Baidu (BIDU) 21 Nov 232023 Q3 Earnings call transcript

Hello, and thank you for standing by for Baidu's Third Quarter 2023 Earnings Conference Call. At this time, all participants are in a listen-only mode. After management's prepared remarks, there will be a question-and-answer session. Today's conference is being recorded.

If you have any objections, you may disconnect at this time. to conference. meeting like host to now today's for would turn over the your I Baidu's of Lin, Juan Director Investor Relations.

today, release Hello, Third everyone, website distributed as earnings services. on on Quarter Baidu's can earlier was XXXX Earnings well a and Conference Call. our welcome copy and you to newswire find as Baidu's

in Rong CFO; and Dou the we Robin have Group our On Baidu Li, our Cloud our Luo, EVP charge call today, Co-Founder of AI and CEO; Shen, ACG.

will we the Act remarks, U.S. prepared a contain Q&A statements that our After Litigation the of the note Please hold provision discussion forward-looking Reform session. Credit Securities of will today XXXX. safe made under harbor

current Forward-looking risks cause uncertainties statements results may subject that and to expectations. from to are our actual differ materially

risks please latest uncertainties, discussions For Hong refer documents with these detailed report annual Stock SEC of our Kong filed and other and to Exchange. and

Baidu does as obligation auditable except any under law. update statements, forward-looking to any not undertake required

of call Our earnings press financial discussions this release include certain non-GAAP and measures. unaudited

Our IR a comparable on unaudited available the GAAP measures non-GAAP most our unaudited press measures reconciliation at directly ir.baidu.com. contains website to of

a conference reminder, recorded. being As is this

the CEO, IR In of available over Baidu's call now this website. Robin. our call turn I will addition, on to will be conference webcast a

us. industry. that, . Presently, of of ERNIE opened on key some ERNIE an I I that open delivered scale of midst the to opportunities handled revolutionize ERNIE in API. is Moreover, valuable product team deploy will navigating margins model's by our stable business and bot enterprise we of refine revenue, and products. large our we cloud to efficiency would received Hello, World each users that the enabling models approval in everyone. AI API ERNIE After foundation in we've then, to and AI and for generative reinvention for every discuss Since us set and On driven a we've performance. that portfolio Baidu the like ERNIE bot showcased by flow and our further August update Today, operational to shift despite to have ERNIE and cash business. a bot. in managed bot marketing up online customers. solid customers ERNIE a witnessed our I'm increase the and a are native broad-based profit full-scale share with climate in XX, QX from At and our both our on and challenging significant bot queries we feedback October, macro new strengthen AI in progress business. ERNIE Baidu maintain received ERNIE highlights proud to platform through

EBX, that displaying GPT-X is or content crucial level in event, model, we advanced understanding that These are most performance for During level X.X native our capabilities believe AI the and and introduced EBX solutions. complex We generation, applications reasoning developing model. foundation retention. ERNIE human memory

input unique In and AI EBX We I and portfolio and our users an earlier helped intense The infrastructure, Model in X our to efficiencies It which end-to-end feedback our launch resulted from are pleased layer training. expectations. from past, as to big experience. said our introduce continue played AI-native the reinvent entire ERNIE customers role. we also to use a products, in than

tests that to multi-round further features results search. direct we bot conversations to initiated answers on the generating new our recommend In On new information Feed-like enable with the Baidu by quarter, the queries, expression. and customer encourage ERNIE user the front, search generated traditional search together enables search complementing

unleasing and monetization Our deepened opportunities. new prolonged time initial promising We have help tests features received these feedback. engagement will spent believe user

native reengagement. legal and care, product Ernie health serves customer willing are to new a advertisers in like verticals and as benefit auto, and invest multi-brand desktop AI both conversational heavily which AI on they assistant particular, as education, acquisition mobile. In travel, bot, versatile in our

confident services. we the large model, language monetizing in our our of exceptional performance Given are

