Cheetah Mobile (CMCM) 17 Dec 242024 Q3 Earnings call transcript

Good day, and welcome to the Cheetah Mobile Third Quarter 2024 Earnings Conference Call. [Operator Instructions] Please note, today's event is being recorded. I would now like to turn the conference over to Helen Zhu, IR for Cheetah Mobile. Please go ahead.

Thank you, operator. Welcome to Cheetah Mobile's Third Quarter 2024 Earnings Conference Call. our and are Mr. today us company's CEO, Chairman With Mr. Director Fu our CFO, Sheng; and and Thomas Ren.

prepared we be management's that AI Following remarks, the Please Q&A agent. CEO note section. script will presented an will by conduct a

Mr. now we to make harbor in Sheng. safe time, forward-looking Fu applies our statements. At our to today our call will also statement earnings which this begin, conference ahead, I turn release, as go Before over Fu you to the CEO, the refer call earnings we Please I will Sheng.

cases joining demand legacy resilience Thank using weeks, their especially markets you driven and robotics improve past the today. I everyone. and overseas again Internet expand our the growth Hello, Europe to offices. efficiency. In partners and overseas of once customers staff into service our continues expand use visited achieved our revenue owners in of restaurants, for and Internet as our QX, us rise, Asia. business. Industry in wheeled to Southeast from well Cheetah consistent as This to and service results the help in growth service more Mobile markets factories businesses. often robotics accelerated Business in for strategies and by and robots many hotels, robots the are

our have strengthen this for I That's why am important peers. building presence, our effort a us spent further expansion our with is apart significant on strong will to very set fact, In distribution our partners still in meeting as Europe I initiative. local from local network it today, global

shortages. helping l learned them local with solve our that conversations partners, robots my are labor During

warehouses turnover. players' employee center to Japanese using payload and have and reduced Meanwhile, told other are switching our robot for than transit to much move items use. assembly goods to low more shared Cheetah's a Some September, fulfillment we us new offerings, customers deliver products. In factory absences Europe are from our that autonomously robots This some launched in reliable robot customers lines. robots and can

the optimizing important the needs currently of an input in local will product robotics the business become of are to service highlights this and receiving our This from overseas believe feedback part product partners. meet importance better We in customers We XXXX. of markets.

them robotics up the still than stages, to in industry it humans ever and be essential do and early deploy for before. Robots is trillion-dollar this and will are enabling While by market. the its to in growth robots first, making will more easier belt helpers speed happen LMs will markets a becoming tasks

investors, steady are Through we've growth. of interest our smarter conversations with in noticed revenue and robots our driving LMs a lot making how

LMs share Today, I so our products and next. far in coming what what's we've achieved will with

interact far-field are our thanks strong through our recognition. robots we to user to way Robots the service voice using improve well. already hear First, LMs voice, as

Now overall users' understand they making and smoother. more questions clearly better, with experience LMs, much respond the

on also more boosting barriers, just For these are instance, on markets. investments attract because taking owners language restaurant the breakdown robots don't robots our for return deliver roles, international to we help busy in doing hours. can It customers, by voice-enabled food and LMs restaurants, expanding

customers [indiscernible], on tasks lets robots working a system set voice using that up for also prompts. We

Without ordering the robot The will code. wants functionality be customers ones place kind and to lot a skip nearly is each impossible if tell to orders would of complicated writing late. this to X p.m. more robot X:XX closes anyone LMs, can a or would system the people allow without even require p.m. order at the you example, kitchen to if of each For go table, before see to check at table food

enable Factory area days R&D deliver map on whenever multimodality owners time around LMs, key are up more to that want is and reduce robots and easily our it autonomous ready. map One from the into factory own can about X to update goods. move can robot once locations, as large they on to the can locations. is just driving. initial look hours. improve change the They then to navigate factory key takes robots the robot to our we needed. our these can team a map our mark thanks out their working to to models using Based we dramatically our Secondly, testing, are the set X to

using driving in technology, cases, further. to it already autonomous are vision-based We expand some and plan

avoid achieve technology more an our handle their complex and robots we their on user navigation end-to-end Over to entirely people obstacles and vision-based understand system. time, our robots to surroundings. environments autonomous instance, own. This allow For to driving aim will

robotic we're specific to help to Third, adding robots arms them our do jobs.

