Margin Expansion Finally Comes From Somewhere Other Than Delivery, and the Board Adds Another US$500 Million

Results vs. Consensus

Luckin reported before the U.S. open on Tuesday, November 3, with the call at 8:00 AM ET. The currency of record is renminbi; U.S. dollar figures below are the company's own convenience translations at RMB6.7110 to US$1.00, the rate stated in the release. Every income-statement figure in this report traces to the Form 6-K Exhibit 99.1 statements rather than to figures spoken on the call, which for this issuer pass through live interpretation.

Consensus is the midpoint of three independent reads that cluster at US$0.82 to US$0.83 for non-GAAP earnings per ADS. Revenue consensus rests on a single provider, and a second source described revenue only as largely in line, so we treat the 1.8% revenue shortfall as indicative rather than as a confirmed miss. Luckin issues no quantitative guidance, and no provider we cross-checked publishes a consensus for operating income, net income or cash flow at this name, which is why those rows carry a year-over-year comparison instead of a variance.

Year-over-Year Comparison

Q3 2025 self-operated revenue and GMV are derived from the growth rates the current release discloses (+16.8% and +19.7%), because our archived Q3 2025 release capture carries self-operated revenue only to the nearest RMB0.1B. Q3 2025 store-level operating profit of RMB1,941.7M is similarly derived from the disclosed +55.5% growth and reconciles to the 17.5% prior-year margin the release prints.

Sequential Comparison (vs. Q2 2026)

Revenue per self-operated store is our calculation: self-operated revenue divided by the average of opening and closing self-operated store counts. It is not a company-disclosed metric.

Quality of Beat/Miss

• Revenue: The 1.8% shortfall against the single-provider consensus is small in absolute terms, roughly RMB333M, and to the extent it is real it is demand-side rather than a mix or currency artifact. There was no acquisition, no divestiture and no FX translation effect inside a renminbi reporting entity. The composition matters more than the variance: freshly brewed drinks grew 17.0% in line with the total, other products grew 69.0% off a small base to 5.9% of revenue, and the "others" line, which carries delivery fees billed to self-operated customers, fell 32.0% as orders shifted to pickup. GMV grew 19.7% against reported revenue growth of 17.0%, so system-level demand grew faster than the income statement shows. The company does not reconcile that 270bp spread, and on our estimate the gap sits in the partnership channel, where Luckin books wholesale revenue against retail transaction value.
• Margins: High quality, and higher quality than the headline. Non-GAAP operating margin of 17.0% strips out RMB269.5M of share-based compensation, which itself grew 80.7% year over year and now runs at 1.51% of revenue against 0.98%, plus RMB21.6M of deal-related expense for the new long-term equity investment. Even on GAAP, which absorbs both, operating margin expanded 378bps. Nothing in the quarter looks like a one-time benefit: there was no asset sale, impairment reversal or litigation credit, the "losses and expenses related to Fabricated Transactions and Restructuring" line is down to RMB1.1M, and interest and investment income actually fell to RMB41.8M from RMB54.1M. The margin came from the stores.
• EPS: Operational, and achieved into a tax headwind. The effective rate rose to 33.0% from 30.5% a year ago and 30.3% last quarter, which cost roughly RMB70M of net income against a flat-rate counterfactual, and non-GAAP net income still grew 50.9%. Buybacks helped at the margin rather than carrying the result: weighted average basic shares fell 1.8% year over year, so roughly 2 points of the 48% basic GAAP per-ADS growth is share count and the rest is earnings. One presentational caution. The printed non-GAAP figure of US$1.04 per ADS is eight times a per-share number rounded to US$0.13, and the release's own footnote allows up to US$0.04 of rounding per ADS. Dividing non-GAAP net income of US$319.6M by 315.7M diluted ADS gives US$1.01. The beat is +25.3% on the printed figure and +22.0% on the exact arithmetic, so the conclusion does not change, but the printed number flatters it by about three points.

Cost Structure and the Anatomy of the Margin Move

This is the table that carries the quarter. Read the two right-hand columns against each other and the whole investment debate of the past three quarters resolves.

Year over year, the margin story is still entirely a delivery story and then some. Delivery gave back 890bps while the other six lines took 512bps of it away, netting to the 378bp improvement in the expense ratio. On that view the bear reading of the past two quarters was correct: the recovery was an accounting migration driven by platform subsidies withdrawing, and the underlying cost structure was deteriorating beneath it.

Sequentially, that reading no longer holds. Delivery contributed 19bps of a 204bp improvement. Every other line except the flat store-preopening line moved the right way, led by store rent and other operating costs at 81bps, which is the line that carries labour, rent and utilities and is therefore the purest read on volume leverage across a fleet that grew 7.2% inside the quarter. Sales and marketing fell 37bps sequentially because revenue grew faster than spend quarter over quarter. Year over year the spend still rose 30.0%, which the CFO attributed to "higher traffic-related investment following the subsidy war," and the ratio rose 54bps. Materials improved 24bps even with green coffee costs still unaddressed.

"Store rental and other operating costs as a percentage of total net revenue was 22% compared to 20% in the same period of 2025. The difference mainly reflects the greater economies of scale in the prior year period when overall volumes were boosted by the subsidy." — Jing An, CFO

That framing is the correct one for the year-over-year comparison and it quietly concedes the sequential point: the prior-year quarter enjoyed artificial volume leverage from subsidised delivery, and the business is now rebuilding that leverage organically. The 81bp sequential improvement is what rebuilding looks like.

Assessment: The single most important question we posed at Q2 was whether Luckin could expand margin without further help from the delivery line, because roughly 300bps of reversion was all that remained and the recovery was, in our words, a one-legged stool. Delivery reverted 19bps and margin expanded 204bps. The stool has a second leg. That passes the test we set. It does not yet prove the leverage is structural, because Q3 is the seasonal peak and the non-delivery cost ratios improved more from Q2 to Q3 in both 2024 and 2025; the fourth quarter, when the season works against the business, is the cleaner read. Even so, it is worth more than the revenue shortfall costs.

Segment Performance

Luckin reports two revenue channels rather than operating segments, and discloses profitability for one of them. Store-level operating profit exists only for the self-operated fleet; the partnership channel's economics must be inferred.

