Hello, I'm Eleanor from Kakao IR. Let's begin the Q2 2023 earnings conference call. Today, I'm joined by Simon Hong, the CEO, and Jay Bae, General Representative of Corporate Investment of Kakao. Please be reminded that the earnings results are consolidated estimates under the KIFRS basis, and are subject to change upon the auditor's review. Also, forward-looking estimates are based on assumptions, so actual results may differ from figures included in today's presentation. I will now invite Simon to present on business highlights for Kakao, the parent and major business results of the platform business.
Good morning, and good afternoon. This is Simon. During the first half of this year, Kakao primarily focused on laying down a robust foundation that will enable long-term growth for KakaoTalk, implementing multiple significant projects. The transformation of KakaoTalk, our priority since last year, is based on the premise that increased connectivity leads to heightened user engagement. As KakaoTalk users create accounts and build relationships through their contact information, the app reflects their real-life social interaction. However, one of the limitations we identified was that without an increase in real-world connections, a surge in user activity was challenging. In Q2, we introduced the Open Chat tab as our third tab, which aimed to elevate activity within KakaoTalk. This new feature allows users to expand their network to strangers based on shared interests, thus offering a broader scope of interaction beyond their existing contacts.
Currently, all users view the Open Chat tab identically. In the second half of this year, we plan to introduce a personalized curation feature that recommends chat rooms based on user interests. Since its launch, the Open Chat tab has quickly established itself with over 10 million daily active users, achieving its goal of securing double the daily active users than the existing View tab ahead of schedule. The Friends tab will evolve to enhance social interactions and to serve as a space where users can discover diverse relationships beyond their real-life circle. In Q3, we plan to reinforce social interaction features by integrating the Timeout feature into profiles. This feature allows users to share daily content with selected individuals, which disappears after 24 hours. Simultaneously, we are planning to launch the local service, offering benefits and useful information around users' geographical proximity.
The Friends tab previously recommended talk channels beyond real-life friends, but with limited discovery potential. Therefore, the local service space will aim to offer small business owners more exposure opportunities despite significant changes this year. The Friends tab has grown its daily active users by 36% from 22 million at the end of the last year to 30 million by the end of Q2, and we aim to reach 40 million by year-end through various enhancements. KakaoTalk now has three tabs, each with 10 million-plus DAU, and we've witnessed a meaningful increase in user time spent. We intend to continue leveraging this momentum in the latter half of the year to facilitate structural changes. Looking at the bigger picture, our goal is for KakaoTalk to evolve into a comprehensive communication service, with each of its five tabs attracting more than 10 million daily active users.
Moving on to Q2 highlights. In Q2, our consolidated revenue surpassed the KRW 2 trillion mark for the first time, reaching KRW 2.4 trillion on the back of the consolidation of SM Entertainment. This signifies a 17% increase from Q1 and a 12% increase from the same period last year. To clarify, excluding the effect of SM Entertainment's inclusion this quarter, we see a 4% increase from Q1 and a slight decrease of 1% from the same period last year, totaling KRW 1.8 trillion. Starting with TalkBiz, Q2 revenue recorded KRW 530 billion, which is a 2% QoQ decrease, but an 11% YoY increase.
First, on Talk Biz advertising, the Bizboard turned to a plus YoY growth by achieving a 2% increase, thanks to the expansion of exposure in the Open Chat tab and increased advertiser demand for CPT products in the Friends tab. As Talk Channel, which showcased excellent reach and conversion rates, solidified its position as the representative product of KakaoTalk, the messaging ad has recorded its highest ever quarterly revenue, with 50% on-year growth. Combined, Talk Biz advertising revenue recorded a 4% on-year growth and a 12% QoQ growth. In the second half of the year, we plan to continue expanding the Talk Biz advertising business by targeting small to medium-sized advertisers, creating a foundation for stable revenue generation and reducing sensitivity to external economic conditions.
For talk channels, the total number of active channels increased to 2.04 million in Q2, and we expect the number of partners to expand substantially through external partnerships in the second half of the year. Furthermore, we anticipate a significant expansion in the number of talk channels with more than 1,000 friends in the second half of the year, thanks to the enhanced exposure and usability through the expansion of the talk channel display space in KakaoTalk Commerce Tab and the Kakao Reservation Service. We also anticipate an expansion into local advertising business through the talk channel. With the launch of local services in the Friends tab in the second half of the year, users will be able to check the news of any kinds of local businesses around them.
On top of that, if local communities grow well in the Open Chat tab as well, we anticipate a further increase in demand from local partners. We will make efforts to instill the perception that business communication is done through the Talk Channel for our partners. Moving on to TalkBiz Commerce, Kakao Commerce's Q2 total GMV recorded KRW 2.3 trillion, growing 7% year-on-year. Starting with TalkGift, a culture of sharing feelings through this platform has expanded even during Parents' Day and Teachers' Day, when offline gifting was traditionally common. The number of buyers in May reached the highest of the year, and the GMV grew 7% year-on-year. In particular, premium food and luxury beauty delivered gifts were very popular among users, resulting in a 10% year-on-year GMV increase for all delivered gifts.
In June, we introduced a new online luxury specialty store, LuxTab, based on the differentiation of direct sourcing from global luxury brands. LuxTab plans to strengthen the gift lineup that can be shared among closest relationships, like family and lovers, by providing limited edition products that can only be purchased from LuxTab. As new relationship networks are formed within KakaoTalk, TalkGift is also expected to grow into new areas. While Gift has been mainly used for gifting among real-life acquaintances so far, for the second half of the year, we plan to expand our target to non-acquaintances, sharing the same interests and diversify the purchase methods for future growth. Kakao Commerce has continued its growth with TalkGift at the center. We believe that various commerce services other than TalkGift are now starting to grow.
Going forward, by maximizing the synergy between TalkStore, Talk Channel Store, and Live Commerce within the Shopping tab, we plan to provide users with a more personalized shopping experience. For sellers, we plan to expand Kakao Commerce exclusive advertising product, which can deliver brand information to 48 million users on KakaoTalk in real time, and to solidify our position as a differentiated relationship-based commerce platform. Next on Kakao's future growth engine, the new initiatives. Let me begin with Kakao's AI business strategy, led by Kakao Brain. Kakao Brain unveiled the multimodal foundation model, Karlo 2.0, in July, and is preparing for the final stages to reveal a next-gen large-scale language model in the second half of the year.
The benchmark test, conducted during the development of the foundation model, has confirmed that the technological gap with global models is rapidly narrowing, and we are currently conducting tests internally and in the process of upgrading the model. We believe that for AI to provide additional value to users when combined with real services and to be a sustainable service, cost efficiency must be thoroughly considered. Therefore, the next-gen large-scale language model is directed toward being an optimal-sized foundation model that achieves a proper balance between accuracy and cost efficiency. In addition, we plan to unveil a lightweight language model in the second half of the year that can be operated more economically and timely applied to different vertical services.
In the second half of the year, by building foundation models of various parameter sizes, we will continue to attempt to combine various vertical services within Kakao Group with AI to deliver new user experiences. Kakao Healthcare is pushing ahead to launch full-fledged services in both the B2B and B2C sectors in the second half of the year. Firstly, in the B2B sector, we have launched a clinical data warehouse equipped with further enhanced AI and statistical functions to standardize and accumulate massive clinical research data within hospitals. The first clinical data warehouse was successfully built at Bundang Seoul National University Hospital in June, and we are currently deploying the system at several different major hospitals in Korea. We expect various sources of demand, including global research institutes and big pharma companies.