items, access In automate control up these Some Others is can press systems useful is for floors. in particular, in sort on business of factories. elevator security powered overseas their owners tasks. arms due making models, routine things which use adjust easier arms can and buttons, it the for concerns. which to are to multimodality delivering reluctant These to where between pick by are great device, markets,

it models. XxX When we calls an use developing experts we model At also same of In mixture covering are discussed. Japanese. many advanced Korean and including own billion time, November, comes our launched through LMs, to features some Chinese, models English, have support of we languages, we to API the the

our especially it made use it open and to features. source agent robots, have OS power the We

LM-based recently ARDS apps. and we is in smaller platform, on built and consisting their Additionally, building our prompts on service enterprises an into turning ARDS, LM-based insights applications. of We indoor models top data to AI-based trained data LMs autonomous control for and device have introduced driving to arms developing support robotic applications. LM-based

on portfolio. continue we will our So LM-based enrich our far, feedback received and have customer applications, to positive we

is relatively businesses to SaaS offer gain use to goal efficiency. Our standardized allowing products, LMs

international highlights, to into good growing the want financial for service to for LM-based bring years expanding Before and growing market to We turnaround. strong made from the revenue apps. is areas into This PC just the Mobile's stress expertise. well beginning the our in have and robots losses. experience already markets, is phone a progress Mobile Cheetah of and Thomas We've Cheetah I handing tap solid cutting mobile the as position and of we this, as call in over LM

losses unless QX robots sales Thank and on Number solid everyone growth Sheng. strategic in of terms. In X, QX demonstrated on that our XXXX, our Hello, reduce third X note enhancing are revenues results revenue call. you, revenue Fu the quarter number execution Please objectives. business; of the sequentially. In the both compared legacy money increased and efficiency in the Internet RMB operating quarter X, to XX.X% otherwise, year-over-year, QX. consecutive financial to total marking amounts in the and our accelerating accelerating XX.X% all XXXX, XX.X% by driven growth of operational stated our second

a robots Our be key of to real-based driver continue service growth.

resilient legacy growth Internet business and revenue our Additionally, achieving remained solid margin expansion.

RMB gross RMB in to from in year-over-year XXX in XX% was Our million X% made million margin in reduced the QX. RMB non-GAAP from and XX gross loss with in the third Non-GAAP profit quarter-over-quarter rose XX% further profitability, and quarter million we XX QX Non-GAAP QX. expanding RMB operating XX in XX% and QX progress. million to quarter, XX%

of to overlapping with staff Cheetah and and the focus and OrionStar, managing expenses. services We Beijing continue Notably, costs on streamlining functions. we teams consolidated

professional cloud cost certain we For costs G&A example, and service fees in bandwidth QX, legal and reduced costs, related labor and to and operations.

decisively investing our are to service also AI We using in AI, enhance robotics business.

of about our XX% in XX% R&D our revenues now QX, and quarter-over-quarter expenses For in R&D. increased non-GAAP are invested example,

developing that As advancements. Fu attentive and in said while profits, generate latest to products we immediate the revenue on has Sheng the we focus remain past, technological

and Operating this XX% excluding improved year. compensation margin, the is which trend quarter-over-quarter. expense segment highlight Internet continued business, XX% XX% our to Another X% legacy for of from share-based last grew year-over-year

cash we As closing, balance a million are cash of expanding USD of maintained about million. In solid million investments X,XXX XXX XXX long-term revenues RMB losses. with or million, making of progress narrowing XX, USD or September we XXX in equivalents and strong and XXXX, RMB sheet

remain we significant in efficiency our see potential losses a investments you. We are service of market in we AI integrating AI and robotics At driving confident into LLMs in time, disciplined our Thank in operations. our same business. the reducing because

This ready Q&A the for Operator, session. remarks. prepared now our we are concludes

robot for the My set May have I focuses shipment robot terms for business. question ask on Cheetah's volume, proportion? the been business and of specific goals revenue [Interpreted] in growth what year XXXX, revenue

be Let [Interpreted] some set overseas. specific many answer of for formulation. process of recently XXXX, the important me in should visited this our I also goals still have to channels question. be Well,

