Coupang, Inc. (CPNG) Earnings Call Transcript & Summary

Hello, everyone. My name is Emily, and I will be your conference operator today. At this time, I would like to welcome everyone to the Coupang 2026 Second Quarter Earnings Conference Call. [Operator Instructions]. Now I'd like to turn the call over to Mike Parker, Vice President of Investor Relations. You may begin your conference.

Thanks, operator. Welcome, everyone, to Coupang's Second Quarter 2026 Earnings Conference Call. I'm pleased to be joined on the call today by our Founder and CEO, Bom Kim; and our CFO, Gaurav Anand. The following discussion, including responses to your questions, reflects management's views as of today's date only. We do not undertake any obligation to update or revise this information except as required by law. Certain statements made on today's call may include forward-looking statements including statements regarding future financial and operational results. Actual results may differ materially. Additional information about factors that could potentially impact our financial results is included in today's press release, and in our filings with the SEC, including our most recent annual report on Form 10-K and subsequent filings. As we share our second quarter 2026 results on today's call, the comparisons we make to prior periods will be on a year-over-year basis, unless otherwise noted. We may also present both GAAP and non-GAAP financial measures. Additional disclosures regarding these non-GAAP measures, including reconciliations of these measures to the most comparable GAAP measures are included in our earnings release, our slides accompanying this webcast and our SEC filings, which are posted on the company's Investor Relations website. And now I'll turn the call over to Bom.