Starting fee marks from implement market. as in was public about November EBX per charges. models X, subscription of from in user at opened company This us Distinguish month. X the USD first bot to China other us through the ERNIE to the a

Our the primary co-pilots, to between era. believe collaboration which new focus trend is key encourage users we is seamless and AI a in

a For shop example, Baidu co-pilot creation needs. into with document has transformed various one-stop an [indiscernible] AI for

a in that coming increase the expect account, we We continue the paying have quarters. already in to trend seen an user

simplifies interaction across with the various The On analysis help analyze was intelligence we faster need front, of recognition industries language bot. AI introduction using for GBI, GBI data business the customers generated operation. the GBI decision-making to have more prompted them copilots ERNIE data facilitating for enterprise-facing product that natural a by recently introduced business efficiently.

collaboration productivity ERNIE we our opportunities earnings Baidu During enterprise and how customers. create last to bot of present call, cloud discussed platform. them focus boosting we products our gains and and These upsell AI also our each coding on communication assistant use and InfoFlo, our for efficiency [indiscernible],

cloud traditional number and the increasing industry interest this in showing of public sector China's version an our customers trial and products fact, in used and In features. have

array and industries. these products us Additionally, customers new to across allow also an wider acquire of features even

ERNIE enterprises drive and AI LLM. AI that development API In native to create empower of the own will through generative to of terms solutions applications ecosystem, and we their leverage

on us As likewise of more customers, AI ERNIE will successful. successful, our and native ERNIE more or whether by be by become developed applications top built

at regulatory August. are received enterprises a monthly green through since XX,XXX over on light Now basis. API been has of using ERNIE end number we This quickly the actively the growing

period lot November, has same in now, the of building large the products pioneers handling of products of Baidu every In queries a millions daily number rising of XX% number we this family Right is come our half effort from of have the over compared of tens day. October. reinventing increased into of put AI been and as native external and queries ERNIE a to Currently, have first in the offerings. queries by

As rise in credits cloud we our in the native continuous will applications, future. customers we actively external there creating AI in believe assisting and be a are significant

attracting services to possible actively ins more cases. of their with ins. people We're With bot ERNIE developers also and help a through plug tasks more ERNIE wide to information range can plug use connect unlocking and

As future, quarter, Justice and to of Autohome in progress and Mind. this users already resources during [indiscernible] realign and Tree pluggings New thus just today, efficiency improve have ctrip.com, from in been shift Oriental, will of ERNIE. Data lower to batch using invest made Press include accessible China growth Gen plug-ins Institute, summary, revolutionize significant is and third-party that In the we usage product believe priority and initial Group, for We AI and the to away The for CITIC business opportunity balancing transform this margins. customers. our In operations businesses, the investment beginning. efforts existing we Big

customers, possibilities excited for our users, partners for Baidu and are ecosystem. about entire We the the

the Now metrics the in by each Core's quarter. increased experienced distributed exhibit flow financial App's Baidu content let App strong feed increased Search to search a consistently of and generating both Baidu performance by online and by by steady ecosystem Baidu year-over-year and third marketing app queries million particular, group. for of quarter. in the profit resilient. remained September. me key recap and double-digit for the distributed Mobile to user revenue third year-over-year highlights growth our businesses. X% both cash XXX continued Baidu in videos X% growth in MAUs In quarter, within

language platform. Gen to This continuous the use from marketing their our others. Starting among more growth more quarter, was and them. enabling the and effective and care on conversations targeting continued can Moreover, our In campaign strategies new to made enhancements as focusing by formulate our multi-round driven health us platform. advertisers help monetization we advertisers This platform verticals advertisers supports ROI the improving system. help auction engage such natural articulate capabilities AI we ongoing in comprehensively, recovery increase input and system, on September, travel, to conversion requirements which for with to

conversion enhanced education, capabilities. after platform XX.X% company and IT education XX.X% and example, increase boost achieved professional an of an ROI For this using Tarena in in rate

still stages Gen achieve to We and advertisers our ROIs of on AI in the higher conversions platform. are early using help

