Baidu (BIDU) 28 Feb 242023 Q4 Earnings call transcript

Hello, and thank you for standing by for Baidu's Fourth Quarter and Fiscal Year 2023 Earnings Conference Call. [Operator Instructions] Today's conference is being recorded.

If you have any objections, you may disconnect at this time.

I would now like to turn the meeting over to your host for today's conference, Joe Lin, Baidu's Director of Investor Relations.

Hello, everyone, and welcome to Baidu's Fourth Quarter and Fiscal Year 2023 Earnings Conference Call. distributed our and on website was a can newswire release on copy as earnings earlier you today, find services. Baidu's well as

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

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

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

For undertake detailed SEC of risks obligation forward-looking any any with Exchange. as and report annual refer and uncertainties, Stock discussions to our under update latest to filed except applicable these law. other statements, Hong not documents required does and Kong please Baidu

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

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

As this conference reminder, recorded. being is a

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

with in fiscal our expanded year-over-year, solid business, for serves core We ERNIE concluded resilience our XX%. the by of foundation XX% for and as and strong ventures the core This which Baidu's XXXX to bot. revenue year the total from non-GAAP quarter. a margin X% increased ERNIE operating improvement highlighted

the economic the opine of benefit and in and models Gen and opportunities talking progressed of economy. activity bot about to shift from will process We that greatly revolution. Gen sector well to XXXX bring scale. is from marked ERNIE AI technology a renaissance foundation this and real reengineer at year ERNIE groundbreaking prepared about foundation Baidu models, massive AI a of a business every

X.X market as standing in model from ERNIE company a powerful most model call public ERNIE EBX foundation runner to by globally API probably as became in the GPT recognition As gain with evidenced AI, launch costs to companies. our the front Baidu China. the well-known first multiple continues

its to Notably, series. [indiscernible] Autohome API its leadership ERNIE Galaxy The IGC SX Samsung our efficiency and API API Furthermore, further the using app. ERNIE earnings. MAGIC Gen is uses power of ERNIE X.X in continued we solidify uses Owner on AI. improving XG partnerships

costs, users accessible inference EB increasingly only is the XXXX about example, March For the X.X inference X% cost and ERNIE become to of version. With of enterprises. will lower

are our unique We end-to-end because strong and optimization. that in able X-layer our to do AI of ability architecture primarily

our customer between Additionally, AI earnings to RMs Since This provide believe native the potential are We Recently, small we better incremental are and ad unleashing offering revolutionize of bot. model and and earned full business in in their business needs. review more ERNIE serving revenues models. the efficiency, and services, section. own various several we million from have we began to will we generate helping actively balance detailed XXXX, In technology applications experiences. essential smaller the and build real a products QX ERNIE quarter, to sizes. bot. ERNIE primarily creating utilized our performance enterprises improvement hundred I fourth RMB ERNIE from explanation can

revenue incremental to Cloud businesses. this RMB, and from believe we primarily AI billion advertising XXXX, into Looking were multiplied several

up applications ERNIE, an AI in foundation and ecosystem for foundation future the opens party technology. models, which While envision millions monetization us ERNIE we see groundbreaking enable third This revenue to to serving as by beginning are native this potential we and sources. will We commercialize create of enormous the around AI Baidu. developed Gen system various foundation as paradigm

in a platform Now an Equipped AI express has a create of Baidu helping co-pilot, range become for with the application. wide Wenku content native formats, AI key highlights their users in them Baidu's ideas. to

In them inspiration automatically Wenku's assisting features addition particularly for users by content, Baidu their and creating into turning expanding PowerPoint. in appreciate summarizing to

We have been consistently improving and enriching features. its AI

For mine organizing for example, a X we maps new information. such introduced as recently

are offer feed results, Baidu about using ERNIE experience. used results quarter, reconstructing at Search the how to scale and Last information introduced QX, one we useful relevant talked the to generate improving explore. ERNIE. we Baidu In Search We search alongside shop content such like new generated to users

understanding wider creation facilitating Gen expanding the we range search Overall, certain complements multi-round AI-enabled also content traditional of dive and to third need information into began TAM questions. of Baidu by queries related answering serving QX, In with our users' conversations deeper search, intention. Search

to BXB merchants actively encouraging Search. platform. lending with are to X,XXX pages directed and buyers often driving AI our about We purchasing users chatbots by At education, auto Currently, where transactions in new potential offerings AI AI help chatbot our particularly serve as time, interact can them are build about to their SMEs increase enable before merchants Baidu them making industries health chatbot in care, productivity. on better AI same to us, service the called merchants decisions. using legal, chatbot, travel, learn our