Thanks, everyone, for joining us today. Consolidated revenue grew 10% year-over-year in constant currency, that's a step up from the growth in Q1 and in line with the guidance we provided. Adjusted EBITDA margin also came in within the range of guidance. First, Product Commerce saw its revenue growth increase to 8% year-over-year in constant currency. Let me spend a moment on the customer behavior behind that number because the reported rate blends groups moving in different directions. The vast majority of our customer spend never moved. That group is spending at the highest levels in our history and compounding similarly to before last year's data incident. The spend that did leave was a minority of the total and most of it has already returned. Some of these customers were away for months, long enough to settle in somewhere else. These customers not only came back, they returned to their full prior spend levels and have since gone beyond it, compounding their spend at similar high rates as before. And because the customers who returned were, on average, higher spenders than those who haven't, the picture in spend terms is even clearer than in customer counts. The vast majority of the spend, the incident disrupted is back and growing the way it did before. On top of that, new customers keep arriving. Total WOW membership, for example, now exceeds its levels before the incident. New members begin at the early stage of the spend curve where new members always begin. So record membership shows up in revenue on a lag. Every cohort before them has climbed that curve, and we expect these to do the same. The spend of all customers, excluding just those that left during the incident and haven't returned is growing around 16% year-over-year, which is closer to the spend growth product commerce delivered in Q2 last year before the incident. Spend growth does outpace revenue growth, but the gap between the 16% and the reported 8% revenue growth is driven mostly by the missing spend of the cohort that hasn't returned. We'll keep chipping away at earning them back. And after we lap the affected periods, we expect the spend growth for all of product commerce to reflect the growth rate of this underlying customer base because the cohort that has not returned will no longer be in our year-over-year comparison. On margins, we're continuing to work through the disruption. We plan capacity and fixed costs against the predicted demand curve, and much of that capacity has long lead times. With revenue temporarily below that plan, those costs represent a larger share of revenue today. We could cut them significantly, but we've chosen not to because the right long-term decision is to grow into the capacity and support our customer experience that has always been our North Star. There are also meaningful volume-based savings in our supply chain that we're missing this year that we expect to recover next year. We've also deliberately increased marketing spend to reacquire customers but we plan to reduce it next year after we've lapped the period. We run this business with precision on capacity utilization and volume economics. That's what allows us to deliver double-digit growth along with expanding margins in normal times. And it's also why a sudden shock is more visible in our numbers than it might be somewhere else. And the same discipline that makes the disruption more visible is what will enable us to reverse it. We've seen elements of this before, coming out of COVID, a sudden shift in the demand curve pressured margins and we return to normal levels. The shape of the recovery this year won't move in a straight line. Holiday timing and seasonal cost patterns will affect the reported improvement quarter-to-quarter. Gaurav will walk through how that shapes the back half. The arc we're managing to runs through next year. After we lap the affected hperiod next year, we expect to work our way back to the growth and margin structure that product commerce ran at before. I want to also note a few trends that we track closely and the broader opportunity we see before us. Our cohorts have continued to increase their spend with us year after year, including through this past year. Our oldest cohorts, the customers we acquired about 1.5 decades ago are still increasing their spend today. They now spend nearly 10x what they spent in their first year and climbing. Our newest cohorts are growing fastest of all at the beginning of the same curve. Each year adds a new cohort at the start of the curve, while every earlier cohort keeps climbing it. All 3 trends, our oldest cohorts continuing to grow spend, the spend of our subsequent cohorts climbing the curve to convert to higher levels and new cohorts joining power our growth. The past year tested us and all 3 trends held. And the spend of our cohorts climbs because spend growth is wallet share growth. As we expand selection, customers find more and more of the things they buy for which we've broken the trade-off between price, selection and speed. We're saving customers more money and more time with every item we add on Rocket Delivery. Our wallet share or penetration of the overall retail spend in Korea remains below the levels of penetration we see from global peers in markets like the U.S. And we don't view these levels as our ceiling because penetration follows the strength of the offering. Every trade-off we break brings purchases online that were never in reach before. As the offering improves, the addressable