The channel rows do not sum to total net revenues because the company presents revenue two ways. Product sales of RMB13,648.7M plus partnership revenue of RMB4,242.7M equals the total; self-operated store revenue of RMB12,939.8M is a subset of product sales that excludes e-commerce and offline sales. The "Others" row is a component of product sales shown separately because of its signal value on delivery mix.

Self-Operated Stores

The self-operated fleet is where the quarter was won and where the open question lives. Revenue of RMB12,939.8M grew 16.8% on a store count that grew 34.5%, which is the arithmetic of the same-store decline. Store-level operating profit grew 55.5% to RMB3,019.3M, and the margin reached 23.3% against 17.5% a year ago and 21.3% last quarter.

Set that 23.3% against the series we hold. Store-level margin ran 19.8% in Q4 2024, 17.0% in Q1 2025, 21.5% in Q2 2025 on the restated basis, then fell through 17.5%, 15.0% and a 13.6% trough in Q1 2026 before recovering to 21.3% and now 23.3%. The current print is the best of the eight quarters on our file. Those eight quarters hold no like-season quarter from before the subsidy war, because Q2 2025 already carried the subsidy-driven delivery mix, so we make no claim that the margin sits above its pre-war level, and no all-time claim.

"Store level operating margin was not only notably higher than a year ago, but also largely back to pre-subsidy levels." — Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

That description is accurate. On the same call managers made the harder version of the point themselves, noting that against the third quarter of 2024, the last like-season quarter before the subsidy war, the delivery mix remains much higher and the store network has nearly doubled, yet store-level margin has returned to the earlier level. The live interpretation garbled the end of that sentence, so we do not quote it, but the comparison is the right one, and the growth rates in our archived Q3 2025 release imply a Q3 2024 margin at about today's level on the pre-2026 definition. Returning to a pre-war margin with a delivery mix still materially above the pre-war baseline and a fleet nearly twice the size is the achievement, and it is enough.

Assessment: A 23.3% store-level margin on a fleet growing 34.5% is the most important number in the release. It says new stores are not structurally destroying unit economics, because if they were, adding 6,520 self-operated locations in a year while average unit volume fell 14.8% would have dragged the margin down rather than to a cycle high. The disclosure to settle the cannibalisation question still does not exist, but the circumstantial evidence now runs clearly one way.

Partnership Stores

The partnership channel is the one place in the print where the trend genuinely deteriorated beyond the base effect. Revenue grew 11.7% to RMB4,242.7M on a store count that grew 30.7% to 13,507, so revenue per partnership store fell 16.2% to roughly RMB325,000 from RMB388,000. The channel's share of total revenue slipped to 23.7% from 24.9%.

The composition explains part of it. Partnership revenue comprises sales of materials at RMB2,811.2M, delivery service fees at RMB606.3M, profit sharing and royalty fees at RMB580.8M, equipment sales at RMB214.5M and franchise and other service fees at RMB30.0M. The CFO identified the drag precisely.

"This growth was mainly driven by higher cup volumes at partnership stores, partially offset by reduced delivery service fees due to a lower delivery mix." — Jing An, CFO

So the same delivery normalisation that rebuilt the self-operated margin cut the wholesale revenue Luckin earns from passing delivery through for its partners. That is a mechanical and largely benign explanation for the revenue line. What it does not explain is the implied capture rate. Total GMV grew 19.7% while partnership revenue grew 11.7%, and on our approximation, which treats self-operated revenue as a proxy for self-operated GMV, non-self-operated GMV grew roughly 25% while Luckin's wholesale take on it fell from about 61% to about 55%. We flag that as directional only, because self-operated revenue is not identical to self-operated GMV and the company publishes no channel GMV split.

Assessment: Neutral to mildly negative, and under-disclosed. A franchise channel whose unit revenue falls 16% while its unit count rises 31% deserves more than one sentence, particularly when its retail volume is apparently growing faster than the revenue Luckin recognises from it. If the mix shift is the whole story the line should stabilise as delivery stops normalising, which on this quarter's evidence is already happening. If it is not the whole story, the partnership channel is quietly becoming a lower-take business and nobody has asked.

Key KPIs

Revenue per monthly customer and revenue per store are our calculations, not company-disclosed metrics. Same-store sales figures are stated on the definition Luckin adopted in Q1 2026, under which prior periods presented were restated; figures published before that change differ and should not be mixed with this series.

Store Network and International

Net new openings of 2,599 split 2,543 in China including Hong Kong, 14 in Singapore, 40 in Malaysia and 2 in the United States. China ended the quarter at 38,630 stores comprising 25,277 self-operated and 13,353 partnership locations. The overseas fleet reached 279: 103 self-operated in Singapore, 22 in the United States and 154 partnership stores in Malaysia.

Two facts about this expansion deserve more attention than they received. First, the pace barely slowed in response to the same-store decline. Luckin added 2,599 stores against 2,714 last quarter, which is a 4% reduction in net pace while the comparable-store metric deteriorated 580bps. Second, the global store count of 38,909 now sits within 2,698 of Starbucks' 41,607 at its 2026 fiscal year end, and Luckin added 2,599 in the quarter against 303 for Starbucks. On the current trajectory Luckin passes Starbucks by total store count sometime in the first half of 2027.

Singapore crossed 100 stores and is now the country's second largest coffee chain by store count.

"Our store level economics in Singapore have largely proven out and the store performance continues to improve." — Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

Assessment: The international business is now three years into validation and still carries no disclosed revenue, margin or profitability figure for any market. "Largely proven out" repeats the kind of qualitative assurance management has given about Singapore before, with no figure attached. At 279 stores, or 0.7% of the fleet, international is not yet a modelling input, and the company's own reticence suggests it should not be treated as one. The Starbucks crossover is a narrative milestone rather than a financial one, but for a pink-sheet issuer seeking a U.S. main-board relisting, narrative milestones are not worthless.

