an EBITDA basis this year, a goal Korean coverage of the February briefing described as having been pulled forward from 2027.

I did not write a condition on that break-even, and the reason is structural. EBITDA does not appear in the standard Korean filing format. The EBITDA field in my own data source is empty for this company, because depreciation lives only in the notes, outside the standard financial statement response. So the company defines and computes the figure by which it asks to be measured. When the entity being judged also holds the definition, the condition stops testing facts and starts testing definitions. A break-even announced on that basis will not move either of my conditions.

A sales multiple of 6.94 needs a peer group to mean anything

Three different sales multiples for one company

Trailing twelve month revenue here is KRW 91.82 billion, which I built by taking the first half of 2026 and adding the second half of 2025. Investing.com displays the same KRW 91.82 billion, a useful check that I constructed the trailing window the same way the screen did.

Divide the KRW 637.4 billion market value by that and the sales multiple is 6.94 (derived). My Korean data screen instead shows 7.67, which reproduces if the denominator is the fiscal 2025 revenue of KRW 83.13 billion. Investing.com shows 7.1. Three numbers, one company, one day, and the whole difference is which revenue window sits underneath. I use the trailing figure because the peer screens I am about to compare against use trailing figures too.

A note on where this company is listed

Lunit trades on KOSDAQ, the smaller of South Korea’s two main boards. KOSPI carries the large industrial and financial names most overseas investors have heard of, and KOSDAQ carries younger technology, biotech, and media companies at generally smaller capitalizations with wider daily price swings. A KRW 637.4 billion company is a small capitalization even by KOSDAQ standards, and nothing here should be read as though it were about a KOSPI blue chip.

Where Lunit stock sits among five loss making imaging names

The group I built and the rule I used

My rule for this table was narrow on purpose. Listed companies that lose money and that sell into medical imaging or oncology diagnostics, compared on two axes only, a sales multiple and a revenue growth rate, with no currency conversion between them. All four comparison rows come from public company screens checked on September 3, 2026 (Thu).

Dividing each sales multiple by its growth rate gives 0.302 for Lunit, 0.616 for Butterfly Network, 0.159 for Tempus AI, 0.141 for Nano-X, and 0.136 for RadNet, all derived by me. Lunit lands third of five, the middle of the group. Butterfly Network is paying for growth at roughly twice Lunit’s ratio and RadNet at less than half of it. Whatever case exists here, the sales multiple is not it.

Eight places this table is loose

I would rather list these than let a reader assume the comparison is tighter than it is. One, my row is my own arithmetic and the other four are screen values. Two, my growth figure is a first-half year-over-year comparison and the peer growth figures are trailing. Three, I did not convert currencies, which is fine for the ratios and wrong for the level columns. Four, RadNet’s core business is running outpatient imaging centers, with its DeepHealth AI unit one part of the whole. Five, Butterfly Network sells handheld ultrasound hardware. Six, Nano-X combines hardware with teleradiology services. Seven, a large share of Tempus AI’s revenue comes from sequencing tests, a different business from image reading. Eight, my net result is a half year figure and the peer figures are trailing or annual.

Seven of those eight loosen the comparison in ways I cannot fix without data I do not have. I kept the table because the ordering survives every one of them, and because a Korean small capitalization with no domestic analyst coverage is otherwise hard to place at all. I also dropped one candidate: iCAD, which sells breast imaging AI and would have fit the rule, but the screen I use carries a July 2025 quote date for it, and I do not put a stale price in a table dated to one day.

Equity has been refilled twice since 2022 and drained once between

Five balance sheet dates

Total equity at the five reporting dates runs KRW 68.91 billion, KRW 235.11 billion, KRW 164.86 billion, KRW 137.43 billion, and KRW 331.64 billion. That is a jump, a two-year slide, and then a larger jump. Subtracting each period’s net result from each period’s equity change tells me how much arrived that did not come from earnings.

Billions of Korean won. Every figure in the third and fourth columns is derived by me from the filed equity balances and net results. The fourth column is a residual, so it captures share issuance together with anything else that moves equity without passing through profit, including share-based compensation and translation differences on the overseas subsidiary. I am not claiming it is all cash.

Summed across the four steps, roughly KRW 470.7 billion arrived from outside since the end of 2022, against KRW 208.0 billion lost through the income statement over the same span (both derived). The check closes: KRW 68.91 billion plus KRW 470.71 billion less KRW 207.98 billion gives KRW 331.64 billion, which is the reported figure.