AI-enabled Gen reinforcing will on help search user increase our intention position for entry seeking as and services. primary Baidu information the believe users point the We

easily set reach to and meeting the cloud application to public tools the introduced prioritization, we such to comprehensive by native ERNIE view online AI-enabled capabilities up. for data ERNIE. We of are organizing enabling tools enterprises analysis. messages enterprises and incorporate December, our facilitating AI on summarizing on swiftly ground draft, as from to also Gen divest expanding In AI-enabling support Builder creating App notes applications of enable These enterprises

enterprises several promptly have validating to ideas. to as applications AI for their enable such agent templates Additionally, create included ready-made we

in pioneering Such experience progress part in on own large our AI applications. based native developing is

have the XX efficient the cloud to with and and the day. are That's enterprises with monthly API ]. models XXX% addition In for we In quarter-over-quarter is divestment, handling XXX% cost-effective a most tools on apps. for through about powerful December, XX,XXX million partnered enterprises most ERNIE us foundation most ERNIE more because ERNIE platform every app basis, And a third-party significant quarter-over-quarter. using costs [ queries increasing up rise in actively than now

AI As a cloud rise. native number to continue public hosted applications our on of

We gap enhancement widen feedback performance more AI further successful applications are and around ERNIE ecosystem our ERNIE it, poised peers. growth customers. and ERNIE steerings on between a the popular, incorporating of forward. multiyear likewise This more by our will model support will from robust and and to accelerate the as users our be Moreover, domestic native become

a we amidst conclusion, model. tremendous In opportunity find Gen and foundation ourselves AI in

we our remain core revenues meaningful, serve and development next we the resilient. become are while business started have to preparing ERNIE more wave the XXXX, revenues. Moreover, incremental ERNIE will of generating expect In AI bot, the With contribution to in ERNIE. we

will then potential millions of with vast there unlock on that majority believe monetization we expanding future, a AI ERNIE of them numerous developed ecosystem the The ERNIE customers. be applications for cloud will by In our Baidu.

highlights ecosystem me for the Core's exhibited care, and ad quarter, solid recap the driven ad services the marketing a others. by business revenue increased revenue hundred we by cash tech. online let RMB and several QX, revenue, Now X% improvements due In incremental flow. for verticals in performance in of in margin In travel, to million generated XXXX. key each year-over-year, Mobile across fourth quarter fourth Baidu health

capability We bidding and continue quarter. including to our systems use ERNIE in to enhance system, targeting the monetization

has for text-based system queries Starting technology able and our real-time targeting demonstrating search been both improvements January, monetization to ads ad from generate creation. significant in

QX. platform As about adopt in began industry our our marketing attracted advertiser today, marketing to An new has the new of platform in advertisers. XX,XXX medical aesthetics

further With and co-pilot conversations clarification natural the AI the explanation. through multi-round can its articulate for requirement advertiser comprehensively language and

The feed content. generating in experience campaigns assisted has optimized search building conversational and relevant by ad

using drive advertiser conversion budget an and cost of achieved also Our dynamically them The XX% rate has audiences helped while capabilities. reduction conversion. highly a platform for X% sales this enhanced targeted in and acquisition reach lead platform of increase in allocating after to

a Currently, initiatives their of improve marketing entire growth. these helped our opportunity advertiser for quarter. small base the this platform, portion leading future budget there ad a All platform new conversion to have our has means in an their increase adopted significant on is which advertisers in only

models We expect ad from in and and growth in Gen foundation AI XXXX continued beyond. revenue