Key Topics & Management Commentary

Overall Management Tone: Confident on profitability and conspicuously philosophical on demand. Where the Q2 call conceded the same-store pressure and framed the third quarter as a base problem to be endured, this call reframed the metric itself as a management dial rather than a scoreboard, which is a more assured posture and a more evasive one at the same time. The two places management was least convincing were both places it has been unconvincing for several quarters: the absence of any quantified store runway, and the silence on why a company with RMB14.3B of liquid resources keeps borrowing.

1. The Margin Engine Changed Hands

The central event of the quarter is that delivery stopped being the source of margin improvement and margin improved anyway. Delivery expense fell to 10.00% of revenue from 10.18%, which is 19bps of reversion against the roughly 300bps we estimated remained available at Q2. On the mix-reversion thesis the quarter should have delivered about 20bps of margin. It delivered 204bps.

"In the third quarter, we delivered solid revenue growth against a high comparison base from last year's subsidy war. More importantly, we saw a further improvement in profitability." — Jing An, CFO

In absolute renminbi, delivery expense fell 38.1% year over year to RMB1,789.0M, which the release attributes to "a notable shift in order mix from delivery to pick-up." That shift is now largely complete. The 10.0% run rate is above the roughly 7% pre-war baseline, so some reversion remains, but the sequential improvement has slowed to 19bps from 72bps a quarter ago.

Assessment: This is the confirm the thesis needed and the one we explicitly set as the test. It shows the earnings base no longer depends on give-back from the delivery line, and it removes the bear argument that the recovery was a migration between two expense lines. Whether the remaining leverage is structural or seasonal is the fourth quarter's question, because Q3 is the seasonal peak and the non-delivery cost ratios improved more from Q2 to Q3 in both 2024 and 2025 than they did this year. It also removes a crutch: from here, margin expansion has to come from volume leverage and cost discipline, because the delivery tailwind is nearly spent.

2. Same-Store Sales at (11.1)%: Worse, and Not a Record

The comparable-store metric fell to (11.1)% from (5.3)%, a third consecutive negative quarter and a 580bp sequential deterioration. On the definition Luckin adopted in Q1 2026, the series now reads +9.2%, +13.8%, +14.3%, +1.3%, (0.1)%, (5.3)%, (11.1)% across the seven quarters the release discloses. This is the steepest print of the current cycle.

It is not, however, the worst in the company's recent history, and that matters for how an investor should read it. Our own source archive records same-store sales of (13.1)% in Q3 2024 and (3.4)% in Q4 2024, with the full year 2024 at (16.7)%, all on the pre-2026 definition and therefore not strictly comparable. The last time this metric sat near (13)%, it was (3.4)% a quarter later, positive the quarter after that and +14.3% four quarters later. The metric has been here before and it came back.

"As platform subsidies fell sharply this year, third quarter same-store sales for self-operated stores were significantly impacted, declining 11.1% year-over-year. However, we have also seen the decline narrow month by month since July with the pace of improvement continuing to pick up. This was fully in line with the trend we had previously communicated." — Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

The intra-quarter trajectory is the useful disclosure. July was the trough, September the best month, and management says the improvement picked up further as the fourth quarter began. The comparison base cooperates from here: the prior-year prints that become comparatives are +1.3% for Q4 2026 and (0.1)% for Q1 2027, against the +14.3% that this quarter lapped.

Assessment: Our Q2 commitment list put "whether same-store sales stops deteriorating in Q3" as the single most important confirm, and on the printed number it failed. On the intra-quarter path and the fourth-quarter commentary it partially passed. We score it partially met, a miss on the printed number with a mitigating trajectory, and note that the arithmetic of the base makes Q4 the real test rather than Q3.

3. Management Reframes Same-Store Sales as a Dial They Choose Not to Turn

The most substantive new idea on the call was a reframing of what the metric means for a business at Luckin's stage, delivered at length in answer to the only question about it.

"Looking at our own performance over the past few years, we believe Luckin's same-store sales performance may follow a wave-like pattern as we continue to grow rapidly. When competition is intense and we are opening stores at a fast pace, same-store sales may come under some short-term pressure. But as the pace of store openings moderates and consumer demand continues to grow, we would expect same-store sales growth to quickly improve. We have seen this before. For example, from the second half of 2023 to the first half of 2024, we accelerated store openings, which put some pressure on the same-store sales performance. Then in the second half of 2024, as the pace of store expansion moderated, same-store sales growth quickly improved and turned positive by year-end." — Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

Read that carefully. Management is saying that comparable-store sales are substantially a function of its own opening pace, that it has demonstrated the relationship once before, and that it could improve the metric by slowing down. It then declines to do so: "our top priority remains growing market share over the long term while maintaining healthy store economics."

The historical claim checks out against our archive, with one correction on timing. The company opened aggressively through late 2023 and early 2024, same-store sales ran deeply negative across 2024, and the metric improved from (13.1)% in Q3 2024 to (3.4)% in Q4 2024 before turning positive in Q1 2025. "Positive by year-end" may describe the exit month, but the quarterly figure turned a quarter later. The precedent is real.

Assessment: This is the most important thing management said, and it cuts two ways. It is a genuine and verifiable defence of the current print, and it reframes (11.1)% as a chosen cost of share capture rather than a symptom of demand failure. It is also an admission that the metric investors watch most closely is partly an output of a decision management makes, which makes it less informative as an independent read on business health. Investors who want the same-store line to turn should understand they are asking for slower growth.

4. Average Unit Volume: The Dilution Gap Narrows, the Decline Does Not

The bear point we added to the thesis at Q2 was average unit volume dilution: revenue per self-operated store was falling faster than the same-store metric, and the gap represented the drag from new stores opening below system average. The gap narrowed this quarter. The underlying decline did not.

Revenue per self-operated store, on average store counts, was roughly RMB526,700 against RMB507,900 last quarter and RMB618,100 a year ago. Year over year that is (14.8)%, against a same-store decline of (11.1)%, so the dilution gap narrowed to 3.7 points from 4.2 points. That narrowing is the genuine improvement, and it is the only one. The metric did rise 3.7% sequentially, but the second-to-third-quarter step is seasonally positive and rose 10.2% in the same period last year, so a 3.7% step is roughly a third of last year's seasonal lift rather than evidence of underlying recovery. The year-over-year decline also widened, from (9.5)% last quarter to (14.8)%.