The refill and the burn move together

Line the two tables up and something uncomfortable appears. The largest single-year cash burn in the record, 2024, is the year immediately after the largest arrival of outside capital. The two smallest inflow years, 2024 and 2025, bracket the period when the cumulative gap started closing.

I am careful about what that does and does not show. It is not evidence that capital availability causes spending, and there are obvious alternative readings, including that a company raises when it plans to invest and invests the year after. What it does establish is that in this business the funding schedule and the spending schedule are not independent series, and that reading either one alone gives an incomplete answer.

The most recent refill, and one thing I could not source

The KRW 235.58 billion that arrived in the first half of 2026 is mostly a rights offering, which Korean coverage of the subscription result put at approximately KRW 211.5 billion, with the stated purpose being the repayment right on that May 2024 convertible bond. That leaves the balance of the residual to share-based compensation, translation effects, and anything else, and I could not reach the note-level detail to split it.

The KRW 203.00 billion residual in 2023 I could not source at all. Something arrived that year that was more than five times the company’s revenue for the year, and I did not find the filing behind it. I am printing the number because it is arithmetic from two disclosed balances, and printing my failure to explain it because that is the honest state of my work on this name.

Lunit stock and a revenue base that is 95% overseas

What that mix does to a dollar-based reader

The company’s own first-half release puts roughly 95% of revenue outside Korea, with North American cancer screening revenue up 30% year over year. For a US-based reader that cuts two ways at once. A Korean medical AI company whose revenue does not depend on Korea’s national reimbursement schedule is easier to underwrite than one that does. At the same time nearly all of that revenue is exposed to dollar strength and to whatever happens to imaging reimbursement in the United States, and the company reports in won, so both effects land in the same reported line without being separable from outside.

It also complicates my own cumulative table. Four and a half years of won-denominated revenue from a base that is overwhelmingly dollar-linked carries an exchange rate history inside it that I have not stripped out. The gap I am tracking is a won gap, and part of its movement is currency. I have no way to quantify how much without segment-level currency disclosure.

The Volpara question I left open

The acquisition of Volpara Health Technologies is what put much of that overseas base in place. I wanted quarterly revenue by segment to see whether the acquired business is growing faster or slower than the rest, and semiannual disclosure does not reach that level. So I cannot tell you which part of the screening growth is the acquired base and which is the original one. That gap matters more than most of the gaps in this piece, because it is the difference between a business compounding and a business that bought a step up.

What owning Lunit stock requires from a US brokerage account

No receipt, one board, and a liquidity figure I could not extend

There is no American depositary receipt for this company that I could locate. The broad Korea exchange traded funds that US investors typically reach for track large capitalization indices, and I did not verify whether this specific name appears in any of them, so I am not going to assert either way.

The practical constraint is size. On September 3, 2026 (Thu) the entire company was worth roughly USD 469 million and that day’s volume was 251,089 shares, about USD 1.6 million traded at the closing price (derived). I was unable to obtain a longer daily volume series, so I will not characterize typical liquidity beyond that single session. For an institution that figure is the whole story. For an individual account it is a reason to size orders carefully, though hardly a barrier.

Thirteen places where the case against Lunit stock lives

Seven from the two tables above

1. Cumulative operating cash consumed since 2022 exceeds cumulative revenue by 4.58%, and that has been true in every one of the five periods (derived).
2. The gap widened in each of the first three years, from KRW 39.15 billion to KRW 50.55 billion to KRW 62.07 billion. Only two reporting periods of narrowing exist so far, which is a short record to extrapolate from.
3. Equity arriving from outside earnings totals roughly KRW 470.7 billion since the end of 2022, against KRW 208.0 billion lost through the income statement over the same span (both derived).
4. The worst burn year in the record follows the largest inflow year. Funding and spending are not independent series here.
5. Absolute cash consumption has not fallen in a straight line, running between roughly KRW 36 billion and KRW 66 billion across the four full years. The improvement is in the ratio, not the level.
6. The KRW 203.00 billion that arrived in 2023 I could not source to a filing. That is the single largest number in this piece that I cannot explain.
7. The May 2024 convertible bond carried a holder repayment right that came due in 2026, and Korean press tied the equity raise to it. Capital raised to retire an obligation is not capital raised to fund growth.