we while promote customer brand introduced and its ] [indiscernible] have consumers. [ multi-round can enabled Facebook increased valuable capturing product last and understand This further of insights attention local enriched chatbot brand to has enhanced an QX. service customers, on for recognition customer a our significant gain The also technologies, AI building to insights leading brands chatbot formula top our customer more demand. more expect brand AI interactions chatbot AI has for Since enhance helped chatbot [indiscernible] and October, It future. also product ad company with enhance has AI infant also innovative in user from our its the brand. in have products, strategy. engagement We chatbot use and and we into customers' interest driving ad AI ad its of the one While chatbot brand we potential marketing bot has advertisers, ERNIE AI and and China [indiscernible] profound to portfolio. embrace adopted to

continuous users will previously advertising share both turn, of product to gaining and online engagement in assist our we're our believe in I will more and apps. native platform. more With AI come us Baidu mentioned, offerings, build enhancement using in we As ERNIE revenues. more This, to market user

AI for accelerated generate steady by a increasing cash in played along X.X% building Cloud demand AI from Looking and profits to XXXX, grew foundation The with ecosystem revenue Gen model profitability distributions year-over-year improve into fourth to in in our flow. continue and and RMB XX% represents inference. should continue significant QX. Cloud of to revenue increasing quarter. X.X Revenue growth, role billion model Mobile revenue from AI this the

We have cloud companies tech build to seen public to models. particular, a enterprises, of turning our in growing their number

group other about as AI such revenue Core, billion X.X QX. the business Intelligent by Mobile Baidu Additionally, groups, the and Driving cloud Group was RMB in Ecosystem generated the

revenue, total foundation AI cloud for internal contributing and million. around QX XX%. was foundation about internal the and basis, X.X AI in with combined Cloud Gen revenue XXX a the external QX, model RMB Within and model On billion RMB AI accounted Gen

Gen strengthening expanding in enterprises reshaping within AI have AI computing the a shift competitive is industry our driving computing. and dynamics evolution and many general cloud market. addressable considerations purpose AI LLM This to cloud for pivotal the lead our total the become industry from in

XX% infrastructure Both these of Last explained metrics a was training LLM valid. ERNIE. on XX% benchmarks. us, running we [indiscernible] choose our AI most X-layer unique gains in thanks to both years infrastructure model to Thanks infrastructure Enterprises our training efficiency cost-effective AI China. our of for training set is that have powerful experience GPU Moreover, time of LLMs. of consistent at position quarter, networking our for and to in drive the and AI and utilization industry

math, our offer that of models Within app tools apps. X set which and help builder are we develop effortlessly model enterprises build builder, and

Since thousands enterprises of facilitated XX,XXX has Builder App have creation built native AI As now, of apps. our about its on map. the inception, of models

As revenue map, forward. more and built we on are our applications more potential will going have greater being

Looking revenue in profit level. AI should and growth non-GAAP into a strong operating maintain generate XXXX, Cloud as

conditions intelligent in metric orders driverless January. were provided business during operating months. largest because complex Wuhan, increase in area breakeven operations our achieving few on Go's traffic peak further we to Go. XX% surpassed XX% QX. focus expanding Our operation, by past and new Apollo fully in areas In This is in our intensified The with the hours Apollo continued about vehicles driving for of

technology from about through Apollo safely In resulted improve rise the -- from marking operations. our China, collected XXX,XXX cumulative The to public provided Apollo the up increase. development early in by help operating Go In public million. in rights the substantial ongoing QX, This operations doles. offered safety January, Go us of year-over-year safe on efforts Go X enhance data exceeded will Apollo efficiency through further XX%

remain focused to expenses in to UE successful target in plan to on our managing regions. Apollo intelligent Go's swiftly Wuhan our driving. Looking closer reduce will and reaching costs UE breakeven, to and we into other operations getting replicate XXXX, losses we breakeven Upon

AI Gen and summary, are models. tremendous facing foundation In in we opportunities

invest strive optimize cost to these line business time, at to will expense will operational and in same each improve we We opportunities. efficiency. to continue the And structure our for

to With Rong the that, results. financial the let over call through go me to turn

you, Thank Robin.