share of retail expands with it. And the same logic applies to margins. The long-term margin drivers keep compounding. Automation continues to improve productivity across our fulfillment and logistics network, and margin-accretive offerings like advertising and FLC are still early in their scale. And AI raises the ceiling on both. We think of AI as a multiplier and what it multiplies is a set of assets we've been building for 15 years, the physical network, operating data from billions of orders picked, packed and delivered and direct relationships with tens of millions of customers. Applied to the customer experience, AI improves discovery, personalization and service. Applied to operations, it compounds productivity and lowers the cost to serve. And applied to margin-accretive offerings, it raises the returns for the merchants and brands who are using them, which expands the addressable opportunity itself. Turning now to developing offerings where we're running the playbook I just described in new markets and categories. In Taiwan, we've built out and continue to expand our own end-to-end fulfillment and logistics network that now delivers the vast majority of our shipments, next day, 7 days a week, the only service in Taiwan to our knowledge that does. We also began rolling out our Dawn Delivery experience, which became a defining part of the customer experience in Korea to our first neighborhood in Taiwan. And we're building it faster than the first time because Taiwan didn't start from scratch. Taiwan inherits over a decade of technology and process innovation from our Korean operations, the design, the systems, the operational playbook refined shipment by shipment. It took us 4 years into our logistics journey in Korea before we were able to launch Dawn Delivery, Taiwan reached it in just 1 year. With the network in place, the work shifts to the input we know best, selection. Our selection in Taiwan today is a fraction of Rocket Delivery's in Korea. Every item we add is another purchase where our customer saves both money and time. And each of those purchases earns more of their wallet. That's the same dynamic that as our Korean cohorts still climbing 15 years in, and we can already see it taking hold. Our early cohorts in Taiwan are retaining and growing their spend tracing the curve Korea's cohorts produced at the same age. Taiwan is on the same compounding curve as Korea, just earlier on it. Two things to keep in mind as this scales, First, the path won't be linear quarter-to-quarter. Sometimes building selection at the right cost structure means stepping back in a category to rebuild it. Second, today's economics reflects the stage of our build-out, not the destination. As we work out the inefficiencies of an early supply chain and our volumes grow into the network, volume economics engage, and we expect Taiwan's P&L to follow the path that Korea paved. Let me turn to our on-demand delivery service, which includes both Eats in Korea and Rocket Now in Japan. We've shared in the past our approach to developing offerings. We make disciplined initial investments where we see the potential for meaningful long-term cash flows. We look for early proof points in customer behavior, and we scale investments only as results to validate the opportunity. This cycle is complete when an offering stops drawing on the portfolio and starts funding it. Eats has traveled that entire arc. We entered a category most considered settled with a modest investment and a simple thesis that the same propositions that customers valued in commerce, price, selection and service would matter just as much in food delivery. Customers responded at every stage, and we invested behind that response. Today, Eats has grown to serve millions of customers and the category itself has grown with us. Food delivery in Korea has more than quadrupled in size since we launched Eats, now reaching a meaningfully higher share of total restaurant spend than when we entered. And we've been a significant driver of that expansion. And the capital story has come full circle. Rocket Now, our on-demand delivery offering in Japan is in its early investment stage. Eats and Rocket Now are today sustainable on a combined basis. That is the model working end-to-end, disciplined entry, validation, scale and then an offering that carries its successors. We're also extending what Eats built. We've begun rolling out nonfood on-demand delivery, the same network and speed customers already trust apply to new use cases, offering customers even more opportunities to save time and money. What I've covered today is 1 model running at 3 different stages. Product commerce is furthest ahead with years of investment in infrastructure and selection, customer cohorts still compounding 15 years in and the margin expansion that follows scale. Eats has now run that same cycle in a second category. And Taiwan is midway through it. Building the network, filling in the selection, moving through the same stages that Korea moved through. They represent the same playbook at 3 different points on the same curve. I'm proud of our teams for continuing to build for our customers at every stage. Our ambition from the very beginning has been to build an experience that wows customers so much that they ask themselves, how did I ever live without Coupang? Every item we add, every offering we build and every market we enter is another chance to build to that standard. With that, I'll turn the call over to Gaurav.