Six from the screens and the peer group

8. The sales multiple lands third of five once divided by growth, so the multiple offers no margin either way and the case has to rest on something else.
9. Lunit is the only row in my peer table whose growth figure had to come from a half-year comparison instead of a trailing window. That inconsistency favors nobody in particular and weakens the ordering.
10. Two of the four peers sell hardware, so the group is not a clean software comparison and part of the ordering may reflect business model as much as valuation.
11. Two Korean market data screens carry no analyst estimates at all for this name, one of them stating no published opinion within three months.
12. The one overseas screen carrying an analyst view shows a single analyst with a sell rating. I am recording both the absence and the single view, and adopting neither.
13. The same company shows a 6.94 or a 7.67 sales multiple on the same day depending on which revenue window the screen uses, which is a reminder that the multiple I chose is a choice.

Nine of the thirteen count against the company, three are gaps in what I could obtain, and one cuts both ways. I have written much shorter opposing cases for companies I liked less.

One classification I had to change

Equity issuance sat in my head under corporate actions, in the same drawer as dividends and buybacks. Things a company does occasionally, that I check for, and that adjust my per-share arithmetic when they occur. That filing works fine for profitable businesses.

For a company like this one it is the wrong drawer entirely. Across the five periods in my first table, external capital is not an occasional event punctuating the operating story. It is that story’s funding source, arriving roughly on the schedule the burn requires. Filing it under corporate actions meant I kept treating each arrival as news, when the pattern was the thing worth reading. I have moved it. In a business that consumes cash, issuance belongs next to revenue, and I now build the equity history before I open the income statement.

What would prove me wrong, and what I do then

Condition one, written on the two running totals

If the 2026 annual report shows cumulative revenue since 2022 exceeding cumulative operating cash consumed since 2022, I am wrong about how far this business is from carrying itself.

Checking whether it can fire: the remaining shortfall is KRW 10.18 billion, and the half just reported closed KRW 23.93 billion of it. The second half needs to produce revenue exceeding cash consumption by KRW 10.18 billion, which is 43% of what the first half delivered on that measure (derived). It is not already true, and it is comfortably reachable, which makes it a soft condition, not a hard one. I am fine with that, because a soft condition that fires on schedule tells me the trend is real and a soft condition that fails tells me something broke.

The judging document is the annual report’s income statement and cash flow statement, both in the standard format, neither subject to the company’s own definitional choices.

If it fires: I stop reading this name as a cash question and start reading it as a margin question, which means rebuilding the peer table on operating margin instead of sales.

Condition two, written on the share count

If the 2026 annual report shows total shares outstanding still at 74,458,180, with no further issuance, I am wrong about how quickly this company needs to come back for capital.

Checking whether it can fire: KRW 235.58 billion arrived in the second quarter and the whole first half consumed KRW 21.83 billion of operating cash, so there is ample room for this to hold through year end. It is a real possibility, which is what makes it worth writing down instead of assuming.

The judging document is the share count on the cover of the annual report, which is about as free of interpretive discretion as a disclosed number gets.

If it fires: I stop treating the share count as a moving denominator and rebuild every per-share figure on a fixed base, which is the only route by which this company becomes comparable to my peer table on anything other than sales.

Expiry, and what happens if neither can be judged

Both conditions expire with the 2026 annual report in March 2027. If a restatement or a segment reclassification makes either comparison impossible, I will record it as undetermined and say so, because a condition that could not be judged and a condition that did not fire are different outcomes and I do not want to read the first as the second a year from now.

I have looked at other Korean companies where the choice of financial statement line decided the conclusion, including one where reported revenue grew faster than the unit it was supposed to track, and a loss-making Korean name I did decide to hold. What separates those two cases from this one is what the losses were buying, and here I can measure what they cost before I can measure what they bought.

Prices and multiples reflect the September 3, 2026 (Thu) close as I checked them at the time of writing. Korean won is the reference currency throughout, and USD conversions are approximate at roughly KRW 1,359.3 per dollar on that same date, per the Seoul market close reported by Korean financial press. Financial figures follow consolidated statements filed with Korea’s electronic disclosure system, and anything marked derived is my own calculation, never a figure lifted from the source.

Sources consulted: the company’s first-half results release, August 11, 2026, in Korean · Cheongnyeon Uisa on the first-half numbers, in Korean · Digital Today on the revised issue price, in Korean · Digital Times on the offering and the guidance, in Korean · Bizhankook on the convertible bond repayment right, in Korean · Newsway on the subscription result, in Korean · WiseReport company screen · Alpha Square company screen · Investing.com company screen. Korean-language sources are paraphrased in translation, never quoted directly.