In the digital front door business, which is the digital touch point for all medical care services, we plan to launch a service that seamlessly connects the beginning to end of medical care, from hospital reservations to recession and prognosis management for large hospitals using KakaoTalk chat. Lastly, in the B2C area, we are preparing a blood sugar management service that is more convenient and predictable, connected with a continuous glucose monitor for patients with diabetes or those who need blood sugar management. In June, we signed the final contract with Dexcom, the top company in the U.S. CGM market, for global business cooperation. Together with Dexcom, we plan to start launching in the domestic market in Q4, and to actively promote a hyper-personalized digital blood sugar management service in various global regions, including the United States, the Middle East, and Japan.
Other than Kakao Healthcare, Kakao is actively supporting the journey to overcome diseases through Rare Notes, a service by HumanScape that we invested two years ago. Rare Notes provides information on rare and, rare and intractable diseases to patients. Furthermore, through Mommy Talk, the largest pregnancy and parenting community in Korea, Kakao is making efforts to contribute to overcoming the low birth rate challenge faced by our society. Continuing on to cloud, Kakao Enterprise is restructuring its business structure with a focus on the cloud business, which has high future growth potential and investment value. Through close collaboration with AMD, Kakao Enterprise has accumulated technical competitiveness on par with the level of global CSPs, preparing to target the high-performance premium cloud market.
The company's cloud product was the most selected by participating companies in the high-performance computing resource support project run by the government in the first half of the year, thereby proving its competitiveness. Kakao Enterprise is working to gradually increase its customer base by targeting the public cloud transition projects and the K-Cloud project run by the Ministry of Science and ICT, while accumulating references in various business areas through collaboration within Kakao Group. As it has established a foundation for a new leap forward this year, it expects the financial effects of the business structure reorganization to become visible from next year and to show the substantial growth of the cloud business. Next, I would like to talk about Kakao's ESG effort in Q2.
As part of the Active Green initiative announced last year, Kakao has publicly disclosed the status of Kakao Carbon Index, an environmental contribution indicator developed in-house on our website, and is conducting campaigns to promote users' environmental awareness and active participation. Kakao will continue to explore various ways to form a consensus with users for the transition to a low-carbon society and create changes. In addition, through the Kudos for Credo project run by Kakao to support small merchants nationwide and foster coexistence with the local community, we have provided benefits of about KRW 14.2 billion to over 47,000 small business owners through Talk Channel message cost support until Q2, striving to strengthen digital communication of small businesses.
Together with this, through the My Favorite Neighborhood Market project that supports the digital transformation of our traditional markets, 587 stores and 16 traditional markets have opened KakaoTalk channels. We aim to expand support up to 100 markets this year. In August, we plan to further disclose video content containing the essence of the Kakao ESG Report 2022 and the ESG Voice Report produced to enhance the accessibility of the visually impaired as part of the barrier-free initiative. We would appreciate your continued interest and encouragement in Kakao's sustainable management endeavors.
Next, Jay Bae, General Representative of Corporate Investment of Kakao, will brief you on key earnings and business highlights.
Hello, this is Jay. I will begin with the performance highlights for platform and others. Q2, platform and others revenue is up by 8% QoQ and 6% YoY to KRW 396 billion. Mobility revenue was up 11% on quarter and 24% on year, driven by the consolidation of KM Park last July and steady growth of all business segments. In the taxi business, despite a gradual recovery in user demand following the fare increase, we have continued to experience sustained growth due to the increasing preference of mobility platform among taxi drivers. Parking, which is a relatively new business, saw steep growth from increase in users as more parking lots interlocked with Kakao T application. As of Q2, it has become second-largest revenue source following taxi hailing business, contributing to total mobility revenue growth.
In the global market, we've been promoting overseas services in line with the increase in outbound travelers, driving large increase in overseas performance. Kakao Mobility will solidify its position not as simple roaming service, but as an unrivaled global mobility platform underpinned by global network of mobility apps owned by Splyt. KakaoPay Q2 TPV was up 18% YoY, reaching KRW 34 trillion. Revenue TPV, which is a revenue contributor, was up 19% YoY to KRW 10 trillion. By expanding the payment network in favored overseas destiny, destinations for domestic users and on enhanced usability of inbound foreign visitors, supported by alliance with Alipay Plus Cross-border payment TPV was 84% YoY, driving the entire payment service growth. For KakaoPay securities, as strategic investment into Siebert Fi nancial drove cost efficiency, we were able to run different promotions, including lowering fees for stock trades
This resulted in an increase in trading volume, as well as rise in KakaoPay money, money balance, improving user activity metrics, with total trading volume up by 2.7 times YTD. KakaoPay Insurance has been recognized for its achievements in overseas travelers insurance, leaping to the second position. As KakaoPay Insurance will become a wholly owned subsidiary in August, its business efficiency will improve and is expected to grow, driven by product line of expansion to meet the needs of daily life. Next is on content, and I'll begin with the music business. On SM Entertainment consolidation, Q2 music revenue was up by 107% QoQ, and 130% YoY, reaching KRW 481 billion. On a pro forma basis, the revenue was up 4% QoQ, and 15% YoY to KRW 241 billion.
Melon focused on defending its five million subscribers and profitability by scaling up content offerings and CRM competency. By introducing Spotlight, a service featuring a new album inside Melon, variety of content is being delivered to the fan base, while Music Wave is growing its into a community service, gathering people of similar taste to get together and communicate within the same music channel. We are also at the forefront of expanding K-pop's global influence by providing streaming and downloaded data to Billboard. In distribution and label business, IVE's first full-length album, which sold 1.6 million copies, and I-dle's mini album, drove growth in revenue and operating profits. We are also seeing synergies start to crystallize between two companies as Kakao Entertainment acquired distribution rights for SM artists starting June 1st.
As SM Entertainment completed the setup of an efficient business framework following the restructuring, we anticipate positive results to come through with the debut of a new boy group, Rise, in September. Concrete talks are ongoing to build a global management system and integration of entities in the United States for the management business. In the second half, aespa and Ive will make their debut in North America. As we solidify the basis for business cooperation, combining the platform and know-how of Kakao Entertainment and SM Entertainment, we will endeavor to bring business synergies in the near future. Next is story and media business. Story revenue was flat QoQ and YoY at KRW 231 billion.
In time for Piccoma's seventh year anniversary in Japan, we ran multiple events to improve user retention and monetization during the Golden Week, and as a result, drove record high GMVN revenue. In terms of content, K-Webtoon IP, God of Black Belt and Saving My Sweetheart, continued its popularity streak following last quarter, and we saw a sharp rise in revenue from completed Japanese series after joining our platform through Piccoma's traffic and platform operational competency. Going forward, Piccoma will further gain competitiveness through its own studios and the network of publishing houses, and will scale up content curation technology to offer better content offerings based on user preference. In the process, it will grow into a platform that users visit more often to find what they like out of greater collection.
The business restructuring in North America that has been ongoing since the end of last year, is now complete. Since then, we focused on expanding the library of content, growing the number of titles by twofold, and under efficient marketing spend, GMV was up QoQ. In the second half, we'll focus on revamping the tapas application, as well as increasing the quality and quantity of IPs. For tapas, UI/UX will get a new facelift, uplifting the operational efficiency through global service standardization. In terms of the content library, we already have web novel IPs, not only Korea, but China and US, from Radish and Wuxia world. We will leverage the strength to curate and expand the content offerings on tapas, enhancing our IP competitiveness in the web novel business. The media business.