year-over-year. non-online year-over-year. X% Baidu and XXXX Core results. closed results. year billion, to Core in increasing in me revenue, XXXX revenues X.X Baidu the up Total fourth for Baidu year with Now X% the were up was year-over-year. XXX.X financial XXXX, increased Total XX.X XXXX X% quarter In generated of billion let year-over-year. Core's QX, fourth revenue quarter RMB billion billion, our the QX RMB was details walk in financial Baidu through year-over-year for was Core's online revenue Core's billion, year-over-year. We RMB revenue X% solid revenues XX increasing in marketing was online X% increasing total Core's of XXX.X XX.X Baidu Baidu marketing full X% accounting increasing revenue X% billion, year-over-year loan. RMB XXXX. marketing RMB X% RMB full

QX, offset partially revenue in a online XXXX, year-over-year. IT increase costs of mainly was marketing RMB the related in year-over-year. due was business. year-over-year, by business. the increasing full in increasing year-over-year. costs X% to The our partially increasing XXXX, costs, X% in was and acquisition to track in in business QX, For billion primarily Cloud the billion business in quantum year increased XXXX, in XX.X billion driven RMB X.X content due billion was related RMB decrease in from to of Cost Revenue an year-over-year our Cost by a business, IT to online cloud revenue AI primarily from XX RMB an increase Revenue X% offset marketing by increasing increase was XX% Cloud X% XX.X AI the costs. decrease causes AI

custody Operating to [indiscernible] marketing server RMB AI marketing which XX.X primarily RMB billion spending support channel XX.X to and expenses an in depreciation year-over-year, and X% support QX, research the fees, due promotional XXXX, expenses. which research were channel and primarily expenses in spending and expenses server and sever and and an improves. in expenses increase depreciation increasing server promotional in were Operating improves AI development year-over-year, increase development due X% increasing billion and expenses an

income X.X operating was billion, X-X, XXXX. income X.X Baidu Core was margin Baidu XX.X income was Core XX.X Non-GAAP billion non-GAAP XXXX. operating in RMB RMB Baidu and RMB income margin was operating in Operating billion, operating and RMB income Core's was was was Operating QX. Baidu and XX% was X.X QX. Baidu Baidu in XX%, QX. operating X.X operating in QX. XX%, Core's RMB Non-GAAP X-X, billion, income was Core's billion margin operating Core's RMB billion in in

peak year RMB income XX.X RMB Core preferred other operating a other in XX.X same compared billion modification net as was RMB Non-GAAP billion due of Baidu was to equity Baidu XXXX. of and income operating method the underlying RMB XXXX. X.X was QX, margin Non-GAAP mainly tax loss from total income X.X losses billion total is of XX% Core a operating result XX was investments In for of billion last period in of million. non-GAAP benefit to a [indiscernible] shares. Income RMB net

RMB decrease impairment gain year loss and was In loss year last of total X.X billion million X.X XXX value investments. other last fair billion net total XXXX, of year to investments from billion long-term net of long-term RMB to billion our of to a a X.X X.X a compared other due value RMB mainly income RMB fair this of RMB compared in

and income Income X.X X.X non-GAAP Baidu Baidu RMB was net to per In X%. attributable ]. per income net was ADS Non-GAAP Baidu net to billion, margin to was for diluted earnings [ was was billion, Core income ADS RMB net income and was Net Non-GAAP attributable Core XX.XX. tax RMB X.X billion, RMB was Core was Baidu RMB X.X Baidu RMB was attributable non-GAAP QX, to RMB earnings billion. diluted attributable and billion, XX%. X.X margin Core Baidu for net X.X expenses

income RMB was RMB Non-GAAP and attributable RMB Core XX%, for margin non-GAAP ADS was billion, attributable X-X. billion and billion, billion, XX.X Baidu margin for was Baidu In Baidu was diluted to ADS and attributable was was earnings to XX.XX. XX.X XXXX, per net were RMB earnings RMB Baidu diluted RMB XX.X XX.XX. Core Non-GAAP Net per net to XX%. was income Baidu net income to net Baidu net Core Core attributable income XX.X non-GAAP

flow RMB cash were and XXX cash, XXXX, RMB XXX.X restricted cash, was and billion and XX, excluding short-term short-term billion restricted excluding billion, December and XX.X Free in investments, XXXX. XX.X RMB cash cash IT, cash cash IT, equivalents, flow, of was free As billion. cash equivalents, were RMB investments