"Importantly, even as our store base continues to grow, stores opened this year have maintained healthy operating performance." — Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

That is the fourth consecutive quarter in which management has asserted new-store health without publishing a single figure to support it. There is still no disclosure of new-store payback, first-year revenue, or how new-store volumes compare with the mature base.

Assessment: The direction improved and the disclosure did not. A narrowing dilution gap plus a cycle-high store-level margin is about as much indirect evidence as can be assembled that new stores are economic, and it is still indirect. We keep this bear point open at EMERGING rather than retiring it, because the company has had four quarters to answer it with data and has chosen assurance instead.

5. The Buyback Escalates Twice in Nine Weeks

Capital return is now the clearest thing about this company. The April program authorised US$300M over twelve months. By August 31 it had spent US$287.2M, or 95.7%, which forced a September 1 upsize to US$500M. By September 30 cumulative repurchases reached US$391.0M for 95.8 million Class A ordinary shares, equivalent to 12.0 million ADSs. On the print the board authorised a further US$500M for twelve months, commencing when the existing authorisation is exhausted.

"Taking into account our financial position and the business outlook and to better reward shareholders, our Board has approved a new 1-year share repurchase program with a total authorization of USD 500 million. This also reflects our firm belief in Luckin's long-term value." — Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

The arithmetic is material. US$109M remains on the existing authorisation and US$500M on the new one, so US$609M is authorised against a market value of roughly US$10.8B, or 5.6%. The 12.0 million ADSs already retired equal 3.8% of the weighted average basic ADS count, and the balance sheet now carries RMB2,654.1M of treasury shares against nil at year end. Weighted average basic shares fell 47.3 million, or 1.8%, year over year.

Assessment: Our Q2 commitment list asked whether the board would re-authorise once the remaining US$105M was spent, likely inside the fiscal year. It did so twice, and the execution price has risen with the stock rather than being abandoned when the discount narrowed, which is the harder test of intent. The blended US$32.58 is also a usable anchor: management has bought nearly 12 million ADSs at that level, which is a reasonable floor reference for an outside buyer.

6. Borrowings Climb 67% While Cash Builds, and the Question Is Reframed Rather Than Answered

Short-term bank borrowings reached RMB3,250.0M at September 30 against RMB1,945.0M at June 30 and nil at December 31, 2025. Over the same nine months liquid resources, comprising cash, restricted cash, term deposits and short-term investments, rose to RMB14,333.4M from RMB8,964.4M. Interest and financing expenses were RMB13.0M in the quarter against RMB2.0M last quarter, implying a cost of roughly 2% annualised on average balances.

"Our net cash position remained strong at RMB 11.1 billion. This leaves us well placed to invest in our business expansion and continue to deliver shareholder returns." — Jing An, CFO

Note the change in presentation. At Q2 the CFO linked the first borrowings of the cycle to shareholder returns and left the rationale unexplained. This quarter the CFO introduces a net cash framing, which arithmetically disposes of the question without answering it. Our best read remains an onshore and offshore treasury bridge: the cash is substantially in China and the buyback is paid in dollars offshore. That is a reasonable explanation, it is cheap at 2%, and it is still our inference rather than a disclosure.

Assessment: Economically immaterial, governance-relevant, and now 67% larger than when we first flagged it. A company that generated RMB4.1B of operating cash flow in a single quarter is not borrowing because it is short of money. The objection is not to the facility but to two quarters of a growing debt balance that no filing or call has explained, at a company whose history makes unexplained financing items expensive to ignore.

7. The RMB644 Million Nobody Mentioned

A new line appeared on the balance sheet: long-term investment, net of RMB644,256 thousand, or US$96.0M, against nil at December 31, 2025. The non-GAAP reconciliation carries a matching first-ever adjustment of RMB21,636 thousand for "deal-related expenses for long-term equity investment," which management describes in its own definitions as a "one-off item."

Neither the release nor the call says what was purchased. There is no counterparty, no sector, no stake size, no strategic rationale and no mention in the CEO's or CFO's prepared remarks. The only acknowledgement of its existence is the expense add-back.

Context from the quarter's other filings is suggestive without being conclusive. On September 1 the board upsized the buyback. In early September the controlling shareholder and Mubadala made a reported US$1 billion joint investment. On September 29 Luckin appointed Mubadala's Director of Private Equity China to its board, taking the board to ten directors of whom three are independent, and the chairman told media the company was "currently looking at" Gulf-country markets. A first-ever long-term equity investment, with deal expenses, landed in the same quarter. We cannot connect these facts from any disclosure we hold, and we are not asserting that they are connected.

Assessment: This is the most material undisclosed item in the print. RMB644M is small against RMB38.3B of total assets, so the sizing is not the concern. The concern is that a company rebuilding credibility after a fraud deployed capital outside its operating business for the first time, incurred deal expenses it chose to adjust out of non-GAAP earnings, and said nothing about it on a call where the operator closed questions after two analysts. We expect this to be explained in the 20-F and we will treat the answer as a governance datapoint.

8. Input Costs: Materials at 38.3% and a Fifth Silent Quarter

Cost of materials rose 23.7% to RMB6,854.2M and sits at 38.31% of revenue, up 207bps year over year though down 24bps sequentially. The CFO gave the cleanest attribution yet.

"Cost of materials as a percentage of total net revenue was 38% compared to 36% in the same period of last year, mainly due to changes in average selling prices resulting from the shift in delivery mix." — Jing An, CFO

That is a price story rather than a cost story, and it is internally consistent: delivery orders carried a higher average selling price per cup, so as the mix shifted to pickup, realised price fell and a stable input cost became a larger percentage of a smaller per-cup revenue. It also means the materials ratio should stabilise as the mix shift completes, which the sequential improvement supports.

What is still missing after five quarters is any treatment of green coffee bean costs: no hedging policy, no quantification of the bean component, no expected duration, no commentary on crop or futures exposure. The line is 38.3% of revenue and the largest single item on the income statement.