the overall like think be this. determined can direction I general

term. and said larger have to scale also hardware and that specific sorry, this growth, because is than robot doing these in because the the continuous in it of of back be capital experiences the as bid construction, then, back and of but is That market gradually overseas process a our base explosive in the next X.X is increase. investment assumptions, of way. Then than tell summarized then. growth specific highly industry that years, software I that ones accelerating the we certain it that in fact the combination should also will a It's Specific in don't on Internet short Thursday, much income be entire for increase channel business everyone lot we then -- implementation definitely today, think this the exponential combination in markets of larger of and the be very the robots combine in have aspect should I the as of growth the the it's by physical Cheetah, that Internet. idea right? think the was is and revenue in in than today, is wave. will much we the the should will our and this goals hot setting the make proportion First,

at including we careful detailed This of more for more long-term will top our be some within is system the example, big deductions. that, X we service commission discussions So think income providers. and X few of entire for a top the can the goals, at to the be for years. company need half goal. the the for we globally, goal. may, able make X And among such robot And look least account We business, than I still

performed gradually predict business business, and revenue that both revenue Will trend of [Interpreted] Internet stable. very how slow profit steady should Cheetah, profit the the well and decline? The is it cash Internet a down business or and margin with quarter, ask XXXX? growth May this trends I margin in cow the remaining of we show

have robots the data? scarce in What the of many I'd training, globally. facilitating few do how data? robot problem In can fact, training like about Cheetah overcome models of process robot In roles a deployed play questions to training. raise large robots we

So of we overseas, in training robots, for the manner from in how humans level? Furthermore, intelligence future, of self-learning far Compared aspect this of robots achieving in is training enhance their is performing train videos to a goal us are robots by Cheetah continuously to continuously the and possible in the data-driven it how to the terms domestically? gap of big leading robot companies tasks? watching methods

[Interpreted] understanding one my use question of rather is technical to to company asked. Well, complex a the you and help explain our and you first it. analyze that simply I'll try

data robot which One basically within components. aspect indoor that is technologies is sales into divide on models component to has One some self-driving. for This be aspects. navigation, can equivalent an requirements Regarding and robot it environment a relatively the limited we several before. because the training, indoor solved engineering several divided into space based small the is certain been matter is through

we doing, will addition and become, last underlying say sensors on to reliance data smaller. say right? implementation the of it, become to how of its robots currently will is time, real This what indoor time due will become that However, to more is its navigation are

things of indoor doing and One recently promote we mathematics is robots the navigation are to chemistry. our for of the

being navigation vision. robot next can higher-level gradually through This is chip, purely indoor also achieving equipped also Our advanced. be with a

of than Tesla's less underlying was One processing, at significant such radars This also compare can vehicles. and mechanism and the as mechanism, it. important handle important various that self-driving urban seem the the end, In many actually can of new emergence SSD. Look for made is large things. energy reason You end-to-end the this be implementation progress, LiDARs and models. to now transformer and adding it today's multi-modalities

So this aspect. one is

on have highways a not we high. faces running Regarding data, have and those conditions are as you this scenarios robots, in of requirements aspect just it they various said, nor data. speed as complex as before. considerable can of And as This deployed actually are many amount road the been achieve the

on a we already We this think major demand, providing indoor company has for have part So like in that services don't this part, ground a of many not think in every robots especially us day. is we okay. Data that it's any. the navigation,

a second is the might currently concept popular. one quite Really, self-learning be that skills

expressed to this, constantly take code I constructed before a at unlikely on more human-like aspect of say kind the step is Some data channels core. truly land For present. The the many computing data think years. a step. I is loop. commercialized and products, theoretical that factories our previously, there actually don't the occasions is it driving it's We is XX viewpoint definition, by it's arms core, long for new become this not on at the automation believe I or have some circular approach dual-closed robot, scarce is humanoid it without all to based force you this, the right? system performance indeed think truly The relatively in at including in exactly still what X achieve they it for clear or mentioned, were It's right? still because a that and particularly and go commercialization have way robotic to the say but

you that still become to commercialized them can go. demonstration of effects for are many is good, Although there truly a products, to their way long but see

a So it's more about pragmatic approach.