Media revenue was up 8% QOQ, driven by the production and promising new IPs, including You Have Done Well, slated for release on Netflix. With key titles mostly scheduled for releases in the second half, there was 38% decline YOY, with revenue reporting KRW 74 billion. Drama Namnam, which is based on Kakao Webtoons, is currently on air and receiving high praise, which goes to show our OSMU strategy, supported by original Webtoon IP is becoming stronger. In the second half, there will be a release of 10 titles, including the drama Castaway Diva, The Worst of Evil, and the non-scripted content, Zombie verse. While 20 or more dramas and films and entertainment programs will go through the planning and production phase.
With recent success of K content, not only in drama but also in entertainment programs, Kakao Entertainment aims to strengthen its studio capabilities across various genres. By doing so, we will expand our global achievements and continue to thrive in the entertainment industry. In Q2, game revenue was KRW 269 billion, up 9% QoQ, but down 20% YoY. In Q2, growth in mobile game revenue and profits, driven by ArcheAge War and Mamluks first anniversary event, resulted in QoQ increase. In the second half, we plan to maintain growth through major updates of Odin and ArcheAge War, while anticipating the success of new MMORPG title, Ares. With a combination of established flagship titles and new releases, along with strong global performance, we expect to continue our quarter-by-quarter growth in the latter half of the year.
Next, I will walk through consolidated operating expense and profits. In Q2, consolidated operating expense was up 16% QoQ and 17% YoY to KRW 1,929 billion, with booking of KRW 230 billion of additional operating expense, including PPA amortization following SM consolidation. On a pro forma basis, due to cost control, the increase was limited to 2% QoQ and 3% YoY. Going into detailed expense breakdown, labor cost was up 6% QoQ and 11% YoY, reporting KRW 472 billion due to SM consolidation, despite conservative hiring spends across the entire company. Cost of revenue was up 22% on quarter and 24% on year to KRW 810 billion.
Compared to last quarter, despite declining one Pmix for Kakao Gift and lower marketplace fees and outsourcing costs for the media business on year-on-year basis, the consolidation of SM Entertainment drove up cost of revenue. Cost of revenue as a percentage of the total revenue recorded 39.7%, up 1.4 percentage point QoQ. Outsourcing infrastructure expense was up 70% QoQ and 20% YoY to KRW 282 billion, due to increase in fees linked to the growth of Kakao Mobility taxi businesses, as well as the higher AI infrastructure fees on top of SM consolidation. Driven by the marketing activities in Piccoma and Mobility, marketing expense was up 5% QoQ. On year-on-year basis, Kakao Entertainment's cost-efficient marketing lowered the spending 20% YoY to a total of KRW 120 billion.
The marketing expense as a percentage of the total revenue was 5.9%. With consolidation of SM, there was KRW 23 billion of PPA amortization. On a higher AI-related CapEx, depreciation was up 23% QoQ and 36% YoY to KRW 186 billion. All in all, Q2 operating profit was KRW 114 billion, with OP margin at 5.6%. Operating profit, excluding the impact of SM consolidation, was KRW 101 billion. For your information, Kakao separate basis revenue, comprising advertising and commerce business, reported KRW 635 billion in revenue and KRW 117 billion of operating profit, with OP margin of 18%. Next is non-operating revenue and expense. Other revenue was KRW 17 billion, comprising of disposition gains from equity method shares and other miscellaneous items.
Other expense was KRW 28 billion, which includes disposition losses from equity method shares and advisory fees. Equity method gain was KRW 28 billion, decreasing KRW 6 billion QoQ on lower net profit from Kakao Bank, and equity method loss was down KRW 11 billion QoQ to KRW 29 billion, on decline in losses from the equity method companies. Financial revenue was KRW 62 billion, mostly from interest and dividend income, which was down KRW 18 billion QoQ due to lower dividends. Financial expense was KRW 51 billion, which are mostly interest expenses, and due to the rise in FX translation loss, following higher one dollar FX rate, it increased by KRW 4 billion QoQ. Q2 corporate income tax was KRW 56 billion, accounting for additional tax amount determined by the tax authority following tax audit of certain subsidiaries. Consolidated net profit reported KRW 56 billion.
Lastly, for CapEx, on back of AI and data center investments, there was KRW 196 billion of tangible asset investment, including machinery and construction in progress, and KRW 42 billion for content and intangible assets, amounting to a total CapEx of KRW 238 billion.
This ends the earnings presentation for Q2 2023. We will now begin the Q&A. As we are mindful of the time constraint, I would like to ask that you limit your questions to two per person.
[Foreign language]
Now Q&A session will begin. Please press star one, that is star and one if you have any questions. Questions will be taken according to the order you have pressed the number star one. For cancellation, please press star two, that is star and two on your phone.
[Foreign language]
The first question will be provided by Jingu Kim from Kiwoom Securities.
Please go ahead with your question.
[Foreign language]
I would like to submit two questions. First has to do with your plans regarding internalization of generative AI. I would like to understand as to what your specific execution strategies are for each of the verticals, and when can we actually see AI-powered platform and services? Second question relates to your infrastructure investment costs for AI and data center. Would like to understand as to the size, the extent of the size, for this year, and when would you foresee that it will start to normalize? At that point in time, what will be the amount of infrastructure investment?
I would like to submit two questions. First has to do with your plans regarding internalization of generative AI. I would like to understand as to what your specific execution strategies are for each of the verticals, and when can we actually see AI-powered platform and services? Second question relates to your infrastructure investment costs for AI and data center. Would like to understand as to the size, the extent of the size, for this year, and when would you foresee that it will start to normalize? At that point in time, what will be the amount of infrastructure investment?
I would like to submit two questions. First has to do with your plans regarding internalization of generative AI. I would like to understand as to what your specific execution strategies are for each of the verticals, when can we actually see AI-powered platform and services? Second question relates to your infrastructure investment costs for AI and data center. Would like to understand as to the size, the extent of the size, for this year, when would you foresee that it will start to normalize? At that point in time, what will be the amount of infrastructure investment?
I think right now, the way I see this, this situation is it's not about who comes out with the hyper AI, generative AI model first.
I think it's more about finding that optimal and reasonably priced AI model that could be used for many different verticals. When I say most optimal and appropriate AI model, that means from the application side, you would have to go through many verifications and validity tests and fine-tune those applications. We require innovation on the application side. You asked about a specific vertical. I think you would have to look at the B2B, B2C, domains as well as SaaS and PaaS. Right now, what we are doing is we're testing different sizes of parameters, starting from 6 billion, 13 billion, 25 billion, up to 65 billion parameters. In so doing, we wish to develop a cost-efficient AI model.
[Foreign language]
I believe, one of the most important areas for Kakao, it would be powering KakaoTalk with AI technology. Basically, KakaoTalk is currently growing into a comprehensive communication platform, underpinned by 3 major pillars of personal communication, social communication, and business communication. I believe that, we will be able to see, firstly, synergies come out of business communication pillar. Up to now, business partners, and you would have experienced this through, Talk Channel, the communication mode was from 1 to many. Going forward, we believe that we will be able to provide more personalized messages by using AI technology across many number of users. We will be able to build out a business communication pillar that is 1 to 1 in nature.
For example, in the areas of ordering, making appointments, counseling and making payments, these transactional services, if we could power that well with AI, that would be beneficial. For instance, in terms of delivery, travel, and accommodation, people could. When people are making appointments or reservations, we will be able to bring together the meta information that Kakao has and combine that with these non-standardized requests that are being made by the users, and we will be able to fully leverage that data and opportunity, be able to provide competitive services and opportunities to our user base. In terms of going forward, our plans, Kakao Brain, I think it will be some time in the second half of the year.