bot, revenue munifization to future merchants quarter, Our on to our Baidu. the capabilities sell help With stream Baidu. ERNIE improvements contribute to that SMEs shopping. digital used allow introduced and efforts the cost own live more AI will we generate One in sales we also ultimately their lead a them for help significantly easily to to with scripts grow capabilities live In growth. on human, we live streamers example barrier significant and how merchandise tool create is their and lowering

test than GDP same AI continue marketing growth. potentially gives the growth of our time, [indiscernible]. open traditional we will marketing marketing continue This we more in to are long-term exceed opportunities that that prospects. forward, to products revenue general the growth Baidu's up At China's could Looking optimistic native will us confidence online online

AI to cloud. Turning

We continue on remained focused the operating in on quarter to as basis of positive we a generate business. the non-GAAP our healthiness profit

of cloud AI Gen API increased have strengthened advantages brought competitive their and ERNIE lot native develop us AI to solutions. and enterprises number applications a growing opportunities, which have and own A LLM in our [indiscernible]. of using

GPU number are complicated. and is a helping impact can failures [indiscernible] our unique working and our the their using process. leveraging building of simultaneously. models. also AI efficiently experience GPUs foundation very of by requires We ERNIE X-layer build in customers infrastructure large models years entire own training It

developed in ways a have failures reduction and quickly GPU leading to training We to costs. significant identify address

on of XX% platform industry about our benchmark. Now the training valid, is setting time an

enterprise ERNIE A generative easily us because new have fine-tune cloud time, set a of customer using some our for models. our of number At existing API cloud customers also of cloud large has We resources, us increased spending have to of with AI including data base. their are their customers customers grow customers. toolkits, sets AI. customized different same helped the Gen

in AI the cloud AI-related fourth year, We year-over-year generative have in an in due to rebound in projects. we the the slow since businesses. to growth quarter, easier year-over-year X% mainly smart Also, demand believe weak revenues transportation will driven third of increasing cloud this the positive last quarter, base in by year. to by in AI comp should revenue momentum transportation started revenue QX smart down QX declined

to Moving driving. intelligent on

we years before which pivotal also unchanged, concentrating taxi the on the achieve autonomous a a providing remains population largest regional turning operational To operationally we is million. Wuhan resources unit services covering currently regions. couple [indiscernible] this it in breakeven globally, of target drive Our X.X strategically region for are economics global profitable. our area, to end, remains operation on our is and about largest believe of

orders of third overall within XX% That's up XX%. exceeded the Wuhan in the the in portion In fully portfolio quarter, from driverless order QX.

Wuhan Go Apollo in We pleased to involves Kinko airports [indiscernible] the the Apollo Go economics. provide are that first opportunity the to highlight the at continue excellent into August, extended transfer of company presenting to of also distances, to autonomous expand. services China. International in travel general longer Airport, late remains operation Central public an busiest unit particularly the one for In improvement future The reach China airport

these developments we All years. in and of regional the UE UE reach aim to couple of contributed breakeven to a improvement,

Apollo XXX,XXX In and million the the order of year-over-year. increase QX. Go surpassed rides by X.X has provided a XX% end QX, volume to public, cumulative a marking

executive VP Intelligent business program, since great take rotational been leader who has Ethics for has of of taken autonomous XXXX As lead I named CEO driving the and XXXX. pride debuted Corporate Assistant Chairman another and been within recently [indiscernible] as has in resale our a us [indiscernible] Committee. we the group. will Baidu. part of have Technology the as business position seeing [indiscernible] with responsible Dr. Driving Baidu, since

Rong's proceed with let's review. performance Now financial

Robin. you, Thank

increasing sober Baidu me billion, Core period compared unchanged an X billion, bot year-over-year. and let through was from our quarter XX.X marketing financial X% Cost Revenue billion, RMB depreciation billion, [indiscernible]. the year-over-year Core's billion, marketing XX.X billion, RMB server same Total RMB expenses increasing increase revenue year-over-year. was expenses, revenue due RMB walk results. last X% increasing year-over-year. Revenue which year-over-year. up were Core's RMB XX.X remained was which in year. from was revenue online X% the Baidu details X.X ERNIE Now XX.X IT billion, support of RMB promotional the spending, custody expenses XX.X Operating year-over-year. China to primarily third essentially RMB of online increasing revenue fee, X% was Baidu X% increasing X% was marketing to