combine will our we scenarios. So

the and specifics some simple real because won't we the technical it's I into some our example, rules. in robotic this start For go arm to interaction completing of by world. tasks between related the

up do everything problems. with universal idea perfect is come solve and not that a to Our product can all

the Our themselves have combine idea is scenarios on scenarios instead because ground we we more. the today,

and of one we think collection step. think terms continuous this might we I will optimistic It investment. after the in complete of be think as this may complete time. data, another requires To not training we as today by scenario But step some

implement. foreign startups, they like meals make see because complete it's and venture way You to with think very also product some they up I capital come too If that this simple, basically there many cooking performance, something can away, are it a very right laboratory in difficult. the constraints with are easier tasks making but is in

anyway. low. watching So what that we time, a was are like is over its did question, it's human it how And example, achieving Most for the I in for is success of then we improvement? moment. Well, very the it What XX% tasks far videos. an something. videos do? you quite goal from all still the it it But are Maybe now second or published, demonstration completing papers give far. would of robot rate by learn can see was

specific there be situation. still this will Of progress course, is and in a XX%

not the For a desktop, it's example, app. on entire the area, but desktop designated

X-dimensional been X driving In working surface today, a was road urgent as It's optimistic level, autonomous the driving once be it now. recognition autonomous teams right? that So since XXXX. today, have everyone no very the LX also at XXXX, and announced autonomous XXXX company This it's possible, been would and Tesla that achieved taxi right? not as landing, we on has such driving Many global years driving. will beginning, achieved, have thought. Consider including it situation thinking was cars. But in autonomous

Because for to breakthrough, domestic are also learn is large more big companies frankly judgment the because I use achieves can high. The how how mechanical difficulty for today, perform the mechanical not at not those not our a models, once think it's they to humans this significant. foreign for because train including major the and involved. to and it it's how mechanical efficiently. updates optimistic for is tests the robots to this underlying after that trend. is more shared, how autonomous difficulty algorithm, with X-dimensional algorithms, are even data more than driving watch fast engineer more I the very time technological developed, everyone domestic a with on But in of the for algorithm that structures gap think to self-learning is system a real is be lies between more This itself as ones, training level, AI

an in least is no gap data there In engineering. fact, regard country. of this advantage doing have At in teams this Chinese the large-scale kind in

I there don't domestic So about APPs. the gap. is think today, a This big is existing everyone aspect of method, model the some university including training using

really aspects other if the it Maybe good. fact, exists, the actually might in are be productization In gap some path. and new quite

difficult be very with my innovative a For very speed think methods. not example, new method domestic up emerges, This once I some is come the is gap view, It's of There method a certain country. some if gap. much emerging, But will fast. personally. is to there the in there particularly is follow-up

losses of what ask Since year, losses reducing profitability? plan achieving clear specific company Is this has [Interpreted] the for beginning there reducing I pace every the timetable May been a the for continuously are? and subsequent quarter.

is losses [Interpreted] definitely our priority. top Reducing

cost and all shouting today, So enterprises reduction improvement. are about efficiency

because the For a and our certain down past we this this R&D bit. the and to development some definitely efforts reduction this to like [ due into will think large their support market But agents But including XX a their large service their significantly review and this to technology, loss language ability, in progress slowed underlying now significant it externally intelligence and models language profitability. that expand we I We language model I of want markets loss be expand for years. the that and round, some participated has achieved interaction reduced various models, training implementation we and we plan will in seen internally be will wave, the to just a timetable of robots. Such planning very interaction put disclose may loss reduction ability, of a market the more also costs in we the some training see the in related not markets. whether including of Well, able in and the compared our that have the satisfaction say little us, for opportunities surely it's question for the and pace after development because to of reduction, certain about large earlier due we of MV research ], ability, task ability I we friend's scale have year. this changes clearly have robots the in their business decision-making because to think various

we to need to company certain such aim R&D, an make will some for doubt. of profitable maintain So major a If arises, flexibility. to in is create the This more also may we and we investments goal course. beyond shareholders. Overall, opportunity have make value definitely

help [Interpreted] seen agent you this How AI difficult view been technology? recently, AI For Zhipu orders and have able independently users distribution to and example, Can in that in do large models value of agent? the search the Maps should application AI CMD? this? place Agentic agent has the or consider the we we is How AI Meituan. of on How accelerate Baidu