Thanks, Bom. Our Q2 results demonstrate a continuation of the momentum we started to see last quarter. Before I walk through the numbers, I should highlight 2 items that shape the reported numbers this quarter. First, the Korean won weakened significantly versus the U.S. dollar during the quarter, reaching its weakest level in more than 15 years. As a result, our reported growth rates in U.S. dollars understate the underlying growth of our business and we believe it is especially important to evaluate our results on a constant currency basis this quarter. Second, our product commerce results this quarter include $410 million in administrative fines recently imposed by Korean regulatory authorities. While these fines are still subject to judicial review, and we plan to appeal them through the courts, we recorded the expense this quarter within OG&A in the P&L. Where relevant, I'll explain our results both with and without the fines. Let me now walk through the segment results and then cover our consolidated performance. Product Commerce segment net revenues were $7.4 billion for the quarter, growing 1% on a reported basis and 8% in constant currency. This represents a sequential improvement from the 5% constant currency growth rate we reported last quarter. Product Commerce active customers for the quarter were 24.7 million, growing 3% year-over-year and up from 23.9 million last quarter. As we noted, the sequential decline last quarter reflected the lag effect of the data incident on our trailing 3-month active customer definition. This quarter, we saw the dynamic reverse driven by the number of returning customers and new customer additions. And as Bom noted, WOW memberships now exceeds the level we saw prior to the data incident and returning members are spending today at higher levels than they did before the incident and increasing their spend at similar high rates as prior to the incident. Product Commerce gross profit for the quarter was $2.3 billion with a gross profit margin of 30.5%. This represents a contraction of approximately 210 basis points year-over-year but an improvement of 25 basis points quarter-over-quarter. The year-over-year contraction is driven primarily by supply chain headwinds and temporarily elevated levels of promotional activities to accelerate customer reacquisition. Segment adjusted EBITDA for Product Commerce was $382 million for the quarter, which excludes the fines I mentioned earlier, resulting in an adjusted EBITDA margin of 5.1%. This represents a contraction of approximately 390 basis points year-over-year, resulting from the gross profit margins impact I just described as well as the headwinds from our current capacity and fixed cost structure built against a pre-incident demand curve. We believe the margin pressure we are experiencing to be relatively short term in nature and not representative of a structural change. Our conviction of the drivers of our long-term margin expansion potential, operational efficiencies, supply chain optimization, continued investment in automation and technology and the scaling of our margin-accretive categories and offerings remains firmly in place. Within developing offerings, we reported segment net revenue of $1.4 billion growing 20% on a reported basis and 24% in constant currency. Growth continues to be led by Taiwan, Eats and Farfetch driven by increasing levels of customer adoption of these emerging initiatives. Developing offerings generated $226 million in gross profit for the quarter with a gross profit margin of 15.8%, expanding both year-over-year and quarter-over-quarter as these offerings continue to demonstrate the path to sustainable economics. Segment adjusted EBITDA losses were $219 million, an improvement of $110 million versus last quarter and over 440 basis points of margin improvement over last year. At the consolidated level, we reported total net revenues of $8.9 billion for the quarter, growing 4% on a reported basis and 10% in constant currency. This is in line with the constant currency growth range we guided to last quarter. Consolidated gross profit was $2.5 billion with a gross profit margin of 28.2%. This represents a contraction of 188 basis points year-over-year, but an expansion of 115 basis points quarter-over-quarter. As with Product Commerce, the year-over-year margin compression continues to reflect the near-term headwinds we have discussed. OG&A expense was $3.1 billion for the quarter or 34.4% of total net revenues, representing an increase of approximately 610 basis points year-over-year, excluding the $410 million in fines, OG&A expense was approximately $2.6 billion or 29.8% of total net revenues, an increase of approximately 150 basis points year-over-year and down slightly quarter-over-quarter. This underlying year-over-year increase primarily reflects temporarily elevated marketing spend, our continued investments in developing offerings and a cost-based position against an expected demand curve from prior to the data incident. We view the elevated marketing and promotional spend as a deliberate near-term investment to accelerate growth, not a structural change in our cost base. Operating loss for the quarter was $556 million. Excluding the fines, the adjusted operating loss was approximately $146 million, representing a quarter-over-quarter improvement and an operating loss margin of approximately 120 basis points. Loss before income taxes was $533 million or $123 million, excluding the fines. We incurred income tax expense of $37 million this quarter. Our tax dynamics continue to reflect the fact that losses in early stage operations in Taiwan and Japan do not generate an offsetting tax benefits at the consolidated level. In addition, the fines recorded this quarter are not deductible for tax purposes in Korea, which further impacted our reported effective tax rate this quarter. Net loss attributable to Coupang's stockholders was $570 million or approximately $160 million, excluding the fines. This resulted in a diluted loss per share of $0.32 or approximately $0.09, excluding the fines. We generated $163 million in consolidated adjusted EBITDA this quarter, which excludes the $410 million in fines recorded this quarter resulting in an adjusted EBITDA margin of 1.8%. This represents a contraction of approximately 320 basis points year-over-year at the low end of the guidance range we provided. The vast majority of this year-over-year contraction is attributable to 3 temporary items, supply chain dislocation, elevated levels of marketing investments, our primary fixed cost base size to a pre-incident demand curve each of which we expect to work through over the next few quarters. On cash flow, on a trailing 12-month basis, we generated operating cash flow of $1.4 billion and free cash flow of $105 million. The decreases versus prior periods reflect the lower profitability resulting from the data incident, increased level of investment in developing offerings and elevated levels of capital expenditures. During the quarter, we repurchased an additional 23 million shares of our Class A common stock for approximately $459 million. We continue to be opportunistic in our capital allocation as we identify opportunities to generate long-term returns for our shareholders. Now a few comments on our outlook. We expect Q3 consolidated constant currency revenue growth to be 8% to 9%. With the year-over-year comparison pressured by the timing of the Chuseok holiday season in Korea, which negatively impacts Q3 of this year compared to Q3 of the prior year. As Bom indicated, excluding the portion of customers that left during the data incident period and not yet returned, we are seeing growth in customer spend at 16% year-over-year. After we fully lap the affected periods in Q2 next year, the cohort that has not returned will no longer be in our year-over-year comparison. And we expect the spend growth rates of Product Commerce to reflect the spend growth rates of the underlying customer base. On margins, we expect underlying improvements in Product Commerce to continue their progress in Q3. However, we expect these to be offset by higher weather-related seasonality and the timing of Chuseok holiday that is different from last year. As a result, despite our expectation of making meaningful improvements in the underlying drivers we noted this quarter, we anticipate consolidated year-over-year adjusted EBITDA margin contraction in Q3 of 300 to 400 basis points, similar to the contraction we guided to for Q2. Looking beyond Q3, we expect the improvements to become increasingly more evident with product commerce adjusted EBITDA margins by mid-2027, returning to margin levels approximating those we generated prior to the data incident. We plan to provide more detailed guidance on the 2027 margin expectations towards the end of this year. We continue to estimate full year developing offerings adjusted EBITDA losses of between $950 million and $1 billion. The largest contributor of the developing offering losses this year remains the long-term investments we are making into building our retail offering in Taiwan. As always, our investment in developing offerings is anchored by our commitment to rigorous analysis, operational excellence and disciplined capital allocation. Operator, we are now ready to begin the Q&A.

