Luckin Won the Subsidy War, Lost the Comp, and Bought Back $500 Million of Itself

Luckin Won the Subsidy War, Lost the Comp, and Bought Back $500 Million of Itself

Luckin's third quarter is one of the cleanest examples you will see of a company deliberately trading a headline metric for a unit-economics metric. Revenue rose 17% to RMB 17.9 billion, GMV rose 20% to RMB 20.7 billion, GAAP operating profit rose 55% to RMB 2.8 billion at a 15.4% margin, and non-GAAP net profit hit a quarterly record above RMB 2.1 billion. Self-operated same-store sales fell 11.1%. Both facts are true, and management knows exactly which one it wants you to look at.

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.

The Trade: 2,600 Stores for 11 Points of Comp

The mechanism is not complicated, and management was unusually candid about it. Last year's subsidy war created a high comparison base for July-through-September, because delivery orders carry a higher average selling price per cup than pickup orders. When platform money evaporated this year, delivery volume went with it, and pickup volume recovering did not fully plug the hole. Same-store sales turned negative.

But the same channel collapse is precisely what repaired the P&L. Delivery expenses fell from 19% of revenue to 10%, down 38% year-over-year in absolute terms to RMB 1.8 billion. That single line item is worth roughly nine points of revenue, which is more than the entire 3.8-point gain in operating margin plus a cushion. The delivery mix normalization was the margin story this quarter — not operating leverage, not pricing power. Store-level operating margin at self-operated stores reached 23.3%, up 5.8 points, and store level profitability at the self operated store base is now broadly back to pre-subsidy levels despite a network that has nearly doubled.

That framing hides two less flattering lines. Cost of materials rose to 38% of revenue from 36%, and store rental and other operating costs rose to 22% from 20%, both attributed to the mix shift away from high-ASP delivery and to the loss of last year's volume-driven scale economies. More pointedly: sales and marketing rose 30% to RMB 976 million and G&A rose 36% to RMB 1.1 billion — both roughly double the 17% revenue growth rate. The CFO attributed the marketing step-up to “higher traffic-related investment following the subsidy war” — Jing An, Chief Financial Officer · 2026-11-03. So the profit record is a mix-shift artifact, while the operating expense base is being rebuilt to defend traffic in a post-subsidy world. That is a margin that needs monitoring, not celebrating.

An Argument, Delivered Twice

The Q&A ran two questions and was then closed for time. Both were about same-store sales. The CEO's translated answer resorted to history — a “wave-like pattern” — Nancy Song, Head of Investor Relations · 2026-11-03 in which rapid openings pressure the comp, a moderation in openings revives it, and 2024's second half proved the sequence. That is a legitimate model; it is also a promise that store additions, which management says will continue at a pace of store expansion ahead of the industry (net additions of 2,543 in China alone, 38,909 stores globally, up 33% year-over-year), will keep suppressing the very metric analysts use to value restaurants.

Worth noting: none of this was a surprise. On the August call management had already flagged that “as platform subsidies were particularly intense during July and August of last year, we expect the resulting high base effect to persist” — Jinyi Guo, CEO · 2026-08-03. And back in November 2025 they had described an entire doctrine: delivery as a supplemental channel, pickup as the long-term primary format, with a warning that this would not be quick.

We believe the coffee business will naturally return to a pickup-oriented model over time. Although this transition period will take a longer time to happen.

It happened faster than advertised, and it happened because the platforms pulled their money, not because Luckin willed it. The company gets credit for having the store density to absorb the shift; it should not get credit for having predicted the timing. Note too that in that same call they told investors “our same-store sales growth metric in the fourth quarter and even next year will face some short-term fluctuations and pressure” — Jinyi Guo, CEO · 2025-11-17 — so the negative comp is a delivered forecast, not a miss.

What Fell Out of the Script

The most informative part of this quarter is what management de-emphasized. A year ago, Singapore's 100th store was a headline item; now store in Singapore is a single clause inside a single paragraph. The whole overseas footprint is 279 stores — 103 self-operated in Singapore, 22 in the US, 154 partnership in Malaysia — after 56 net additions in the quarter. Compare the February call, which reported 9 US stores at year-end and 80-plus in Singapore. Thirteen US stores in ten months is patience bordering on abstention. The international expansion narrative that dominated early 2026 has quietly become a rounding error against 38,630 China stores.

Also gone: the relisting question. In November 2025 management said they remained committed to the US capital market with “no specific time line or schedule” — Jinyi Guo, CEO · 2025-11-17 for a mainboard listing. This quarter they answered that question with capital return instead — a new one-year, USD 500 million repurchase authorization, to take effect once the existing USD 500 million program is exhausted, of which USD 391 million had already been spent on roughly 96 million Class A shares (12 million ADS). Against a market cap near USD 10.5 billion, that is a real authorization. Cash stands at RMB 14.3 billion versus RMB 9 billion at year-end, with RMB 11.1 billion net cash and RMB 4.1 billion of quarterly operating cash flow.

And conspicuously absent: coffee beans. In November 2025 management flagged elevated green coffee prices as a next-year margin headwind. This quarter, with cost of materials up 24% in absolute terms, the coffee bean cost line was not discussed at all. When a named headwind stops being named while the underlying expense line rises, that is usually not because it resolved.

The Tape Is Short a Different Story

Luckin's own price tape is unavailable in this dataset, but the global keyword baskets that carry the name are not flattering. lower tier cities — a basket that includes Luckin alongside NIO, MNSO, BZ and 3690.HK — is a 360-day decliner. The tape has been voting on the China down-tier consumer as a short while Luckin's actual store economics quietly repaired themselves. Chinese consumption reporters on the same day were framing around traffic and volume indices rather than headline comps. That divergence between the basket and the unit economics is the setup here — either the tape is early or the margin is late.

Customer metrics argue the former is more likely. Average monthly transacting customers exceeded 120 million, a record; 28 million new transacting customers were added, taking the cumulative base past 520 million; cup volume and cups per customer both set highs. Thai Milk Tea sold 13 million cups in its first week. Those are not the numbers of a demand problem. They are the numbers of a company converting subsidy-driven demand into habit — and charging the transition through the same-store sales line.

The negative comp is narrowing month by month, and management says the improvement “has picked up further” — Nancy Song, Head of Investor Relations · 2026-11-03 into the fourth quarter. If that holds and marketing intensity normalizes, the mix-shift windfall becomes durable margin. If marketing keeps compounding at 30%-plus while the comp recovers slowly, the record quarter will look like a peak in a channel transition rather than a new baseline.