Baidu quarter Baidu's in XX% cost the XX% expenses Core R&D which XX% year. period SG&A RMB Baidu same were in in last XX.X year-over-year. the billion, expenses the increasing expenses revenue Core RMB increasing Core Core the were unchanged maintained last year-over-year. SG&A R&D XX%, were revenue billion, operating billion, of Baidu Baidu's compared same X% to for from XX% quarter, period RMB X.X X.X of accounting accounting increasing year. year-over-year. for

loan long-term billion. other loss RMB net Operating was the fair operating same loss RMB to billion. operating year RMB was the and X.X in million for RMB XX%. RMB gain XXX a period operating margin RMB income X.X first RMB same income X.X in X.X Non-GAAP by from non-GAAP XX%. mainly income X.X of last RMB was RMB Baidu Core was versus year, to billion, X.X and Total investments; X.X last billion net change impairment and recognition investments billion was of of Baidu for was was period margin second, X.X decrease billion income due value billion, the billion a core income operating compared Baidu Baidu operating total other non-GAAP the Core

tax was Income Baidu for ADS earnings attributable profit net to due billion to billion, to attributable income XX% was ADS Non-GAAP per X billion XX.XX. an were Core Net Baidu billion, Core income attributable net non-GAAP earnings increasing Baidu per XX.XX. [indiscernible] XX%. billion, margin increase RMB X.X to to Baidu RMB diluted income and RMB RMB attributable was and tax. RMB Core Baidu net RMB margin X.X [indiscernible] net XX%. for was Baidu Core were before was expenses X.X and Net primarily and was was diluted X.X income year-over-year, RMB in

operator, billion, was cash IP, cash, investments employees call X and XX, billion, cast, that, September free excluding XXXX, shorting were restricted the open Core to cash equivalent, billion equivalents, as cash let's short-term and XXXX. cash cash X.X billion. XXX.X was and investments flow RMB With and XXX.X IP XX, was now XX,XXX RMB cash September questions. flow, of of had restricted Baidu RMB As approximately excluding RMB Free

comes question [Operator Citi. Instructions] Your first from with Alicia

Robin, Thank on question management My and is you. team. [indiscernible] advertising.

of the So the is AI seen how some it change ad then fourth revenue outlook. the so any Baidu growth? e-commerce revenue the growth other seems management contributed sector contribution? that advertising looking Internet than reasons can What you demand quarter, to any macro peers, the is will softer into And like slower ad elaborate picking up? tracking besides And have

platforms marketing relatively from for platforms Robin. our also e-commerce online QX, one weak. is Revenue the accounted In of weakness, this Alicia, total apart of is from macro top are e-commerce on revenue online marketing we our e-commerce revenue. building contributors, our Like from was revenue business about growth Revenue as to platform many our e-commerce improve strong own shopping tracking Baidu. continue very native we business. other the experience is Internet from companies, XX% native

through made would in the I we highlight like have business to ad our Gen strides AI.

the of restructuring We this reach efforts the kind are construction, initiatives investments. and which level basically platform, growth China's from continue And ad that looking And online the of creative These to to current this are optimistic are of forward, revenue of incremental -- millions we of the marketing bidding quarter, the RMB including growth. the in have QX overall our to ad will off. is GDP hundreds you. pay Thank started the revenue exceed year. targeting expected

question next Yao JPMorgan. comes with The from Alex

XQ, monetize the return despite And cloud we in revenue. cloud should in moderate decline the revenue -- rate revenue the projects, I regard have AI half to the to on will then questions I then cloud start further when the first of at Smart risk? mentioned territory will cloud believe any will meaningfully. of a that expect you revenue from to growth Robin few into to there, And AI importantly, guys still start projects QX. revenue City are are to as contribute positive there With more capability, more the accelerate XXXX.