I'm not sure at this time, but it will be probably after month of October, we will be able to release our foundational model that strike a good balance between performance and cost efficiency.
Thank you.
좀 피크에 달하고, 내년에는 요 성장률이 크게 감소할 것으로 이제 보여지고요. 그때부터 이제 저희가 이제 요 인프라 비용 관련해서 안정화가 이루어지는 시점이라고 판단하고 있습니다.
This is Jay, responding to a question about, about our infrastructure spending. Just to provide you with a background as to what it is comprised of. It comprises of depreciation cost on our GPU servers, as well as depreciation and amortization on the leased assets, including the, the shelf space, and all the infrastructure related fees that we pay. In the case of infrastructure expense last year, basically, well, basically it had posted a increase of about 50% in a year-on-year basis. This year was the peak, and we expect starting next year, there will be a significant drop in the amount of infrastructure spending. It will be at that point in time, we will see stabilization of infrastructure spending.
네, 다음 질문 받겠습니다.
Next question, please. The following question will be presented by Stanley Yang from JP Morgan. Please go ahead with your question.
네, 감사합니다. 저 두 가지 질문이 있는데, 첫 번째 질문은 그 친구, 친구 탭이 그 프로필 변화 이후에 이제 트래픽이 크게 상승 중인데요. 향후 수익화 전략이 궁금합니다. 그리고 전망치에 대해서도 말씀해 주시면 감사하겠습니다. 그에 대해서 이, 이 두 번째 탭하고 첫 번째 탭의 어떤, 어, 매출이 굉장히 많이 차이가 나는데, 첫 번째 탭이 이제 얼마나 빠른 속도로 이제 매출이 올라갈 수 있는지 궁금합니다. uh, 두 번째 질문은 그 아까 대표님께서 그 AI 관련 그 비용 합리적인 어프로치에 대해서 말씀해 주셨는데, 그러면 이제 올해 그 AI 관련된, uh, 추가 비용, uh, 투자, 이 천삼백억, 그 예전에 말씀해 주셨던 부분이 조금 더 내려갈 수 있는 건지, 아니면은 아까 말씀해 주신 대로 뭐 내년도에 이제 비용 효율화가 더 이루어지는 건지에 대해서 한 번 더, uh, 확인해 주시면 감사하겠습니다.
Yes, I have two questions that I would like to ask. If you look at your Friends tab, after the change in the profile page, we've seen a significant traffic increase. Would like to get some color on what your monetization strategy is and what your outlook is with respect to this matter. Also because there's a big revenue generation difference between the first tab and the second tab, I just would like to get some understanding as how quickly we will be able to see that revenue increase from the first tab. Second question. Simon has mentioned that you will take a cost efficiency approach when it comes to AI investment.
Would that mean that you will be able to further lower that KRW 130 billion of investment that you've mentioned, which will be an additional investment with, with regards to the AI, AI? Could you be able to further reduce that, or would the timing be next year?
스톤 양님, 질문 감사합니다. 첫 번째 질문에 대해서 제가 답을 하고요. 두 번째 질문에 대해서는 Jay Bae, General Representative of Corporate Investment가 답변드리도록 하겠습니다. 현재 Friends tab이 30 million 이상이 이제 찾아오는 곳이 됐는데요. 노출되는 광고 개수는 현저히 작고요. 저희가 이제 Chat tab에서 느꼈던 경우는, Chat tab의 이제 Bizboard라는 것은 굉장히 퍼포먼스가 좋게 광고 설계가 돼 있습니다만, 지면에 노출되는, 지면에 의존하는 광고 성장 방식은 좀 한계가 있다. 오히려 이제 이게 서비스로 진화하면서 그 지면에 노출되는 광고 개수가 아니라, 그 Friends tab을 계기로 다양한 광고 모델들이 나올 수 있는 방향으로 설계되어야 된다고 인식을 하고 있고요. 일단 지면 노출도 합니다.
지면 노출도 이제 그 CPT products도 수익성이 좋아지고 있고 하지만 기본적으로 현재로 그게 보시면 다 알겠지만, alphabetical order에 anonymous list가 들어가 있는데, 그 anonymous list는 그 사용을 이제 많이 안 하시는 걸로 나와서 그걸 이제 accordion menu라고 해서 그 선택하면 그 anonymous list가 나올 수 있게 하고. 그 지면에 이제 그 저희가 48 million users들을 그 이제 동의를 많이 해주셔가지고 이분들의 어떤 지역 정보를 알 수가 있거든요. 그, 그분들의 지역을 이제 세분화해서 그 세분화에 맞는, 세분화된 지역에 맞는 그 콘텐츠도 추천을 하고, 이를테면 날씨 같은 것도 있을 수 있고, 교통 정보도 있을 수 있고요. 편의점이라든지, 마트라든지, 동네 주민들이 일상적으로 찾는 그 단골 가게, 또 시간이 지나면 가치가 소멸하는 그 소멸성 재화를 취급하는 신선 식품이나 정육점, 이런 것들과 이제 관계를 강화를 하고, 그분들하고 이제 커뮤니케이션을 하는 수단이 Talk Channel이 될 걸로 생각을 해서, Talk Channel Messaging Ads revenue도 이제 저희도 기대하...
뭐, 일단 그게 이용자와 그 자영업자, 소상공인들한테 유용한 소통 채널이 되는 게 목표고요. 소통이 되면 거기에 이제 부가적으로 광고수입도 파생돼...
Thank you, Stanley, I would respond to your first question, and then Jay will tackle the second question. If you look at our Friends tab, it's visited by more than 30 million people on a daily basis, but if you look at the number of advertisements that has been shown through it, the number is extremely low. The lesson that we were able to learn from the chatting tab is that the billboard itself was designed such so that we could maximize performance. There's a limitation in terms of growing our advertisement revenue, which is just dependent on that surface and that specific slot. As we are evolving our the services that we offer, we are at this point coming up with various different types of designs that could really help for us to use different advertisement models.
Of course, we will continue to improve on the CPT and we will continue to expose different advertisements. If you currently look at the tab, it's a list of names in alphabetical order. What we can do is we can change that into an accordion type of a menu, menu configuration. And thanks to the consent that we receive from our 48 million user base, we know their location information, their, their geographical information. The data then can be used for us to recommend the most appropriate content to our user base, including, for instance, weather information, traffic, and the most nearby convenience store and other grocery stores. So all the neighborhood stores and shops, that information related to them can be provided to the users of KakaoTalk.
Especially for, for stores that sell perishable goods like fresh food and other poultry, we can. We will be able to then further facilitate the use of the KakaoTalk Channel as well. Our first stage would be to create that communication channel between the users of KakaoTalk and the vendors and the sellers in that neighborhood, based on which we will be able to then create advertisement effect.
We then will be, you, once we reach that level, the traffic from the Friends tab will be able to be further uplifted, and we will be able to use it for advertisement purposes by leveraging the local or the neighborhood information, and based upon the relationship that people form with the partners that reside within that neighborhood, which will then later on bring us, positive advertisement impact as well as revenue stream from messaging. Before I let Jay respond to the second question, let me just elaborate as to what I mean by reasonable level of cost. When you look at, building out of an AI, there are two aspects. One has to do with, laying out the infrastructure for training the AI model.