Assessment: The pricing explanation is credible and reduces our concern about the ratio's trajectory. It does not reduce our concern about the disclosure. A vertically integrated roaster with four roasting facilities and 155,000 tonnes of planned annual capacity has a view on bean costs, and five quarters without articulating it is a choice.

9. Customers, Cups and the Product Engine

Average monthly transacting customers reached a record 123.0 million, up 9.6%. The company added more than 28 million new transacting customers in the quarter and the cumulative base passed 520 million. Revenue per monthly customer rose 6.8% to roughly RMB145.

The deceleration deserves naming: monthly customer growth ran +22.9% last quarter and +9.6% this quarter, because the comparison now laps the subsidy-war peak of 112.3 million. On a sequential basis customers grew 9.1%, which is the healthier read.

"What's even more encouraging is that average cups per transacting customer has continued to increase this year, reaching a new high in the third quarter as well." — Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

On product, the quarter brought 41 freshly made beverages and more than 20 snack items, an Ethiopian Gesha cold brew line positioned to reinforce coffee credibility, and a Thai Milk Tea series launched in August that sold more than 13 million cups in its first week. Brand activity included World Cup sponsorship of the Spanish and Portuguese national teams and a Disney villains collaboration.

Assessment: Positive and the cleanest counter to the cannibalisation thesis available. A record customer count, a rising cup count per customer and rising revenue per customer together describe deepening engagement rather than thinning demand. The frustration is that "a new high" in cups per customer is asserted without a number, for the second consecutive quarter, which leaves the single best metric for separating base effects from demand damage unquantified.

10. The Partnership Channel's Falling Take

Covered in detail under Segment Performance, and listed here because it is the one trend in the quarter that worsened for reasons management did not fully address. Partnership revenue grew 11.7% on a 30.7% larger partnership fleet, revenue per partnership store fell 16.2%, and total GMV grew 19.7%. The delivery-service-fee explanation accounts for part of the gap and not evidently all of it.

Assessment: Watch item, not yet a thesis item. At 23.7% of revenue the channel is large enough that a persistent decline in unit economics would matter, and small enough relative to the self-operated margin recovery that it did not change this quarter's result.

11. Cash Generation and the Balance Sheet

Operating cash flow nearly doubled to RMB4,131.8M from RMB2,069.2M, and ran RMB7,547.8M for the nine months against RMB5,526.3M. Investing outflows of RMB4,142.2M almost exactly offset the operating inflow, leaving cash and restricted cash down RMB50.1M in the quarter, which is a term-deposit and short-term-investment rotation rather than a spending event: liquid resources including those instruments rose RMB3,407.6M sequentially. Financing was a net outflow of only RMB21.0M despite US$195.9M of buyback, because the incremental borrowings funded most of it.

Total assets reached RMB38,327.1M from RMB30,301.6M at year end and shareholders' equity RMB17,469.3M from RMB15,624.7M, with RMB1,514.7M of convertible senior preferred shares still sitting in mezzanine equity and outside the ordinary share count.

Assessment: The cash engine is working harder than the income statement suggests, with operating cash flow at 1.9x non-GAAP net income on heavy store investment. That is the financial capacity behind both the opening pace and the buyback, and it is the reason the borrowings question is a disclosure complaint rather than a solvency one.

12. International and the Starbucks Crossover

Covered under Store Network. The headline is that 38,909 stores puts Luckin within 2,698 of Starbucks globally, closing at roughly 2,300 per quarter.

Assessment: A milestone with real option value for a company pursuing a U.S. main-board relisting, and no modelling consequence this year.

13. Relisting: Said Off the Call, Not On It

The U.S. relisting has now gone four consecutive earnings calls without a mention, including this one. It did not, however, go unmentioned in the quarter. In late September the chief executive told media that the company is "actively pushing the process of relisting on a U.S. mainboard," subject to Chinese regulatory approval and with no timeline.

That is a materially different position from the dormancy we recorded at Q2, when we marked the catalyst latent after three silent quarters. The company also now has a sovereign wealth fund represented on its board, a second sovereign investor holding 6.4% disclosed earlier in the year, and 22 U.S. stores open.

Assessment: Reactivated, and still uncapitalised in the price. The oddity is the channel: an intention material enough to state to a broadcaster is material enough to state on an earnings call, and management did not. We treat the catalyst as live but unscheduled, and we do not underwrite it. If it arrives, the re-rating from a pink-sheet quotation to a main-board listing is worth more than any single quarter's operating news.

Guidance & Outlook

Luckin issues no quantitative guidance. It publishes no revenue, margin, earnings or store-opening targets, and it did not start this quarter. The guidance table that would normally sit here has no inputs, and the forward markers available are qualitative and narrow.

The one forward number worth constructing is the implied fourth quarter. Nine-month revenue of RMB45,772.6M grew 25.4%, and nine-month non-GAAP net income of RMB4,584.7M grew 31.1%. The fourth quarter is seasonally the weakest: in 2025 it delivered non-GAAP net income of RMB698.6M, which was 49% of that year's third quarter, on a store-level margin that fell to 15.0% in the winter off-peak.

Implied quarter-over-quarter ramp: no guidance exists to back-solve, so the useful exercise is the reverse. If Q4 2026 non-GAAP net income lands between 55% and 75% of this quarter's RMB2,144.8M, full-year non-GAAP net income falls between roughly RMB5,760M and RMB6,190M, which is 37% to 48% growth on the RMB4.2B of 2025. The seasonal step down should be shallower than last year's because the 2025 fourth quarter absorbed elevated delivery costs at 12.8% of revenue against the current 10.0%.

Street at: published twelve-month targets we can observe cluster between roughly US$44 and US$55, with aggregator consensus in the high US$40s, against a last close of US$34.45. The rating distribution is heavily positive. Consensus for the quarter itself sat at US$0.83 per ADS on earnings and, from the one provider that publishes it, roughly US$2,716M on revenue, so the Street was modelling slightly more revenue and materially less profit than the company delivered. In each of the past two quarters it under-modelled both.

Guidance style: there is no guidance to characterise, and that absence is itself the pattern. Luckin manages expectations through qualitative directional commentary delivered on the call and through intra-quarter 6-K announcements on capital allocation. For three consecutive quarters the directional commentary on same-store sales has proved accurate, including this quarter's claim that the Q3 high base would persist, which it stated in advance. We give management credit for calibration while noting that an investor gets no numbers to hold them to.