Lastly, cloud any outlook on view revenue XXXX? . preliminary for growth

is for Doug. this Alex, question. you Thank your

our improving in So achieved we focusing non-GAAP the operating on the we quarters. health business as development. of been sustainable actually, result, And have for as few past before, have we a profits mentioned

and As revenue for a in due cloud the Robin QX. right slight to transportation experienced already intelligent demand the weak mentioned decline

business will AI while So cloud fourth small the quarter transportation, growth. And our cloud in of pretty a return the the revenue AI showed believe the road. to will trend put and rest solid continue excluding growth we down

for side, keep across more us Generative came to even is various more What said remodels. quarter, application model exciting we enhancement. more we brought training, new seeing and [indiscernible] by opportunities and customers development large sectors and solution Actually, AI last already that

Although from LLM very it's AI still small, revenue and fast. very is current growing the related business but Generative

proactively productivity enterprises for have seen gain. efficiency more and more these technologies adopting We new and

customers, sectors, gains to see the from [indiscernible] they Some efficiency working have through with those tech these the especially education of and theirs. started

on them their cloud spending As a services. some result, gradually increased have of our

and QX, the attract And long-term we our margin improvement. increase to to also to believe in but aim Thank not this you, leadership [indiscernible] our Alex. Looking cloud. lead their continuously to and existing large we to Baidu spending leverage AI growth, customers generative on should continuous new models customers encourage revenue only [indiscernible]

The Bank next from question comes [indiscernible] America. of with

Congratulations about My is question earnings. the on results.

rollout us can the bot And with among mentioned, the ERNIE charge And So after [indiscernible] biggest think various also driver. any on started contract revenue X? since user subscription you fee? feedback consumer do find ERNIE And for for management X one color become adopting the you is opportunities share last ERNIE how which and to the the a lastly, can on month? feedback and the --

reached devised has have gained out with the impressed me Many questions. capabilities. a in positive to for reputation since and both advanced EBX from receiving test And be in us. lead we're it's the release have the have first we let feedback in to to we've market? of generates to growing users we been customers users AI its leverage charge opportunities sets your peers. and our the EBX EBX your first regarding that see other mid-October, the answer are enterprises LMs company customers, [indiscernible], customers. introduce willing monetization EB coding. pay or our that abilities. and EBX X, and enterprises and LM and November in applications, responses from Baidu reasoning for other native LMs complex user and over either clearer capabilities. noted And EBX. for other in Comparing by widen GPT-X seen fees using We our in started who end of to [indiscernible] AI number us We're question And . From about market, to structured opportunity. more further and China. apart proud model understanding significant charging to we EBX a level users excels by

address If search complex the and questions and at projects. to users more conduct topics enables significant products, we in-depth various a The in ad see on the also were It unanswerable. our research look previously and traditional complements revamped own that search opportunities new It platform. can search. new personalized you

to us. enable multiround will users soon with We have conversation

will natural in As users language will it and create on with multi-round, in side, our be commercial the search talk to able potential too. more

allow transform on our our effective Baidu. will advertisers at capabilities We a product for ad type our And eventually capabilities their a will are to model from help better only also model. to us drive operate the -- time, brand. this and We CPS business with conversion, believe same not a CPC chatbot to help but experimenting SMEs

the revamp team durable AI we leveraging have growth already will advertiser achieving on to platform Baidu. effort ongoing positive ROI to results, generative assist shown ad Our and continue and

needs mentioned, empowering customers Doug of just AI, terms different our with Gen as In are customer now.

eventually will that. their put to especially LLM that GPU export to will emergent customers training break but when difficult, from very clear prefer the model, become train Some achieve It own is a scratch on trying restrictions still abilities.