Second aspect is when users actually make use of this AI-powered services, when they're retrieving and when they're making certain prompts and commands into this AI model, there is an infrastructure expense that's going to be incurred. Typically, when users are making use of these AI models, if it's a free services, then that retrieval unit cost has to go below KRW 1 in order for it to be effective. If it's a paid-for fee service, then if even if it's about KRW 10 or KRW 20 on a unit basis, then as long as we can, then we'll be able to recover the cost that is necessary for the use of that infrastructure.
Based on our past experience, when we have the profiles or the editing of the pictures and images powered by AI, we were able to recover that retrieval cost, but that was only for on a temporary basis. Once again, we are developing our AI model under the objective of bringing that unit cost below one for free services.
네, 두 번째 질문에 대해서 답변을 드리겠습니다. 지난 Q1 저희 실적 발표 당시에 AI 관련 투자 증가로 인해서, Kakao Enterprise, Brain Healthcare를 포함한 이 유 이니셔티브의 손실 규모가 연간 KRW 300 billion대까지 확대될 수 있음을 말씀드린 바가 있습니다. 이번 Q2 Kakao Brain의 영업손실은 파운데이션 모델 구축에 필요한 인프라 수수료 지출이 증가하면서 전분기 대비 확대되었습니다. second half of the year에도 AI 관련 연구개발 인력 증가와 현재 집중적으로 개발 중인 차세대 언어모델 구축에 따른 인프라 수수료의 증가 역시 확대될 것으로 예상하고 있어서 first half of this year 대비 second half of the year에 Kakao Brain의 손실 규모는 확대될 것으로 예상을 하고 있습니다. 다만, Kakao Enterprise에서 제공하는 AI 학습 트론 관련 클라우드 인프라 활용을 확대하면서 집행되는 투자 비용을 내재화해서 투자효율을 높일 예정이고, 또한 해당 회사의 경영 효율화를 통해서 기존에 말씀드린 KRW 300 billion의 손실 규모보다는 줄어들 것으로 예상을 하고 있습니다.
현재 AI 산업의 급격한 변화 속도를 감안하면, 투자 규모에 대해서 향후 투자 규모에 대해서 언급하기는 좀 조심스러운 부분이 있으나, 앞으로의 투자 규모 수준이 Kakao의 체력 내에서, 체력에서 감내 가능한 수준으로 이제 집행되도록 적정 수준으로 조율해 나갈 예정입니다.
This is Jay, responding to your second question. During the first earnings conference call, we mentioned that due to the AI related investments, our new initiatives, including Enterprise and Kakao Brain, the size of the loss could actually increase to about KRW 300 billion on an annual basis. In the second quarter, Kakao Brain's operating loss due to the build out of the foundational model and the following infrastructure fee spending on a QoQ basis, we've seen that loss increase. In the second half of the year, we are also planning on additional hiring of R&D talent for AI. Also we expect the infrastructure spending to actually go up as we build out a next generation language model.
Compared to the first half of the year, we expect in second half, Kakao Brain's size of loss will go up. Now, having said that, Kakao, the AI training and inference related computations, which uses the cloud infrastructure of Kakao, basically, we, what we want, what we are seeking to do is to internalize all that investment within the Kakao group of affiliates so that we can further enhance investment efficiency. And also through some rationalization process, the company, we will believe that we will be able to reduce the size of the loss below KRW 300 billion that we've reported in the past.
However, in light of the very fast changing transformation in the AI industry, I have to be a little cautious as to making future predictions, but we will do our utmost to make sure that all the investment requirements that the company has to deal with is within the level that is tolerable, in light of Kakao's capabilities.
[Foreign language]
Next question, please.
[Foreign language]
The following question will be presented by Rose Ju from CLSA. Please go ahead with your question.
예, 질문 기회 감사드립니다. 두 가지 질문인데요. 그 올해 Talk Biz 연간 매출 가이던스를 16%로 이전에 이제 주신 걸로 알고 있는데, 지금 상반기 보면은 지금 한 10% 초반대여서 아직까지도 이 목표가 달성 가능하실 것으로 보시는지 아니면 조정이 필요할 것으로 보시는지 좀 궁금하고요. 두 번째는 그 AI 관련해서 지난 분기 때 아마 올해 하반기에 그 코챗GPT를 그 출시하신다고 한 기억이 있는데, 그게 지금 여전히 그 유효하신지, 아니면은 지금 좀 약간 비용 효율을 감안을 하셔서 이 부분도 조금 연기가 될 수 있는지 궁금합니다. 감사합니다.
I have two questions. Previously you guided that your Talk Biz guidance would be growth of 11%, but I see that in the first half it was in the early 10%. Would you be able to achieve the target that you previously communicated? I also remember that when it comes to AI, that you will be releasing Kakao's version of Code GPT in the second half of the year. Considering the fact that you want to be rational in spending for AI, should we expect a delay in this time schedule?
로승주님, 질문 감사합니다. 그, 두 가지 다 제가 답변드려도 될 것 같아서 제가 하겠습니다. 그 TalkBiz, 뭐, 상반기까지 TalkBiz 누계 성장률은 십일 퍼센트예요. 저희가 이제 제시한 연간 성장 목표가 십육 퍼센트니까, uh, 좀 굉장히 부담스러운 숫자로 지금 보이긴 하는데 현재 Talk, Talk 채널이 견조한 성장을 이어가고 있고, 비즈 모두도 전년 대비 성장률을 매 분기 개선해 나오고 있어요. 상반기보다 하반기에 TalkBiz 성장률이 더 높게 나타날 것으로 전망하고 있어서, 그, 하반기 추이를 좀 지켜볼 필요가 있습니다. 그래서 이번 분기에는 연초 제시한 가이던스를 유지할 예정입니다.
... 두 번째 말씀하신 그 Co-GPT, 뭐, 그 하반기 연기 여부, 연기는 하진 않고요. 뭐, 아주 뭐 불가피한 불가항력적인 상황이 있으면 모르겠지만, 지금 그 열심히 하고 있고요. 십월 이후에는 나올 것 같긴 한데, 근데 저희가 하고 있는 게 그 모델 하나가 아니라요. 아까 뭐 모두에 말씀드렸지만, 유, 그 육십 빌리언, 뭐 백삼십 빌리언, 이십, 이백오십 빌리언 이런 다양한 모델 사이즈를 지금 테스트를 하고 있습니다. 그래서 뭐 이를테면 그 카카오톡의 메시지 요약 같은 거 해주는 거, 이렇게 뒤늦게 들어갔는데, 이제 많은 메시지가 쌓여있는 거, 그거 요약해 주는 거는 육십 빌리언 경량형 모델이 오히려 제일 그 비용 합리적으로 할 수 있다라는 계산이 나왔습니다. 그래서 이게 뭐 어떤 큰 모델 하나에 모든 걸 걸고 있는 상황은 아니고, 근데 그런 모델 육십오, 육백오십 빌리언짜리도 그... 저희가 준비를 하고 있는데, 연, 연내에 나올 것으로 생각하고 있습니다.
This is Simon. I think I will be able to respond to both of your questions, Mr. No. Yes. On the first half, on a cumulative basis, the growth rate was 11%, and we guided that our guidance will be 16% year-over-year growth. we believe, believe that because Talk Channel is very, you know, robustly growing, although the guidance looks a bit high, you know, we think that we'll be able to achieve that. Hence, we're going to keep to the previous guidance, because Bizboard is continuously showing a very high year-over-year growth rate, and so we expect that as we enter into the second half, the growth rate from the Talk Biz is going to be higher compared to the first half of the year.