Analyst Q&A Highlights

The operator closed the line after two analysts, citing time constraints. That makes this the third consecutive quarter with an unusually thin analyst session and the second in which questioning ended after exactly two questions. The call also suffered a speaker-line audio failure during the chief financial officer's prepared remarks, interrupting the revenue breakdown before the line was reconnected. Both questions were substantive and both went to the two live debates, so the session was informative in content even as it was constrained in length. The entries below separate the second exchange into its two distinct answers, because management used it to deliver both a trajectory update and a conceptual reframing.

The domestic store runway and whether scale now caps the opening pace

The first question went straight to the debate this coverage has flagged for two quarters: whether a fleet approaching 39,000 stores still has domestic room, and whether the opening pace must moderate as a result. The answer rejected the premise of a store ceiling rather than answering within it, arguing that the constraint is demand growth rather than a site count, and pointing to penetration and frequency gaps against mature coffee markets. No number, range or saturation framework was offered, for the second consecutive quarter in which the question was asked directly.

Q: "Congrats for the good results. My question is about the store opening upside. How should we think about the -- in the domestic market, the long-term store opening upside? And given our current already very large scale, how should we think about the pace of the store opening?"
— Michelle Cheng, Goldman Sachs

A: "The way we look at it, Luckin's long-term potential for store expansion is really driven by China's coffee demand growth instead of a fixed number of stores. As I mentioned before, coffee consumption habits are still developing in China. If you look at more mature coffee markets, there is still significant room for growth in China, both in terms of how many people drink coffee and how often they drink it."
— Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

Assessment: A consistent answer and an unsatisfying one. Rejecting a fixed store ceiling is defensible for a category at Chinese coffee's penetration level, and the supporting evidence offered on the call was the right evidence: a record customer count alongside rising cups per customer. But the question asked about pace as well as upside, and pace is the variable management controls and declined to quantify. Two consecutive quarters of qualitative answers to the same direct question is now a pattern rather than an omission, and it keeps the runway debate unresolvable from outside.

What drove the (11.1)% and the trajectory into the fourth quarter

The second and final question opened with the observation that the decline was expected given the base, then asked for drivers and for the trend from here. The answer produced the quarter's most useful piece of forward information. Management attributed the print to the July and August subsidy peak in the prior year, explained the mechanism as a loss of higher-priced delivery volume only partly offset by pickup recovery, and then disclosed an improving monthly path through the quarter that continued into the fourth quarter.

Q: "Given the high base for year-on-year comparison, the decline in same-store sales was kind of expected. Within all the incoming questions from investors, I think the first question would always be same-store sales trend. With third quarter performance broadly in line with expectations, can we understand more about the driver or any color that you can share? More importantly, how should we think about the trend from here as the impact of the delivery subsidy gradually fades? And how should we think about the same-store sales trend into 4Q and beyond?"
— Jessie Xu, JPMorgan

A: "First, July and August of last year marked the peak of the food delivery subsidy war. So the high comparison base had a greater impact on our third quarter same-store sales this year. As platform subsidies came off sharply this year, delivery volume, which typically carries a higher average selling price per cup, also dropped significantly. While pickup volume recovered, the overall impact continued to weigh on the same-store sales in terms of both cup volume and per cup average selling prices. But as we mentioned earlier, same-store sales trends improved throughout the quarter with the decline narrowing month by month from July to September. As we move into the fourth quarter, the pace of improvement has picked up further."
— Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

Assessment: The most valuable disclosure of the call, and management volunteered rather than conceded it. A monthly improving path plus an explicit statement that the improvement picked up further into the fourth quarter converts a (11.1)% headline into a decelerating drag, and it is consistent with the base arithmetic, since the prior-year comparatives fall from +14.3% to +1.3% at the quarter turn. The mechanism offered also reconciles the income statement: losing higher-priced delivery cups explains both the comparable-store decline and the 207bp rise in the materials ratio, which the chief financial officer independently attributed to selling-price mix. Two unrelated disclosures telling the same story is a reason to believe both.

Whether same-store sales is the right lens for a business at this stage

The same answer continued into a longer argument about the metric itself, which deserves separating because it is a position statement rather than a quarterly update. Management argued that comparable-store sales is a primary growth driver for mature restaurant businesses but not for an early-stage category and a rapidly scaling operator, described a "wave-like pattern" in which fast opening suppresses the metric and moderating opening releases it, and cited its own 2023 to 2024 experience as precedent. It then stated the trade-off explicitly and chose growth.

A: "So this is an important metric in the restaurant industry, but we believe its significance varies depending on the company's stage of development. For more mature restaurant business, same-store sales growth is a key driver of revenue and profit growth. But for a coffee market like China, where consumption is still in early stages of development and for a company like Luckin, which is still scaling rapidly, growth comes actually not only from the existing stores, but also from new store openings and improvements in overall operating efficiency. So we don't look at same-store sales growth in isolation or as a static measure. We take a longer view and focus on the pace and quality of growth across the business as a whole."
— Nancy Song, Head of Investor Relations, interpreting for Jinyi Guo, Co-founder and CEO

Assessment: Analytically sound and strategically convenient, which is an uncomfortable combination. The argument is correct on its own terms, and the 2023 to 2024 precedent is verifiable: the metric ran deeply negative through an aggressive opening phase and turned positive within a year as the pace moderated. What makes it uncomfortable is that the company is simultaneously telling investors the metric is partly its own choice and declining to publish the metrics that would replace it, namely new-store productivity, payback and cups per customer. An operator who asks to be judged on "the pace and quality of growth across the business as a whole" owes the disclosure to make that judgement possible.