advanced turn on the So we ERNIE to LLMs will market devising applications. like available for

advanced growth through used. should more more ERNIE be native As We create see to more widely customers applications by become applications revenue become able in continuous LLM model AI to using powered inferencing. and

will offerings Over also forward. a in each to and revenue our inferencing fine-tune suit Meanwhile, their we better, effective. our are term, needs faster help should source customers to because existing customized scenario model the more become models long major cost of

business bring Gen summary, and LLM will in So massive us opportunities. AI

made this a We progress good the so already have of far, promising is just commercialization future. beginning and you. Thank in

question Yu from The comes Morgan next Gary with Stanley.

latest rolled advertisers? revenue feedback all advertiser the Can upgrade, on do core boost to of and gets in you even out how about think more the management ad share AI-powered as XXXX system the adds to level the it

to Yes, found our from transformation appreciate help rapid we with thrilled their positive which Overall, the ad are improve and platform. our think more to I our our advertisers AI system advertisers. [indiscernible] be new they by happy are productive. of very Also feedback the our on efforts them help features, of them

system prepared I aims we've a past few and As into of reinvent mentioned lot the Gen the using effort to in AI our over app remarks, put our quarters.

that created Now significantly even generate to marketing a reduces we allows in creative advertiser-facing ad minutes. platform can And it intentions. platform. like materials ad rates deliver interacting humans. rounds Moreover, the These by creating advertisers better engage platform create Because higher . This in advertisers spend by time understand developing have to strategy. of upgraded managers materials. manager with advertising few to proved it our takes than multiple to to campaign our use with that campaigns. us the AI is also higher an able process to integrated And only their an language deliver natural need an materials strategies advertising copilot, now experienced conversation, hours Advertisers conversion ROI. to allows have have spend an

advertisers to As new platform. thousand of a migrated have today, we few of our already

X.X to compared very small fast. it million base, While is our growing this number certainly advertiser is relatively

in directly initiatives end and continue at advertisers AI our bidding capabilities. of not to back happened system system, the In Such observed ad the ad targeting ROI. improvements improve their they've addition, to may use we and them, perceive conversion so but

single-digit increase in QX. a using these advertisers on So conversions probably high capabilities in achieved average

efforts of advertising to their eventually more attract our of Baidu. budgets on should allocate advertisers All

related AI growing users and useful the decision already long this we This our typically and is And revenue particularly previously marketing in I verticals with said, and mentioned, As as online replacement I meaningful. platform upgraded ad research page as has experimenting before rapidly act became beginning, to are been is process could a in engage chatbot a purchasing. chatbot landing the for future. where just in

Imagine to and a With being instead users AI training page. a information, searching about and directed for a program web learn brand-specific product can details our quickly of bot chatbot, other stuff.

Although, the conversation I them key is could connected continue year earlier very engaging. fast. bring trend work revenue growing expected search stay helps direct is initiatives of are sales. hundreds very kind quarter, to Gen AI new certainly potential Thank to think further powered and this also and opportunities. of multiround them with and I more better well and advertisers, that mentioned the decision-making customers the these to QX, believe incremental The it current reach We that at in of should AI of leading from And and chatbot level RMB with us XXXX. for conversion of guide even millions points, strengthen this you.

question Wei next with UBS. from The Xiong comes

follow your on cloud up to want business. I

assess well color vendors as up Internet landscape more to do next with we try to especially the China how share management cloud intensify the the against with when competition development how as year competitive generative us. expect to with and and other also in companies advantage industry could competing cloud [indiscernible] telcos efforts about think peers? So market, competitive as the in model, make large we and Do the our catch the AI and among

question. Great

slowing As the is business you actually cloud traditional right, already noticed, down.