With Core of call let's to XXXX. December had XX, Baidu now approximately as that, the employees questions. XX,XXX open operator,

first from Our Instructions] will Yap Alicia of today question [Operator come Citi.

on what My of percentage management is as How will of revenue XXXX? for the the Baidu in view think outlook does whole? on question the outlook. revenues landscape total about AI-related is China XXXX in also, What's growth management And Baidu a contributed from macroeconomic be the XXXX? for

this Robin. Alicia, is

very our last but revenue outlook, growth. We We in the into expanded take performance. our a Gen challenging very invested back margin a year. quick solid operating but solid demonstrated let's aggressive year-over-year experienced macroeconomic dive we non-GAAP had Before and environment, very business look AI,

importantly, foundation More have this from putting central noticed for economy. Gen to and local to began that revenue generate grow we are incremental AI models. and And the both governments efforts year, the we

during X-day lot a travel environment And in growth Year holiday, macro particularly we're saw a New with uncertainty. sector. consumption, operating of in in Chinese we the the still a But

We stimulus said, lot But are of closely think monitoring opportunities. defacing which is year's achieving significant goals. for being economic this a that essential by plan, we

remains to contribute foundation and RMB from models Gen increase incremental Our AI and will several the revenue. the will in core of to business revenue growth billion XXXX. our solid, This total

LLM keep thanks base have our user our and and we situation AI products More apps, to and [indiscernible] enhancing innovation. are a position through the AI, our we huge ecosystem, Cloud. building Baidu mobile developing And our Gen and increasingly for capabilities renovating in enterprises specifically, leading accumulated models already

long-term will growth. will we China's think be be GDP than you. which I Mobile, and Cloud sustain combine able our think faster Thank we we when growth So to

your come question JPMorgan. open. Alex, Are is of from Yao And line today Okay. you will muted? Alex Hi, our next

Yes.

Sorry, a I I questions on couple was trends. structure cost margin the muted. of have and

the much contribution? look how how expand? intend How optimization, past, be into you and should room at should we look trend you cheap revenue all, XXXX, investments between lag of more AI-related cutting to First we guys see in will margin AI that and In if there at a the discussed do cost investments. we the

questions. you so Julius. for much your thank Alex, is This

businesses. sure I expenses AI We still it's the room. cost pretty in and the our can is there businesses, the Gen alongside investments think legacy our that manage in

continue be reduce look we will to resource also the the and As [indiscernible] to core our allocations nonstrategic will we XXXX, we into businesses, our focus to on businesses.

continuously layers, are the overall we organization organization and Additionally, by enhancing flattening the removing our efficiencies the simplifying structures. executions

keep So this more of aim a we with we to year, our all generated system team. operations, committed Baidu its ongoing heading business into mobile Cloud productive continue profitability. to [indiscernible] are cash deliver AI and ] very ensuring measures have continues our solid strong while to very and with in Core's the HR flows [ we margins earnings plan, the optimization And these

We training a margins to have generative language models. related servers to purchasing our modern primarily mainly year AI remember chips This from in still if and managed and large XXXX profit furnaces. we our you investments the and and reflected in maintain and in invest for to started despite solid on, AI, operation [indiscernible] investment is the

several profit a on a the margins CapEx non-GAAP operating year-over-year know, percentage despite be year-over-year depreciated so you increase X in As still Baidu's saw cost XXXX, basis. years, our XX% growth in CapEx will a over

is AI new Looking inevitable that make will we investments. the developing it of business, new process forward in our

significantly for also margins long are yield expected However, to the our market in as investments not development, these the better we that margins. AI of will stage Cloud business overly profits. early much the And And run, this is margins now prioritize our during we believe affect expected business. or to

there for AI closely monitor promotional but native online we additionally, to may activities And growth. be carefully the balance XC investments products, will and some and the manage the

happy hear in in initiatives the to results. early are We that have our efforts to new investing see begun

from reached mentioned total to Gen AI has AI revenue revenue million incremental AI and several models QX the X.X% of generated As in have Cloud now, tax the Cloud and just hundred the the contributed generated financial improvement also Robin from RMB revenue. incremental

and think of promising confidence our achievements I line these all strategy. have top our strengthened in business