[Operator Instructions]. The first question is from Stanley Yang from JPMorgan. Your line is now open.

Thank you for your opportunity to ask the questions. I have 2 questions. First, on the margin guidance, I couldn't hear very clearly about this point. But did you guide your EBITDA margin of the product commerce will recover to 2025 level in 2027. Can you confirm that? And if that is the case, what will be the major margin expansion drivers and related to this margin outlook, have you seen a more competitive environment post-data accidents that caused cost pressure? That was my first question. And my second question is about this Taiwan. So do you have any visibility into a potential structural decline in developing offering loss in Taiwan over time? If so, what will be the major drivers?

Yes. Thanks for your question, Stanley. Yes, we did guide that we'll recover our margins by middle of 2027. Specifically in Q3, we guided that the underlying drivers of our margin are improving. Some of the drivers haven't fully recovered yet and won't require until next year because of the scale, but we are on trend to benefit from high utilization as the volume grows. Second, in the projected numbers for Q2, this progress is for Q3. This progress is masked by holidays, timing and seasonal cost patterns which is why we guided 300 to 400 basis points of year-over-year contraction similar to Q2 -- beyond Q3, the improvement will become increasingly evident in the reported numbers as we expect this Product Commerce margin to return to approximately pre-incident levels by mid-'27, and we'll provide more detailed guidance towards the end of the year. The reason for our confidence is that nothing has structurally changed. The pressure is from some temporary items that we chose to carry rather than cut and trade away the customer experience. We have also seen this play out before. Coming out of COVID, a sudden shift in demand pressured margins the same way and the same discipline brought them back.

Yes. So just to quickly touch on that a little bit. Yes, Stanley, we have guided margins to recover fully in 2027. The drivers of that margin expansion are the opposite of what led to the compression. The compression -- the large majority of the compression traces back to volume coming in softer than our demand trajectory. As Gaurav mentioned, we run a business with very tight execution. And so when demand deviates meaningfully below that trajectory, we get capacity stranded -- being stranded and we miss out on some of the volume-based savings that affect our economics. All of that is reversible and it's not structural, but the bulk of it is mechanical, and we expect to recover it in 2027. We have seen more elevated competitive activities. But the -- we think the past year has given us an unusually direct test, customers who paused during the incident had a lot of time to try alternatives, months trial alternatives. And when they came back, and most of them did, they came back, not splitting their prior spend with us. They're back, as Gaurav mentioned, spending at the highest levels ever and growing that spend now as fast as they did before. We also have membership now, an all-time high. And as customers who never left are also spending at record levels growing as fast as they did before. So we feel very confident about the position that our service -- the value proposition that we're providing for our customers. And we'll keep focusing on widening the experience gap with broader selection, lower prices and faster service. On developing offerings, the economics of our -- I think we've typically talked about developing offerings as a whole. Taiwan is a big part of it. As I mentioned earlier, Eats has really come full circle and completed the cycle. We're talking about Taiwan, which is in the middle of that cycle. It's on the same curve that Korea is on just earlier. And the economics there reflects the stage of the build out that it's at. And its priority right now is building that foundation for durable growth and economics at scale. And we're encouraged by the signals we look for and customer behavior and response that we invest behind. And our guidance for developing offerings we typically provide at the end of the year. So we look forward to sharing that with you later this year.

Our next question will be from Eric Cha from Goldman Sachs. Your line is now open.