I believe that we should wait and see, but once again, for this quarter, we will continue on with the guidance that we previously shared. Now, for Co-GPT, do not believe that we will be delaying that unless there is any force majeure. We're working very hard on developing these different models, and I think that we'll be able to see them, you know, after October. Once again, if I may elaborate, it is not just one AI model that we are developing. We're developing various different sizes of parameters using 6 billion, 13 billion and 25 billion, so different sizes. One of the services that we are currently envisioning is a message summarizing feature for all the different messages that is within the KakaoTalk.
For this type of a service, we think a more lighter version of an AI model would be more appropriate, using about 6 billion parameters. We think that that cost efficient model will work.
한 가지 정정을 해야 될 것 같은데. 아까 통역을 잘 하신 것 같은데. 아까 65 bil, B라고 한 거, 65 billion 개의 파라미터 수를 650 bil이라고 표현한 것 같아요. 65 B가 맞습니다.
네, 다음 질문 받겠습니다.
Next question, please.
다음으로 질문해 주실 분은 삼성증권의 오동환 님입니다.
The following question will be presented by Donghwan Oh from Samsung Securities. Please go ahead with your question.
아, 네, 질문 기회 감사드립니다. 두 가지 질문이 있는데요. 첫 번째는 저희 친구 탭의 그 변화에 대해서 약간씩 약간 방향이 달라지는 것 같은데요. 예전에는 타임라인형 SNS 도입 얘기하셨다가 이제 타임아웃 콘텐츠 도입도 얘기하셨는데, 최근에는 전략 방향이 로컬 비즈니스 쪽으로 이제 콘텐츠를 강화하는 쪽으로 바뀐 건지 그 부분이 좀 궁금하고요. 두 번째는, 어, 저희 뭐 오픈 채팅방에 최근에 그 트래픽 지표나 아니면 향후에 이제 매출 기회가 얼마나 될 수 있는지, 이런 것들이 좀 궁금한데. 어, 저희가 이제 오픈 채팅 같은 경우는 개인적으로는 이렇게 많이 사용하지는 않는 서비스라서, 이런 부분의 활성화나 포텐셜에 대해서는 조금 체감하는 부분에서는 약간 회심이 있는 상황이긴 합니다. 그래서, 어, 이게 콘텐츠, 그 텍스트 비즈 기반의 커뮤니티라는 점이 좀 한계가 있을 것 같은데, 이거 이제 쇼폼 같은 영상 중심의 커뮤니티나 영상 플랫폼으로 뭐 전환을 한다든지, 아니면 이쪽 관련돼서 뭐 새로운 플랫폼을 출시한다든지, 이런 계획이 있으시면 같이 말씀해 주시면 감사하겠습니다.
I would like to ask two questions. It seems like your overall strategic direction for your Friends tab has changed over the years. In the past, it was more about making a timeline, and now you're also introducing Timeout feature, and now you're focusing more on the local business aspect. Does that signal a shift in the direction for your Friends tab going forward? Second question has to do with open chatting. I would like to ask some questions relating to the traffic metrics and its potential for revenue contribution. For open chatting, there are some questions related to its potential and how active, how activated this whole service could be, because we don't use it as much. Also, there seems to be some limitation in creating a community that is only text-based.
I am wondering whether you have plans to apply or adopt, you know, video-based or image-based, or maybe create a new platform.
오동환 위원님, 감사합니다. 둘 다 제가 답변드리겠습니다. 그 친구 탭하고 이제 그 저희가 프로필 두 가지가 있는데요. 어, 지금 말씀하신 그 타임라인을 지금 그 프로필 업데이트 한 거, 거기 보면 친구 탭이 있지 않습니까? 거기 들어가면 그걸 지금 스와이프 메뉴로 이렇게 볼 수가 있어요. 이걸 이제 스크롤 할 수 있게 하면 이제 그게 타임라인으로 바뀌는 형식으로 되는데, 그 업데이트한 프로필스, 개수가...... 아직은 이제 지- 지인들 상대로 하기 때문에 엄청 많지는 않거든요. 그래서 이걸 타임라인으로 할 그 가능성은 아직은 없는데, 근데 만약에 이제 저희가 아직은 뭐 확정된 건 아니지만, 이 공개 옵션을 주고 비지인한테도 뭐 공개할 수 있고, 뭐 그렇게 되면은 이제 그 많은 사람들한테 보여줄 수 있게 되고, 많은 사람들이 이제 생산하게 될 텐데, 그러면 자연스럽게 그게 타임라인 형식으로 될 것 같고요, 그거는. 그건 이제 그 뭐라고 할까요? 그걸 소셜 커뮤니케이션 영역이라고 할까요? 그, 그래서 그 타임라인 전략을 바꾼 건 아니고요. 로컬 비즈니스 쪽과 관련해서는 뭐 자연스러운 흐름인 것 같습니다.
친구에서 아까 말씀드렸듯이 대면 친구에서 비대면 친구, 비지인 친구로 이제 저희가 관계를 확장하는 과정에 있거든요. 중에서 가장 지리적으로 인접한 저희끼리는 뭐 동네 친구라고 부릅니다만, 동네 친구를 확장하는 건 자연스러운 것 같습니다. 그것 두 가지가 지금 말씀하신 타임라인과 이 지금 추진하는 로컬 서비스가 상호 병립 가능하다고 보고 있습니다.
This is Simon. Thank you for your question. Responding to both of your questions. If you look at the Friends tab, you know, there's the profile as well. In terms of the profile updates, right now, what you have to do is you have to swipe it to be able to see that. If we change that into a scroll type of an interface, then that will then translate into a timeline. Right now because the updated profiles are not that many, and it is only exposed to your friends, your usual friends, there's really no need for us at this time to make that into a timeline.
If we provide a more disclosure option and have users disclose it to people, even to people that they're not friends with, and if more people are able to see it, then I think then we will be able to bring that concept of timeline into it, and I think we could maybe dub that the social communication scene. I can tell you that it is not a shift in our strategy, but the local business aspect, it's just a natural progression, as KakaoTalk expands its relationship from face to face, to online, to expanding the relationship to people with people that you don't, you are not acquaintances with. We call this, it's a natural progression, a natural evolution, whereby in terms of we call them neighborhood friends in Korean.
It is that aspect of the local business that just very naturally can be conjoined with the KakaoTalk strategy. I think both of those approaches can coexist.
두 번째 질문에 대해서 답변을 드리겠습니다. 그 Open Chat tab에는 그 평일 기준으로 하면 한 1,200만 명이 방문을 하고 있고요. Open Chat은 계속 성장을 하고 있습니다. Open Chat이 이제 저희한테는 아까 말씀드린 세 대 그 커뮤니케이션 영역 중에서 소셜 커뮤니케이션에 해당되는 서비스인데요. 그 활동성도 굉장히 높고요. 성장 가능성이 높다고 보고 있는데, 그 말씀하신 이게 그 텍스트 기반이라고 말씀하셨는데, Open Chat에는 그 다양한 콘텐츠를 담을 수 있습니다. 그래서 뭐 제가 그 보고 있는 Open Chat방은 그 보컬 트레이닝하는 Open Chat방인데요. 거기에는 그 보컬 트레이닝 하시는 분들끼리 본인의 노래를 녹음해서 올려서 서로 뭐 평가해 주고 피드백 주고 하더라고요. 뭐 그런 식으로 그 음성뿐만 아니라 쇼폼도 올려서 서로 공유하는 Open Chat방이 있을 수 있고요. 쇼폼 전용의 그 플랫폼을 만드는 거에 대해서는 저희가 지금 그 아까 24시간 안에 사라지는 펑을 말씀드렸는데, 그 펑에서 쇼폼들을 많이 쓰실 것 같아요. 그 보면서 이거를 쇼폼 플랫폼으로 진화할지 다른 위치를 좀 판단할 수 있을 것 같습니다.