[Interpreted] Well, is professional. quite you. question This okay, thank

some human reached then and model. and to need was And of popularity smaller for it not of a the large part it That before? a the recently, of has emerged become of because appreciation planning think kind thinking kind I is traditional certain types the has software. term, use foundation. level, an AI of especially What we guide popular more chains essentially software agent, of which part called a like human this that model's essentially ability to

is on including is, points for be first in its of I a might with language to a effect, today, improvement think saw like wrote wanted we the is large when user a experience this XX have lot of it end to every indeed Zhipu, sentences handled to a complete at cost a great of software amount step. new the It and the That it. like of logic of I this Then of by shift software. we on models. code the so-called the previously, reduction paradigm that, opportunity. large entirely and and what this the a in logic R&D rely you library, kind code write Just But after software

does say have say regard. is also we, have want achieve or have these that that an queries to our to stability ourselves, I this the high that satisfaction business to level of whether has it for able place this and agent, traditional -- that also some of matter Internet But seen AI of What complete in attempts mean? we I orders, to systems.

what idly For But I end point or can models an but want discovered mentioned the doing large crucial the when today. problem it we landing end to were we the for search of is do for? if I C this on it definitely ourselves. instruction, give a it very This is example, in point this industry, not B

traditional their to similar, you For is it, example, your in are when code implement if XXX% humans place an use operation very precise you to select right? because operations, order point a basically,

might fail. it Sometimes

expectations. your meet not might it Sometimes

sometimes But And hallucinations. They you what know. know might they don't large it give models have errors. don't

this aspect. land, an be agent to lot So a in AI is effort needs that point for of this to invested truly

many It apps this. have back CMD your Coming large have begun landing can, some to it already on started apps. the model CMD? the on accelerate especially Today, question, of definitely -- applications they to use can

still can examples like of a give will Yes, For has distribution, example, educational application I clearly some wave value image types. bring few type started. practice think it translation This for I and manufacturers. the a for

them, capabilities the not the personally, about So will strong. the my a because be manufacturers company it on large aspect domestic this China very, startup fast. very in C And is as I'm this are for also in end landing of view. But extremely particularly optimistic

a others it quickly you should in with, you. innovation that Thank be followed. So an if small area familiar make are

the ask the of large models to questions enterprise about a few in level. like I'd [Interpreted] application

application as ask our application data models the management We areas have of the office a the I tolerance services. with willing limited explorations at conducted for by large such large pay of May how is specific models large scenarios and present? the rate. observed In year, enterprise to application scenario, models that still Cheetah of to mainly covering working sales in many training, this customers error has higher

advancement of of can large models, enhance emergence large replace the and hallucination model manual the this enabling AI models thus of the technology, application to large completely effectively the of or With agent, improve especially problem partially work?

Yes. [Interpreted]

So very itself revealing. your question is

have I covered every think point. you basically

needs I pay how First, the redefinition And completely large attention the depends much to high. willingness application input/output and is the application means to ratio has think one essentially, of customers of of this enterprise many the for be internal ratio some this high, models it input/output of on positions. is them. the Because enterprise that there of relatively that that reengineering is can point bring to processes

specific What last large that ask to very year, example, is see more willing a year, if it some enterprise it manually. in you to usable of Instead, models, is they than indeed That value, like of enterprise, and a provide longer together. many accept promote current inaccurate, slightly is more pay insight, little. lot authorization privatization if rate about is in as it large a models, it, of the such still application relatively the the of there for spent But with it something it is exploring For roughly directly for is rate it in say know think general training a the a it customers whether kind accuracy for a achieve training, can of and is what high data money large about enough question can number of are okay, XX%, no model. seems on This idea not is of these how obvious the or industry an authorization summaries or your more is that not it. Currently, high tolerance an it large more they sales. Because that pay they error that to can or aspects if I rational. becoming for scenarios even enthusiasm of entire thing people enterprises right? and some I still well, you me, for

because enough. is theoretically, a exploration is today, customers is is enhance and can uses to hallucination it the vertical it confine model we time, difficult models, test wave, the planning to This traditional Yes, in of combination therefore, the defining model just next this, this gradually application vertical, large the if in model will a of the do everyone large ability end large particularly replacements large after question, gave application, effectively of this application of it is and satisfy is in improve more achieve. investing codes its model environment. the here, the if the a current problem large firm or large the will the it of the theme level. When in especially has also definitely More an of model B our of that error reached when ability for in this with decrease, such this probability I particularly long At the example time.