I have 2 questions also related to the guidance you provided today. So first, the first question is on -- actually your commentary around the product commerce revenue growth. I think you mentioned that without the WOW members that have left the spending growth was about 16%, which compares to the 8% growth on a reported basis as a whole. Just wanted to understand the gap between these 2 numbers. I think you mentioned that the people that have left are minority. So it does seem like this minority seems to have a bit of an outsized impact. So I just wanted to just get a sense why the gap seems a bit large. And also, when we head into the fourth quarter, obviously, we'll be lapping a quarter where it's post the incident. So given that you mentioned that the existing WOW members are growing at 16% level, can we expect a meaningful step-up in growth by the time we reach fourth quarter this year? So that's my first question. The second question relates to the margin. And I think as Stanley mentioned, you commented that the margin will be down 300 to 400 bps year-over-year. Does that include your sort of expectation around a material impact or some impact from the fire incident as I believe it does cause a bit of efficiencies around logistics. So I was just wondering how much of that was baked into the guidance? And also, you mentioned the margin will be fully recovered by midpoint next year. Just wanted to understand the cadence of that improvement, I'm not sure how you will answer the first question, but if -- by fourth quarter, we do have a bit of an improvement in the year-over-year for the top line, can we expect a bit of a step-up in the fourth quarter, seeing the evidence of that sort of reversal kicking in? So largely sort of a linear fashion, I know you mentioned linear is not the way you see. But moving from 5% margin to probably around 8% margin for PC, Product Commerce margin, should we expect a step-up in fourth quarter and then continuously move up to the 8% level? Or do we expect rather compressed level and then step up in the midpoint next year to around 8% level? Just wanted to get color around that.

Eric, thanks for your question. For clarification, the data incident was closer to the end of -- which began in the end of Q4. And the affected period extended into part of Q1, actually most of Q1. So I think that's one point of clarification. The cleanest way to understand the picture is, as I mentioned, to set aside that small group that's still missing. The customers who left during that period and haven't yet come back, everyone else is growing at about 16% year-over-year. As I mentioned, that's close to the right Product Commerce was compounding before the incident. And there are 3 groups that make up that 16%, all of whom are healthy. The largest is customers who never left, they're spending more with us than at any point in our history, compounding at their old pace. Second group is a group that left came back. As I mentioned, they came back are spending now at record levels. And they represent an even larger share of spend that left because the returnees skewed towards the higher spenders. And now spending at record levels growing faster than they did -- as fast as they did before the incident. And the third is new customers who are joining us actually even faster than they did before the incident. So the gap, as you point out, is between the 16% and the reported 8% is primarily driven by that small group of lower spending customers who haven't returned yet. There is, of course, this factor of spend growth outpacing revenue growth as well. But really, the vast -- the majority of that gap is really driven by that small group that is in the base of the year-over-year comp and is really distorting or obscuring the growth rate of the underlying customer base, the customers who are with us. As I mentioned, next year, the mechanics work as such -- as we outpace -- as we lap the period. The group that still haven't returned to us -- won't have returned to us by that point, will drop out of our year-over-year comp base. And so this product commerce spend growth naturally converges to the spend growth of the customers who are with us, our customer base at that point, which will be then compared against a comp base that no longer includes that missing cohort. So hopefully, that mechanics -- the mechanics of that are clear. And as to your point -- to your question of how that reflects in our growth rate, as you might imagine, we had varying levels of disruption that began at the end of Q4 that extended into Q1 after we fully lap that effective period or as we lap that period, you might see some of the absent customers drop out of the base, but you'll see the full effect of that after we fully lapped the effective period.

Yes, Eric, let me take your question on the margins. We highlighted that the margin contraction were primarily driven by the supply chain headwinds, elevated levels of marketing activities to reaccelerate or accelerate the customer acquisition and the headwinds from our current capacity and fixed cost structure built against the pre-incident demand curve. So we expect to make meaningful progress in mitigating these impacts over the next several quarters, including in Q3. But Q3 has this specific timing issue of holidays and seasonal cost patterns. The whole margin step down and recovery is driven by our volumes that came in softer than our planned trajectory since the incident. The mechanics going forward are clear. As the demand rebuilds capacity and fixed cost utilization come back into balance and the volume-based supply chain savings that we are missing this year come back and the incremental marketing, we have deliberately deployed this year begins to normalize. So as you also pointed out, it may not be linear or at least we're not forecasting or giving any guidance on that.but we are confident that we'll -- we should be able to make the whole recovery.