What They're NOT Saying

1. What the RMB644.3M long-term equity investment is: A first-ever non-core investment, net of RMB644,256 thousand, appeared on the balance sheet with RMB21,636 thousand of deal-related expenses adjusted out of non-GAAP earnings, and neither the release nor the call identified the counterparty, the stake, the sector or the rationale. The add-back is the only acknowledgement that a transaction occurred.
2. Why borrowings rose to RMB3,250M while liquid resources rose to RMB14,333M: Two quarters into a growing debt balance at a company generating RMB4.1B of quarterly operating cash flow, the explanation has moved from "linked to shareholder returns" to a net cash framing that disposes of the question. Nobody asked.
3. New-store productivity, payback or first-year volumes: A fourth consecutive quarter asserting that "stores opened this year have maintained healthy operating performance" without one supporting figure. This is the exact disclosure that would settle the cannibalisation debate, and its absence is now deliberate.
4. Any store-opening number or range: Asked directly about pace for the second consecutive quarter, management offered "a competitive pace of store expansion" and nothing quantified, in either direction.
5. Cups per transacting customer: Described as "reaching a new high" for the second consecutive quarter with no level, no growth rate and no history. It is the cleanest available measure of demand depth and it remains an adjective.
6. Green coffee bean cost duration or hedging: A fifth consecutive quarter with no hedging policy, no quantification of the bean component, and no expected duration, on a line that is 38.3% of revenue and the largest expense in the business.
7. International revenue, margin or profitability: 279 stores across four markets, Singapore economics described as "largely proven out," and not one figure for any market. Another quarter of qualitative assurance.
8. The U.S. relisting, on the call: Four consecutive earnings calls without a mention, while the chief executive told a broadcaster in late September that the company is "actively pushing the process." The intention is public; management chose not to repeat it to its own shareholders on its own call.
9. The September board appointment: A sovereign wealth fund's head of China private equity joined the board on September 29, the day before quarter end, taking the board to ten directors with three independent, and the earnings materials did not reference it.
10. Why GMV grew 19.7% and revenue 17.0%: A 270bp spread between system transaction value and recognised revenue, with no channel GMV disclosure to locate it, at the same time as partnership revenue per store fell 16.2%.
11. The truncated Q&A itself: Two questions and a hard close for time, for the second consecutive quarter, at a company facing live debates about cannibalisation, an unexplained investment and an unexplained debt balance. Taking two questions is a choice.

Market Reaction

• Pre-print setup: The stock closed at US$31.20 on Monday, November 2, having fallen 2.2% the prior session. It entered the print down 6.9% year to date against a 10.2% gain for the S&P 500, down 21.2% over trailing twelve months and down 5.2% over the trailing thirty days, inside a 52-week closing range of US$27.52 to US$40.30. That is a washed-out setup: the shares had given back the entire Q1 buyback rally and sat 22.6% below the 52-week closing high going in.
• Initial reaction: Shares gapped up roughly 5.8% in pre-market trade to about US$33 after the release, with reporting noting an intraday advance of as much as 7%.
• Reaction session, Tuesday November 3: Opened at US$32.90, traded US$32.05 to US$33.49, and closed at US$32.90, up 5.4% or US$1.70. Volume of 3.71 million shares ran 4.0x the 30-day average of 0.9 million.
• Follow-through session, Wednesday November 4: Opened US$33.41, traded US$33.07 to US$34.90, and closed at US$34.45, a further 4.7% gain on 5.88 million shares, heavier than the print day itself.
• Cumulative: Up 10.4% over two sessions from the pre-print close. At US$34.45 the stock is 14.5% below the 52-week closing high, 25.2% above the low, and 2.8% higher year to date.

The market read this print the way we do, which is not always true of this name. The release leads with revenue growth, and revenue came in below the only consensus figure we hold. Had the stock been trading on the top line it would have fallen. Instead it rose on the margin and the authorisation, and the heavier volume came on the second day rather than the first, which is the signature of institutional accumulation following the headline rather than reacting to it. Day-one reactions to Luckin prints have recently mean-reverted: at Q2 the shares gave back the entire print-day gain within weeks. A second session of larger volume and a larger percentage move is a different pattern.

What the market is paying for is identifiable from the three facts it did not have before. First, a store-level margin of 23.3% that delivery reversion did not supply, which, if it holds outside the summer peak, raises the sustainable earnings base rather than the quarterly print. Second, US$609M of authorised repurchase against a US$10.8B market value, which is both a demand function under the stock and a statement about where the board thinks value sits. Third, record non-GAAP net income of RMB2.145B and operating cash flow of RMB4.1B, which make the authorisation credible rather than aspirational.

What the market is discounting is the comparable-store line. A business reporting (11.1)% comparable sales and rising 10.4% is a market that has accepted the base-effect explanation, and it has reason to: the prior-year comparative falls from +14.3% to +1.3% at the quarter turn, and management disclosed an improving monthly path. The risk in that acceptance is that it is now consensus. If the fourth quarter does not show the comparable line improving against an easy base, the explanation that currently costs the stock nothing will start costing it something.

Street Perspective

Debate: Is the margin recovery now structural, or did the quarter just flatter a mix effect?

Bull view: The bull case is that this quarter settled the argument on the bears' own terms. The sell-side objection for two quarters was that margin improvement was an accounting migration from the delivery line, and this quarter the delivery line moved 19 basis points while operating margin moved 204. Store-level margin of 23.3% is back to its pre-subsidy-war level with a worse delivery mix and a fleet 34.5% larger, which is the definition of scale economics rather than mix reversion.

Bear view: The bear camp contends the sequential comparison flatters the business because the second quarter was a transition quarter still absorbing subsidy distortions, and that the honest frame is year over year, where delivery gave back 890bps and every other cost line took 512bps of it away. On that view the underlying cost structure is still deteriorating, materials are up 207bps, overhead is up 82bps on share-based compensation growing 81%, and the only genuinely improving line is the one that is nearly finished improving.

Our take: The bulls have the better of this one, and the bears are right that the year-over-year comparison is the uncomfortable one. The decisive point is that the sequential improvement was broad rather than concentrated: six of seven cost ratios improved, led by store rent and other operating costs, which is the purest volume-leverage line and improved 81bps on a 7.2% larger fleet. That cannot be a mix artifact. The bear objection about the remaining delivery runway is correct and now largely irrelevant, because the business has demonstrated it does not need it. We set this as the explicit test at Q2 and it passed. The bears' stronger argument is one they did not make: Q3 is the seasonal peak, and the same non-delivery lines improved more from Q2 to Q3 in 2024 and 2025, so the fourth quarter is the cleaner test of whether the leverage is structural.