landscape siding business large cloud the which actually pricing. large of the with and people models in generative AI reshaping in market focus and right? industry. AI is rise generative competing like In language changing. now on commodity of a things cloud But past, The and the are models, While, on are the language the are the competitive and house, was

have coming help why experience advanced sophisticated to we productivity They of to successful still utilize AI product is because record cloud because new interest some in these and the technology, but track time enterprises and efficiency. because growing using to expanding our but new not services firm to AI among technologies There us edge. to customers stories increasing most increase the also have we is experimental our the Baidu to a high There's came were only problems. products solve in and technology and are we them to in stages, are seeing the belief overseas. technology seen rebuild they in competitive their have That the

mentioned, Robin style first as model. GPT-X EBX is So China's

feedback also had we garnered shared EBX. initial positive the We for

engaged our with [indiscernible] redevelop assisting currently, the ERNIE utilizing new technology So are ones. products team create to dialogues in their our clients, them existing and in their and

has say spending I already attract So helped customers ERNIE that can us new and additional IT from customers. existing

the bring players briefly So for our market. in up to compared the would to advantage points like here, I X

in us our one keep first to at other innovations to be and to with both is X-layer layer or which leaders able make the [indiscernible]. infrastructure, unique to be The adjustments compatible training AI efficiency model gives flexibility every driving

networks clusters just for is valid. for XX% said, Robin more second training GPU AI develop capability the the of one, large training. [indiscernible] time our And our to is infrastructure specifically, on

architecture and strengthen As a prices to tech continue result compelling our cleaning drive motor customers the will customers, competitive to for and Internet AI several their service. It in we gains. of advantage our our efficiency including to Furthermore, leverage leading companies unique investments the to increasing flexibility will in to market. us costs offer help our cloud, greatly further us defers and our edge giving in more our on reduce in

earnings our projects. like focus is with are on cooperating highlight different differentiate least ourselves telecom can to we AI and I Regarding mentioned. our operators, At that since the market we have noting many we competition as we worth would cooperate from capabilities, on segments actually particularly

allow in are cloud leader large set our particularly so from to us sum well recognized ERNIE AI up, and strong capabilities market AI will companies market, by AI in us To Internet our believe other which in and will market. share and generative language capabilities that end China, for As the become the gain and we strong models eventually apart our peers.

next Lincoln Sachs. with comes question Goldman The from Kong

So my question is also about ERNIE.

successful What X.X. of the tech strategy or we given competition the be So in to intensify model the the future? the our iteration industry, for foundation certify either in stabilize upgrade foresee in model future would ERNIE any So to leadership? do

the ahead in we hand the in LLMs our an take allow application-driven have Lincoln, approach. launch already We're chips next to to us lead To the take AI EBX. level, competition. will we of

bot AI in native the apps us We will let ERNIE capabilities. tell to what improve

only the limited are our drive internal cetera. innovation API cost et internal number ERNIE market in the The of a search, apps apps, right on there ERNIE and from the our products apps like right very existing that now, of direction. [indiscernible], of rebuilding native ads, Given majority AI restructuring

important basis, is have using are XX,XXX enterprises that too. equally ERNIE to propels ERNIE API use a we that build enterprises which And we are through their offerings. helping What is improvement, monthly seen costs on over ERNIE's

We also our the of continue to improve models. efficiency

March, of been XXx XX%, same current version compared to example, in [indiscernible] reduced amount hardware [indiscernible] a has in the version power. basically of resulting increase bot cost the ERNIE by For interest in for

do unique our using ability our this do end-to-end to able X-layer architecture to also are leveraging and We optimization.

the into consolidation of offer to advanced select talents a into translate further our industry own them. will strengthened data, us a for of cost to account stage. on soon create advantage, and From such foundation the the the has investments, perspective, will as gives AI compelling enterprises than Continued building there high and a We competitive industry stage flexibility and chips, will scarcity upfront the market. factors rather more high-performance to believe more taking model's it and only In more products presence. huge demand begin one Baidu reduction influence foundation more certainly of will ERNIE language AI be and models resources models spending like leverage this development, models. large few be in long-term their

in that will millions future. the native So apps number based we on the expect reach ERNIE of

with Lee The next question comes from James Mizuho.