The next question is from Seyon Park from Morgan Stanley.

I'll ask a question on developing offering. I think there was many questions asked on Product Commerce margins already, so I'll skip that. But just on -- first of all, on the developing offerings, you did see the losses narrow meaningfully in the second quarter. Can you maybe provide us a little bit of context as to what part of the business saw lowered losses? And how we should think about the losses for the third quarter and the fourth quarter? That's my first question. Second question, I guess, is a little bit going beyond the core. A lot of talk about agentic AI and how AI could change how we purchase items online? And I know Coupang is already using a lot of AI especially for demand forecasting, delivery and the like. Does the company have plans to maybe utilize AI like from an agentic side of things, is that something that's kind of in one of the research pieces that we can kind of expect going forward? That's my second question.

Thanks for your question. On developing offering losses, as I mentioned, Taiwan is in the middle of that cycle. It is where we're primarily investing. I think I wouldn't read too much into investment levels quarter-to-quarter. We provide guidance on an annual basis. And as I mentioned, we currently remain in line with the full year guidance we're providing for developing offerings. And we're still at a very exciting phase of building that foundation for both durable growth and economics. In Korea, we saw the power of building that underlying infrastructure that really powers the quality of customer experience that leads to both compounding growth and attractive economics in the long term. We saw the benefits of that of building the network, the last mile capability, the full supply chain, these things take time to build, but they compound for a very long time afterwards, as you've seen. And the economics of Taiwan and the growth of Taiwan actually reflects the stage that we're at in the build-out. And again, I wouldn't -- the buildout is still under progress, and we don't manage to a quarter -- a quarter. We provided annual guidance. But we see that the build-out is happening in some ways, even faster in Taiwan than in Korea. As I mentioned, it took Korea 4 years of logistics build-out to reach Dawn Delivery. Taiwan has reached it in just 1. We still have a lot of work to reach the quality of experience we're striving for. But Taiwan does benefit from over a decade of systems and operating processes that it's inheriting. And we invest -- we scale our investment as customers validate it, and we're especially encouraged by the cohort response that we're tracking. The customer behavior we see in Taiwan at this stage looks very much like Product Commerce did in its early years. And that's the signal we invest behind. So we look forward to updating our guidance with you at the end of the year around developing offerings. On AI, as you point out, we are already have already deployed AI in many parts of our business. It's already contributing across operations, fulfillment, logistics, supply chain, pricing, advertising, customer service. It's providing meaningful improvements in service levels and cost to serve where it's been deployed. We have customer-facing -- active investments on the customer-facing side as well, for example, in certain discovery where the industry's direction, we think is clear. And AI has the potential, as we've mentioned earlier, to really multiply the asset base that we have underneath it, the physical network, our operating assets or customer experience, and we've already seen and will expect it to improve customer engagement and compound productivity and lower cost to serve and all the benefits that we've mentioned earlier, on margin-accretive offerings, we think it actually has the potential to expand the addressable market itself, the opportunity itself because we've seen signs of it being able to -- and we see the potential for it to raise the return of the merchant brand customers who use them. On the agentic AI part that you've brought up specifically, we think this is still a work in progress. We think the industry's direction or the -- it's not clear that the winning experience has emerged but we're investing. We are investing in teams and the research, as you mentioned, to explore it while being thoughtful about it and investing with the same discipline that we do and all the other initiatives that we have on the exploration front. Whatever form agentic shopping takes, we believe we'll be in the best position to provide the winning experience which we believe will combine AI with all the other aspects of customer experience to provide a complete and seamless buying experience that customers trust. And to build a complete and seamless buying experience, you need more than AI. AI is one input, but there are many other assets that will be part of it. And we believe we should be in a position to provide the best of all worlds.on that front. So we'll continue to invest in exploring this opportunity thoughtfully and when a winning approach emerges we'll be positioned to execute with the same operational excellence and capital discipline we apply everywhere else.