Answering your second question, open chatting is a scene where we are seeing about 12 million visitors during the weekdays. Open chatting is growing quite robustly. As I've mentioned, what now? It is one of our three key pillars for communication domain. It defines social communication. Currently, the activity metric is really good, and there is a very big growth potential from open chatting. You've mentioned that, open chatting is only text based, that there will be some limitations. Actually there are various different types of content that is being used within that space. One of the open chatting that I'm subscribing to is a vocal training related open chat, where people actually record their singing and they upload the file and other people would listen to it and give feedbacks.
You are able to share such voice as well as other short form videos in this open chatting space. In terms of whether we will be making a separate platform for it to support short forms, I've mentioned in my presentation that we introduced Pong, which is a feature whereby you put up a content and it will disappear in 24 hours. I think what we need to do for the time being is to monitor how this product is being used, and then later on make a decision as to whether we will make an exclusive platform for short forms.
...Open Chatting에 대해 좀만 더 설명을 드릴게요. Open Chat 아마 오유엔님께서는 들어가셔서 관심 있는 채팅방을 발견 못하셨을 것 같은데, 지금 현재, 현재는 그 Open Chat은 저희가 이제 피처링한 채팅방들이 나오고 있는데요. 이제 하반기에는 그 관심사에 맞게 추천하는 Open Chat방이 이제 나오게 될 거고요. 그러면 좀 더 오유엔님 관심에 맞는 채팅방을 발견할 가능성이 높아지지 않을까 싶고요. 그리고 이게 사실 Open Chat방은 그 다양한, 그 아주 세세한 주제로 나눠져 있기 때문에, 거기에 가입하신 분들의 그 데이터는 저희한테 굉장히 소중합니다. 그래서 뭐 요리를 좋아하시거나 영화를 좋아하시는 Open Chat방에 가입하시면, 이분이 요리나 영화를 좋아하시는 분이라는 걸 알 수 있고, 그런 분들을 묶어서 저희가 그분들한테 알맞는 뭐 콘텐츠도 추천을 하고, Open Chat방도 추천을 하고, 뭐 광고가 있다면 광고도 추천을 하고 하면서 그 광고 최적화에도 좀 기여할 수 있을 것 같고요. 그리고 장기적으로는 그 Open Chat을 운영하는 데는 품이 들기 때문에, 그 저희는 당장이라고 그러는데 당장의 인센티브가 있어야 된다고 생각하고요.
그 Open Chat tab을 구독을 하거나 거기 광고를 같이 넣어서 쇼를 하거나 그 모델들을 이제 실험적으로 적용해 볼 예정이고요. Open Chat이 이벤트 목적으로 굉장히 적합합니다. 이벤트 방으로도 운영할 수 있기 때문에, 이거는 시간이 지나면 하나의 그릇인데, 이 그릇에 어떤 그 내용물이 담길 수 있을지 굉장히 무궁무진하다고 생각하고 있고요. 이제 Open Chat tab은 시작이라고 보고 있고, 지금 현재 그 Open Chat tab을 한 것만으로도 Bizboard도 있고, 그 안에 이제 문맥 광고도 있는데, 그 퍼포먼스도 좋아지고 있어서 여러 가지로 Kakao에 도움이 될 것으로 생각하고 있습니다.
Just a little more elaboration on open chatting is, you may not yet have found an open chatting room that really suits your interest. Right now, we're only featuring, you know, basically, you know, what we are featuring, but starting the second half, we will be able to make recommendations based on the interest of that specific individual. I think when that time comes, you will be able to find the chat room that best fits your interest. Also there are so many different segmented and very specific topics and themes that people would look for and subscribe to, and the data that we can get gain out of this is extremely valuable for people who's interested in cooking or in watching movies and films.
That information is going to be very valuable, and we will make use of that data to be able to recommend certain contents and other Open Chat rooms, and even use that for optimizing our advertisement. From a longer term perspective, because running a chat room does require a significant level of efforts of the operators of these chat rooms, we will think about how we could incentivize them. We will be experimenting with different subscription models or even sharing of ad revenue. This type of a chatting room is also very helpful for running specific events. I say that this open chatting is like a vessel. Its possibilities is quite limitless.
You can actually put anything into that vessel, and we are only beginning with open chatting, and we expect, and we are seeing that currently because there's also Bizboard and other advertisement services, the performance, there will be quite a bit of contribution to overall Kakao's performance.
네, 시간 관계상 마지막 질문 받겠습니다.
Due to lack of time, we will take the final question now.
마지막으로 질문해 주실 분은 다올투자증권의 김하정님입니다.
The last question will be presented by Ha-Jung Kim from Daol Investment and Securities. Please go ahead with your question
아, 예, 안녕하세요. 질문 기회 주셔서 감사합니다. 두 가지 질문이 있습니다. 첫 번째로는 AI 관련해서, 이게 이전에 비용에 따라서 뭐 타사 모델도 유연하게 내부적으로 도입하시는 거를 고려한다고 하셨으니 드리는 질문인데, 타사 모델을 파인튜닝해서 내부적으로 쓰시는 것하고, 우리가 만든 모델을 사용하실 때 드는 컴퓨팅 비용하고 차이가 어느 정도 나는지가 좀 궁금합니다. 뭐 이미지 모델하고 언어 모델에 공통적으로 적용이 되는 사항일 텐데, 일단은 칼로 있던 제품이 출시가 되었다 보니까 그거에 조금 스페시픽하게 질문드리자면은, 뭐, 칼로 있던 거 내부적으로 에, 이미지 생성하실 때와 그때 컴퓨팅 비용하고, 아니면 뭐 스테빌리티에 UI를 내부적으로 쓰실 때나 아니면 노블 AI 같은 거를 유료 구조로 해서 쓰실 때의 비용 차이가 어느 정도 나는지가 좀 궁금합니다.
두 번째 질문은 SM과의 전략적 시너지 관련해서 구체적인 부분인데, 세 가지로 좀 분류해 보자면은 IP 협력에 관해서는 기존에 카카오엔터에 있었던 IP, 이걸 아티스트분들하고 SM 아티스트분들과의 시너지가 나는 걸 기대를 할 수 있을지, 아니면은 북미 쪽과 관련해서 최근에 통합법인도 세우시고 하셨으니, 그쪽에서 IP 발굴 같은 거를 구체적으로 어떤 형태로 기대할 수가 있을지, 북미 외적으로 보면은 또 어떤 것들을 기대할 수 있을지에 대해서 구체적으로 답변해 주시면 감사하겠습니다.
Well, thank you for taking my question. My first question relates to your AI investment. I would like to understand as to the amount of difference that you are seeing in developing your own computational models, your AI models, versus taking what is already out there and fine-tuning it for your purposes. Of course, there are different AI models. There, this large-scale language models, but since you've introduced Carlo 2.0, I would just like to, you know, confine my question to Carlo 2.0. Would like to get a feel as to what the amount of investment difference is when you're internally developing it and when you are making use of what is already existing out there. Second question is on strategic synergies that you may be able to gain from, from SM consolidation.