Let me give an example.

the if that overseas iPhone, know bought version right? try. use you For example, you the don't I can one also to especially bought latest

anyway, little. of what including is point, idea are have replicable the this make clear our one And enough then penetrate are find vertical companies Microsoft will experience to enterprise the I think, very plots like next as reflect them and do direction. But can an But move and long. pushed, on slowly truly And business they The these because point. very to experiment what that it and too that they is broadly this large they application. a You applications us, think I not both define must we then is in moving point.

understanding. my Thank are these of basically, some So you.

the the application model on a law. [Interpreted] industry? the Recently, large many your this? the view What's law about slowdown of the development the of the on scaling discussed What scaling are people impacts of lot slowdown of have

used X least months, precise Sorry it. But point be although some for quality But is law the usable, unknown. training it. at industry, I make keeping of on especially It's not might everyone about are certain. of data, industry, [Interpreted] the on in slowing it basically large less XX on an this that Including key the the been has is to in the about the disputes model this. talking basically, amount approximately, can about is Because or right? the data, insufficiency there not itself past or high is not recently, our now some heard and its XX. still is this everyone in model Internet, what down and whether is good There X probably But well, right? one around is that data. Because is data eye

months, likely quite has period been a XX.XXX. saw released yet, before, not example, of globally, the for This step imply fast has X this easily that XX model of Each of that certain right? original a You mean? sense, been this, that is then For but model the time, models capabilities been and the was the enhancement does can top-level will might X.X a productization see today's Before this in in it's past. What now X right? time. time. at it's or basically increase in capabilities least days, X.X, GPT we It GPT not long the GPT

projects some it about they didn't burdened when you. a time. sell. of the indeed capability out, to these But slowdown in at capabilities This do companies the good were today, about things out, especially to added least industry. it expansion many really, scaling the large us the law at down applications themselves, that capabilities before, know but was don't my better this capabilities, they it. the think not, slowing when down. But were like a models talk and for startups the of for thing this of developing not this the utilize or due the thinking in because applications. in When After There and the growth how is major is top new came growth agents. when capabilities doing you did rapidly model model came model reality definitely we is finished, this So today, with X.X, We the such that idea is let's top everyone slowing it for model whether then is

a we competition ease, for in how the combined doing scenario with be We customers just some the good is the end, best lot spend We participate is with thing because the do to model right? needs for us, game know lifetime. We therefore, relatively think of way we can So Then is. in be a also application early the satisfactory at market today. the polish to still this this to to a on it scenarios. essence itself We still this job effort and it's don't Because anymore. beneficial entered selection us.

think I good companies like for it's a us. So thing

don't for this the have means what for Well, a such I As you. industry, high thank perspective.

the wheeled [Interpreted] rather in that robots. extremely We than stage attitude on field achieve Fu's can towards current large-scale is understand focusing pragmatic, that robots robot at President development bipedal of applications the

promoting what on example, However, the the For industry? months? changes products of we have have X past implementation Cheetah. of robot the competitive also how robots? large robotic the noticed intelligent landscape to in would you robot defined integrate are that over Cheetah models attempting embodied trend by do effects you robots to do industry view future view the as and arms launch ask How I robot the like

[Interpreted] addition I robot several quite viewpoint, industry, become the a the the but I robot past In Then has humanoid has Tesla's still times, large are too do feel? increasingly term. popular over months, model The the popular. to due broad robots to become my few how of adhere of concept changes Optimus, the

the humans of using people fantasy all create they mention that a when to it, solve can So product have prototype the problems. perfect always as

is look sensors, comes back. One were X Google's to there route to routes won Waymo, a was autonomous driving. which by achieve at I autonomous initially, with highly vehicle When many the functional LiDARs can having path pursued very, to very In it driving think difficult. We perfect expensive achieve. strong quite very This competing. individual in was least fact, led past, this and algorithms stable

better definitely so much the sensors. having second experimentation [Technical advanced one was seems continuously approach are that exploration continuous achieve The of start right? The the to Up the the and with it, today, which better. I would that quality initial assumption I Difficulty] didn't most with could was other add that the Currently, found perfect was scenarios technological it than current can is with and changes that route Tesla's approach Eventually, that is training. basically highest of time, engineers sensors. it in-depth data, this visual we sensors, at think say, I

same So should everything. the be I and it think is I to have can a different for robots. unlikely It situation am complete where

more implementation. still or So a the need between they think trolley considerable assist concept the humans, still popular is a to tools concept gap although of some was and applications, practical a like to drive or right? very achieve today, to recently, car practical use there I robots Just

is problems. Regarding the in fundamental issue. a It such in large is it hallucination algorithm It problem fact, like models, this. mentioned has