Debate: Does (11.1)% comparable sales mean the base effect, or a saturating network?

Bull view: The bull argument is arithmetic plus precedent. The quarter lapped the +14.3% subsidy peak, the decline narrowed every month from July to September, the improvement picked up further into the fourth quarter, and the comparative falls to +1.3% next quarter. The precedent is the company's own: comparable sales ran as deep as (13.1)% in the third quarter of 2024, after a prior opening acceleration, and were positive two quarters later. Meanwhile customers hit a record, cups per customer hit a high, and the gap between the per-store decline and the comparable-store decline narrowed rather than widened.

Bear view: The bear camp argues that management has now effectively conceded the network effect by describing comparable sales as a "wave-like" function of its own opening pace, which is an admission that adding 2,599 stores a quarter thins existing catchments. On this view the base effect and the cannibalisation effect are not alternatives but additive, three consecutive negative quarters with worsening magnitude is not a base artifact, and the company has refused for four quarters to publish the new-store productivity data that would separate them.

Our take: The bears finally have the better question and the bulls still have the better evidence. The question is better because management's own framing supplies it: if comparable sales respond to opening pace, then some of the decline is self-inflicted and will persist as long as the pace does. The evidence is better because a store-level margin at a cycle high alongside a narrowing gap between the per-store and comparable-store declines is not what a saturating network looks like: a network thinning its own catchments would show unit margin falling with unit volume, and unit margin went to an eight-quarter high instead. The honest position is the one we held at Q2: the disclosure to settle it does not exist, and the fourth quarter base reset will settle it for us. What changed this quarter is that the margin evidence got stronger while the demand evidence got weaker, and the margin evidence is the one tied to cash.

Debate: Does the enlarged buyback authorisation re-rate the pink-sheet discount?

Bull view: The bull case is behaviour over three stages. The board authorised US$300M in April, spent 96% of it in four months, upsized to US$500M rather than letting it lapse, and then added a fresh US$500M on the print. It bought at rising prices, from a blended US$31.98 through June to US$33.48 in September, which rules out the opportunistic-bottom-fishing reading. US$609M remaining equals 5.6% of the market value, funded by a business generating RMB4.1B of quarterly operating cash flow with RMB11.1B of net cash.

Bear view: The bear camp contends a discount built on a fraud restatement, a pink-sheet quotation and a controlling shareholder does not close on repurchases, and that the quarter supplied fresh reasons for it: borrowings rose 67% with no explanation, a RMB644M investment appeared with no description, a sovereign investor that co-invested with the controlling shareholder took a board seat with no mention in the earnings materials, and the board remains ten directors with three independent. On this view buying back stock while governance disclosure deteriorates is the wrong trade for shareholders to applaud.

Our take: We side with the bulls on the capital allocation and with the bears on the governance, and we think the governance side is the one that has moved. The repurchase programme is now large, executed and repeatedly re-authorised, which is about as much as a management team can do with this lever, and the blended US$32.58 is a usable reference for where the board sees value. But the unexplained items are accumulating rather than clearing: the debt balance is two quarters old and rose 67% in the second, the investment is new and undescribed, and the board change went unmentioned. A relisting would collapse this discount far faster than any buyback, and the chief executive's September comments make that more plausible than it was. Until then the buyback is a reason to own the stock and a reason to keep reading the filings.

Model Update Needed

Valuation impact: The model moves up, and this time it moves up on the line that matters. A quarter ago we raised the margin and cut the comparable-store line, and flagged that the margin increase depended on a delivery tailwind with roughly 300bps left. We now raise the margin again while removing that dependency, which is a better kind of upgrade: the sustainable store-level margin assumption goes to 22-23% from 20-21%, and the mechanism is volume leverage rather than a finite mix reversion, although part of this quarter's leverage is seasonal and the fourth quarter is the cleaner test. Full-year non-GAAP net income of RMB5.8B to RMB6.2B against RMB4.2B in 2025 is a materially higher earnings base than we carried at Q2, arriving with a share count falling faster than we modelled. The offsets are real but smaller: revenue growth comes down roughly four points, the tax rate goes up 200 to 300 basis points, and we now carry share-based compensation as an explicit and growing driver. At the November 4 close of US$34.45 the shares trade at roughly 13.8x trailing twelve-month non-GAAP earnings, 16.6x trailing GAAP, and about 1.2x trailing sales, or near 12x on the midpoint of our full-year range. Net cash of RMB11.1B is 15% of the market value and the remaining authorisation is another 5.6%.

Thesis Scorecard Post-Earnings

Overall: Thesis strengthened. Of the eight items we set out to watch at Q2, two were met cleanly, one was partially met, one was answered unfavourably, one was an observation that was confirmed, and three were not met. That is a scorecard which looks poor until the two clean passes are identified. They are the margin-source test and the buyback re-authorisation, which are the two the Outperform rating was built on. A quarter of margin expansion that does not depend on the delivery line was the explicit condition we published for calling the recovery broad-based, and it arrived, albeit in the seasonal peak quarter. Against that, the single most important demand confirm failed on the printed number, average unit volume deteriorated further year over year, and the governance file got thicker rather than thinner. The three outright misses are mostly disclosure items rather than operating failures, which is why the thesis strengthens on weak arithmetic. The net is a higher and better-founded earnings base against a slower and still-unresolved demand picture, which is a trade we take at 13.8x trailing non-GAAP earnings.

Action: Hold the position, add on weakness. The stock has already taken 10.4% of the re-rating over two sessions, so this is not an add-at-the-market print. The buyback's blended US$32.58 is the reference we would use: below it, an outside buyer is paying less than the board has been paying with shareholder money, and the margin evidence supporting that purchase is now stronger than it was when the board set the price. We would reassess on three triggers: a fourth quarter in which the comparable-store line fails to improve against a +1.3% comparative, any disclosure that the RMB644.3M investment is larger in scope than its carrying value suggests, or a further increase in borrowings without explanation.