over quarters. I maybe was ERNIE and accelerate the we in how AI potentially next cost especially the inference And P&L. items guys that think these launch and Baidu of it. investments OPM years, costs should few to over investment various we improvement cloud including you next extrapolate how and term, the longer if your quantify higher Core thinking revenue few given the related parts, using the we profitability. also in should to at X.X over continued shift, your as all Can multiple expect that then affects a And AI investments moving about

take me CapEx. as is questions. is AI of and This generative Currently, computing models let the part which for around investments your large engine recorded the primary is [indiscernible]. power, centered Jim,

widely have put I resources think training more also more by resources few [indiscernible]. past small And future, lot which of we the which in to [indiscernible] ERNIE ERNIE new a can as powered is quarters, used, native chips models. our in in the our applications, AI more

impact that the of note . all quite few for please is are this years. all AI-related depreciation over However, a [indiscernible] the hardware spread manageable because on

the through model developments. moreover, either fruit. see For our depreciations. be and engine that externally, to model IND we're internally expenses power, the beginning related the to and linked ERNIE happy which AI the And future usage or highly the are is training example, large by to investment are in inferencing to And the supporting bear courses, all funding in coating of used the models should recorded generative it we

and just businesses powered generated we're As from revenue from more been additional has Robin earnings receiving fast. has approval mentioned quite by regulatory since growing earlier, [indiscernible] the

close one healthy favorable see flow. our And large and time. we a margin, eye cash strong business we on are to to profitability business we expect Baidu we long engine large cloud expect solid. high various achieve to a continues the renovate cloud making and genetic our are in AI the ahead, also the traditional remain its continued new more AI In and While generation margin from are look its opportunity ensuring growth models ecosystem arising of AI earnings term. profitable using that engine And generative keeping sure and quite can mobile stay have QX, the businesses, models cost still

mature businesses. driving will our all, with by noncore our to businesses in long-term them pace. intelligent from invest And growth all we AI-related For continue opportunity businesses, a to concentrate reallocating will resources

you, beneficial All our these James. long-term Thank growth. of quite for were

Chong comes Jefferies. with question . next from The Thomas

of comment and you. the set U.S.? from My on results. about AI if chip of offerings Thank side. product management How experience Congratulations solid restriction export affect does development impact question the is Can chip the on a our the any. further the that on AI after use

Yes, the restrictions on limited impact term. on the have chip export to the China actually Baidu in near

chip users. EBX advanced AI earlier, year milestone other which chips, and the for a reserve have difficulties I inference lot Also AI model for believe successfully well in help China. in a pace just substantial most a keep powerful our . And sufficient the our impacts is we bot the for divestment said support of long end the two. acquiring launched in of of as us can And us. ERNIE advanced chips we AI alternatives native to next chips, in improving most having have requires less or run, as reserves We foundation in It will China. be apps the inevitably as mid-October,

proactively seeking alternatives. are we So

as best in advanced will unique help these strength us our these algorithm and efficiency not U.S., are X-layer as While challenges. options of architecture some mitigate continue AI the chips and to AI in improve the

bot, For AI model and our framework different chips, in [indiscernible], and be innovations foundation both example, training to parts with compatible we allow made of learning deep them our model have to some tasks. better inference ERNIE

positioned service we the But actually believe best companies we face given the that to are other challenge, all Chinese market. same

try know, do layer power We As model by infrastructure sell to to to start-ups. those our some layer write basically you peers, And the Gen to that. those AI optimize they foundation in startups the models. on investing into to computing didn't [indiscernible] to apps. everything and they We the from probably of in layer past, train try the

optimization need more They can just So we of don't do end-to-end on as time cheaper. computing which the market. we've AI develop that they and realize inferencing based models. effectively company can more think foundation in will more cost -- I apps on do approach. native the passed this by, invested model, And kind power, more to training efficiently, is train amount best to and their need Baidu's Therefore, foundation the same and can faster for we we of

the time we happy that front that So many, for invested on many basically this kind optimization I'm us investment years. to for it. of show really worth is And it's end-to-end

that conference does conclude and participating. Ladies for our gentlemen, today. Thank you for

may You all disconnect.