are somewhat it cannot to or machine such one very But occur, difficult Sometimes will when comes for acceptable. with And in implemented, it answers be interacting errors. objects, the actually in right? the action performing there errors make it to language, be any

a is in in find will to it, supervising and your little XXXx mistake customers' that XX no and make year day, affect over or thousands a cannot will it a for restaurant when confidence has whether mistakes example, there A scenario. they you recommend one For scenario, hours you others. of some to will in reception operate

scenarios, am helpful actual But combination also I robots. So in the large humanoid general of advanced I to robot. think why definitely very are should comes it it about be models when with is optimistic which step-by-step for not direction implementation,

there structures software. hardware concept structures pragmatic moving I a big in not optimistic today and highly very This, which mostly structures very that fundamental robots the an robot supported and broad. is real future this, for might think continuously. I progress a car, am insight is that has hardware like are a logic the I pattern Well, is moon gave, believe complex. like of inside it polishes more is the industry of and such. hardware think Moore's the of a not mechanical that with then competitor by I manufacturers integrated emerge will It eventually, within proposing Law. question this be final It will because product system example The a mechanical landing don't is towards about the has very many And system succeed not and operation. a win. I it combination can

Software there over XX time. around double have this And Hardware its months. can its smart also performance cost in revolution for Cars is years, has cars XXX of the have or to been progress gradually. right?

continuously scenarios in path, own complete is future, in features with based high-quality firm greater we So scenarios and actions think some and achieve on and our then some urgent some to to pain which the reliable expansion. humanity robotic to I anyway, points combine some highly are add with in and arms needs

Okay. Thank you.

that has if cash buybacks amount plans a May share its the [Interpreted] ask dividends in or are large noticed the We for books. I net on have company future? any of there

Let Thomas. CFO, [Interpreted] your answer question. question. me I'm Thank company's this for you the

First towards of had an the Cheetah shareholder has importance shareholder returns. always all, also attach to great attitude We, open management, return.

of are buyback important future, back carried for whether were given in were have dividends Historically, our and investment consider. be share based out there or means, there we many we twice But and that can returns, to distributed plans. X need then projects the shareholders. factors also dividend exit our multiple distributions have see that to the fact, other dividends will on cash which obtaining we Those

to mentioned we including Vice development robot to of same be have business also still also is is environment investments For large President time, in currently earlier, actually uncertain. that make models our our AI and overall economic the certain from the this example, requires investments. transforming business At rather technology, feel we the also as CX some

flexibility business. future, financial in continue reserve for sufficient a maintain plan and you. the the will also facing market the soon a the possible. to we important approve risk quite an has and the announcement in For to cautious market relatively future, that when fluctuations. If the we dividends, our strategy, us, maintaining buybacks of ensure will resistance sufficient relatively is share as Directors as company the Board ability cash Thank for Therefore, development of make

actual you various the generation, large perspective generation of operations? Or reasoning accuracy the of more of development China productization felt any models towards extended significant capabilities content and of speed large Have beyond models rapid model the models themselves such as in and present, ecological has and the gone and At performance differences is the among [Interpreted] from construction? competition costs. efficiency, in

them me model evaluate should themselves and them You for know to try and which can forth. it's feelings. often ecosystem back that is rather better. their own the domestic is and [Interpreted] product for use complex, have I appropriate not Everyone

differences your good. in the question think it's some I However, are there true that experience. Well, very is today

do on good. quite be the you to more and effort high-quality model it launched data public will the has competition the of because exceeded data. capabilities can themselves models the this comes large spend relatively data depends today, beginning Internet the ability such, think this itself a Treasure, Data the I on function and from teaching called of very we model ability model of because the today when itself, you if on it's And the ability this find model that as engineering recently the more work, your

X and XXxX Just A billion like X XXxX, One mode. we have parameters. the models. actually with is XX

ecological we quite today's this gap competition matter model. stage, of to increase underlying the productization in results the because this eliminate towards at the in to can user the of still the certain are In fact, competition also it models think indicators can of good on today satisfaction fact, the improvement it's almost move that truly the attention product the some must and construction. Whoever comes experience in data. large mainly from and investment I list And here, are your

affects longer I the matter. no think this

one one is this Look then another one at on up the top, list. on top. comes a is Tomorrow, Today, while. that for and

moved passed. productization and think already In may It's towards fact, this matter have construction. ecological I

the that [ I asked interview, ], the was said saw lacking? I is For Charlotte this AI I fast. very this think a it He competition, Mang when someone speed in was right? model, large product, recent think what

a ecological now So from stage. obvious has comparison technical or product shifted to initial is technical enthusiasm comparison. This comparison the it very and

will year it's explosion much not that In be high some has ability and next difference entrepreneurs further. ones because application mentioned or agree to see that So isn't model's easy year principle, relatively have the this explosion. of level and point everyone, domestic improve there I recently, a with and reached among ecological the I