First is on the IPs. Can we look forward to a synergistic impact, between the SM and Kakao Entertainment IPs or the artists that it has? Second is, since you've also set up an entity, a combined entity in North America, can we look forward to their role in discovering IPs and discovering talent as well? If you could share your views for North American market as well as ex-North America, that would...
[Foreign language]
Ha-Jung, thank you for your question. I will answer the first question, and the CEO will answer the second question. Regarding the AI service, you asked a question about the cost. There is a somewhat counterintuitive aspect to the AI model, which is that the larger the model, the higher the cost and the longer the response time. Especially when switching to Korean, since the global models use more tokens for Korean, there are comments like, "In the U.S., the response time is not that long, but in Korea, why does it take so long?" Therefore, a model optimized for Korean is needed, and you could say that a lighter model is actually necessary. In the United States as well, Meta recently announced a 700 billion parameter model, a foundation model called Llama.
I think that model will be open sourced and fine-tuned, leading to various models emerging in the future. I also believe that efforts to fine-tune a Korean model will definitely come out. We are not excluding the possibility of trying that ourselves or using models from other companies. However, as AI services evolve, they are becoming more like cloud services. It's not that one company dominates the AI market, but rather that many companies provide such services and build them, similar to cloud services. We are also building our own cloud. The reason we do this is that using cloud services from abroad is very expensive. Kakao is also building its own infrastructure, and the cost of serving that compared to using foreign cloud services is about twice as much.
This is because to operate a cloud service, you need to support numerous customers, which requires developing various functions, and those costs are reflected, making cloud prices from a single company very expensive. That's why we need to build our own cloud, and we approach AI services from the same perspective. In terms of specific costs, it is not appropriate to explain everything down to the penny, and I don't think I will. We used Kakao 2.0... Last time, we held an event using Kakao as the background for profile pictures, calling it Heart Background Pictures. We have confirmed that the costs can be significantly reduced.
Yeah, this is Simon. Thank you for your question. Responding to your first question on AI investment or the cost. When it comes to AI model, it's a little bit anti-intuitive, meaning if the model becomes bigger in size, the cost and the response time actually all increases. Also, there is the Korean language element. Compared to global models, basically, Korean model requires more number of tokens. That's why people say that the response time that they would get, when using English is much shorter compared to when people use Korean, as the language. That is why we need optimized, Korean language models, and that is why we think that the lighter weight AI model will better suit that requirement. Meta recently had released Llama, which has 70 billion parameters, and they open sourced it.
I think going forward, a lot of companies will start to fine-tune Llama and will do many things with it. Of course, Korean language localization will also be attempted. We are not excluding any possibility of using other companies' model. It is just that AI service, if you look at how the market is forming, it's very much like that of the cloud market, meaning one company is not going to be a dominant player in the space of AI. Just like you see in the cloud market, there are multiple number of cloud providers and cloud players in that market. The reason is because for us to use a global cloud services is extremely expensive.
Compared to what Kakao uses, our Kakao internal cloud versus subscribing to an overseas global cloud provider, there is a twofold cost difference. The reason is because there's a lot of features that needs to be developed in order to meet the requirements and the needs of the client of that cloud service. That is why we need our own cloud, and I think same will apply for AI as well. It will not be appropriate for me to give you a specific cost difference between our, using our own AI model versus using that of the outside. Through our experience of having this profile event whereby we used Karlo 2.0, we were able to verify that there will be cost gains from this.
[Foreig language]
This is Jay, responding to your question about synergies with SM Entertainment. We previously communicated that the three companies, including Kakao, Kakao Entertainment, and SM Entertainment, will cooperate in areas such as: number one, strengthening the core infrastructure of the music business. Number two, diversifying into the secondary IP business, including the future industries of AI and virtual human. Three, cooperating across the platform that really cuts across the music ecosystem. Four, expanding K-pop to the global stage. In early July, SM Entertainment set up a professional and specialized publishing subsidiary entity, setting up a publishing structure that is supported by multi-label strategy. Going forward, we will be able to really leverage opportunities of synergies, having different composers and lyric writers all collaborate with one another.
Also we are looking at different ways for us to collaborate and cooperate in nurturing the future next generation artists as well, and also cooperating in the existing IPs as well. On the global side, we've set up that global entity so that we could secure efficient local distribution channel, and by expanding partnerships with overseas labels, our domestic artists will be able to go global, and in that process, we'll be able to provide them with efficient support. We think that in terms of planning and production, with these large, global artist studios, we think that we will be able to have ample competitiveness, and we are looking for good opportunities.
Uh, 또한 AI 버추얼 휴먼과 같은 이제 미래 산업과 관련해서도, 어, 저희, 어, 통합을 통해서 이제 더욱 규모감 있는 투자 집행이 가능해졌고, 어, 카카오 브레인의 AI 파운데이션 모델을 버추얼 휴먼에 활용하는 것도 충분히 검토 가능하다고 생각을 하고 있습니다. 어, 팬 플랫폼 사업 또한 이제 SP, 그, Kakao Entertainment와 SM, JYP 등이 이제 DearU를 글로벌, 어, 팬덤 플랫폼으로 키우기 위해서 현재 여러 액션 플랜들을 도출하고 실행하는 단계에 있습니다.
마지막으로 북미에 저희가 통합 법인을 설립한 게, 이 북미에서의 성공은 유럽과 남미 지역으로 이어지는 교두보 역할을 할 것으로 생각을 해서 저희가 먼저 북미를 중심으로 시작을 했고, 저희가 또한 중요한 시장으로 생각하고 있는 일본 지역은 저희가 그 NCT의 유닛 그룹 NCT Tokyo를 런칭하기 위해서 저희가 서바이벌 프로그램 NCT Universe를 YouTube 포함해서 국내 OTT에서 공개한 것처럼, 저희 국내 아티스트의 글로벌 활동 확대에서 한 단계 더 발전해서 K-pop 육성 시스템을 글로벌 시장에서 뿌리내릴 수 있도록 저희가 준비를 하고 있습니다.
Also in regards to the future businesses and initiatives of AI and virtual human, we will be also able to apply the foundation model that had been developed by Kakao. We now are able to make investments in scale, with scale. Regarding the platform business, already SM Entertainment, JYP, and Kakao, in collaboration, are working together to grow DearU into a global fandom platform. At this point, we are coming up with different action plans. Now, for, in terms of, Japan-... Basically, what we've done was that we have been very successfully run a, the survival program, where, survival program, and we are able to really, you know, exert our efforts to build out a platform for nurturing K-pop artists, and also to build out a system on the global stage for nurturing new artists.
Also for new artists, basically, gaining our local artists and exploring them has become an important aspect. For we have taken NCT, the unit group, in order to launch them in Japan as NCT Tokyo. We have, through YouTube, been airing the survival program, NCT Universe, and that has been well received.
[Foreign language]
Before we end, I just want to respond to one of the questions that we received before we started the earnings call. We mentioned that our DAU was up, so there was a question relating to actually the time spent, because some of the survey data showed that the time spent has dipped. If we look at our figure, compared to June of 2022, there has been a 6% increase in the total time spent by the users. This is happening despite the fact that we are now moving from endemic, pandemic to endemic. Basically, people are spending less time online on the back of that, and also a Kakao-specific event of system failure. Despite all of that, we are seeing the usage and the time spent by the users on KakaoTalk actually going up by 6%.
[Foreign language]
Well, this brings us to the end of the second quarter 2023 earnings conference call for Kakao. Thank you very much, everyone, for joining us.