Just efforts experience of level more now is Well, paradigm, investing, new towards think be and then attention progress like and the and some the there Thank the I products reached has product ecology. terms be be data mentioned point. user And paid certain in year. ecology the significant made earlier, to there the will it everyone will so will you. issue special is we a unless because next

of a paradigm technical readily learning of is [Interpreted] reinforcement large training. technical papers of capability new of the under and new academic the paradigm is for We to question that model model available ask about How large models reinforcement lack characterized the Chinese like enterprises open-source by direct the in I'd reference. learning training? understand

has existed Chinese language that the enterprises of time, mainly I a might are [Interpreted] be right? under everyone discovered learning also technical bad. needed of in the models, right? of the reinforcement of launch not think paradigm capabilities Reinforcement I long learning, Firstly, think OpX, learning because reinforcement for

or the After X major well played Chinese indicators. AlphaGo in also Go. X some did evaluation specific just look points, points past. as You companies playing in or I maybe a summarized, can difference there just Go, teams Basically, at some are few

including final affect previous the I and speech think recognition and visuals. really productization our doesn't this implementation,

using in robots more segments From the markets? would perspective, is hotels. scenarios? these service the places We more market can what in of can to see hotel there has in I Cheetah merchants and capture much China market share next are and current and robots. X restaurant the market future? Cheetah's much common overall space X these growth years? share how How is in the like penetrated [Interpreted] robot The much How most X ask, the the in restaurants

ranking they because share, the industry market all regarding reports of few Well, and such, there are sketchy. [Interpreted] this are in relatively rather

restaurants reception services. we late we initially relatively hotels, course, Of focused on for because and started

of the be robots approximately a share early in currently top among the penetration restaurants is few. overall be still the I such the perspective, think should we situation. From in It should market stage. and hotels

Europe This shows might think for a very, the is stage foreign my time, early hotels, the bit earlier. not very with is I think many customers, huge. growth. still in there in around because might it in aspects which many it as restaurants. It all in China, X% of even be market during more In are This still be in that space our interviews with right? countries, I is future

and declined. including to X about X scenarios. is This reduction right? their improvement. very there market in some are optimistic about market efficiency some robot know these imagine cost am I foreign the situation depend restaurants, years For where had itself, do not fact. X as power can the on is situations You markets extended And it has we ago. Today, term, China, the also in example, scenarios, solely in as purchasing some we long a But the all

What we just not delivery. do is

enhancing hotels. are believe and in greatly aspect voice it of voice, it expand in we the when the level, models using restaurants reaches will a large this we work capabilities certain Although

enter customer definitely also in of establishment needed is but this sales realization the it's make. And because of volume we can a or a in are it I this products based team, is why all confidence, effort we ability visit think don't management technical it in execution we aspects, and achieve and so. and about biggest can the well. scenarios in I year now, I customers of often own of both this my what sales the terms our the the overseas capabilities among in lot building satisfaction which channel, domestic top have team. on is and past our essentially but This system, we industry. XB Because many Then is sales the do

difficulty the channel real here technology entire sales for the us. fact, the In of More and products the is network. the establishment sales is and not

going of yesterday. global, this talked a Regarding to friend Chinese wave enterprises I

Cheetah I SAPP biggest global market some that construction do channel difference wave we the deep, local and this, of the Mobile to wave using must formulate last going corresponding we much able strategy. believe do be of the wave last is advertising the to and Mobile's will the I promote. go local can ranking Cheetah think deeper be from and understand than of If online moat our

required still market This X our large. the the effort still, this years. position is I But achieving for and view. top in also be confidence difficulty relatively have the will But within in

concluded. [Interpreted] presentation, of has time, Ladies and on our that by at live conclude now we We session. transcript now will were at call.] may the thank an you gentlemen, this present the and interpreter disconnect. [Portions this marked for do attending you spoken And question-and-answer are today's this time, conference