-- Thank development and commitment we So remain in going of use. language modal our will the Alex. you, development steadfast the large to to Gen AI forward,

Our Gary Stanley. next from of will come question Yu today Morgan

contribution. I have a question regarding AI

prove business. growth generating we search core ] is rate or revenue incremental, way AI-related for quantify contribution And is be from the And is AITC faster for So revenues ad a Baidu about there [ existing growth. we can excluding expecting how be you your cannibalizing contribution, average that purely will AI, incremental think XXXX? purely than should not the AI from if should and

this Robin. Gary, is

We are in largest search engine China. the

presence XXX the on to established robust a users information. close comprehensive rely have Internet for users. reliable They and we We million and us among active mobile and monthly Chinese have brand very

and of we So also are we coverage has macro strong revenue But have business and because a very stable broad profits. base sensitive advertising different our verticals. to very

AI and the I the mentioned But macro are front at Gen engagement earlier and environment. with still new opportunities unlocking both front. uncertainties are that monetization LLM the user there

quantify several earned upgraded presenting this and think AI initiative. And to easier Gen QX kind it's generating monetization allowed revenue has ECPM. has front. the I thereby I system earlier capabilities, increase millions revenue will incremental incremental us targeting more to grow monetization a on advertising year. that helped billion relevant mentioned this from of And We ads. several our hundred improve the in already the to

quantify to user helping the part, harder us engagement Gen AI experience. is It's user improve

us us pool, larger seen will in one help time We forward. we mind will initiatives even the time user initial features These share continue and outcomes going new spent bring and have introducing improve potential. and over search and

incremental side will both you. engagement from come So and purely user the Thank the side. I think revenue

Goldman from Our Kong Sachs. Lincoln come next question today of will

at product driven at is question revenue How exactly we drivers growth And AI looking then that should overall mix what as this how revenue Cloud year regarding And Cloud well. AI, year Generative what the business. is the should look and incremental will here. the that's the when Cloud? AI my for So growth expect offerings trend we're growth be margin we the XXXX, are the this by primary so Generative that

Thank for is This question. Julius. you the

Robin Gen in the related already revenue the total to from revenue internal businesses, RMB several just both foundation RMB the AI reached and As full mentioned, from million QX, this number XXXX. our external XXX model should the and including year -- for and grow billion

We features. new using have models increasing to AI To and and so in seen their interest enterprises applications from this Gen goal, achieve LLM building and to develop solutions. they power products are actively

external So the is generate customers. were we from how Meanwhile, revenue majority a seeing of increase this in revenue from the model external customers. we significant inferencing

inferencing term, the revenue We over is small. significant still the So will become and believe revenue a that sustainable it driver. from long

the generated is revenue quite AI is all businesses so to use important Baidu a for internal also is We significant of think switch reconstruct and Gen customers LLM products. by model first revenue inferencing, and company because portion the to

continue significant increase of features ILM number a the with to increasing rapidly reached customers powered So and AI have API costs and products have magnitude. been internal ERNIE Gen by and from

and drive revenue As ERNIE proven use also their to growing that apps efficiency in number ERNIE external has customers applications. enhance external of well our such productivity we the ERNIE believe the and bot a future. And and in real-world development can own develop will

inferencing and So your offerings, product question we have training our in powerful the model China. for regarding AI about most infrastructure

and helps build infrastructure customers effectively. this run cost our models So

of months app various in a for builder model models set and mentioned builder our full application tools offers terms Robin of and and building Additionally, development. as earlier, model

developed So generative intelligence. native our a in is such own AI we to business addition as that, have GBI, solutions, that

applications and enterprises productivity those helping increased So efficiency. are

and bring opportunity incremental foundation opportunities to Gen our the businesses. to related new to legacy also addition AI In models AI, cloud

So and tech are win recognized offerings, AI and to highly education and continue projects cloud in being because centric CPU-centric and the for customers first, especially sectors. we super the cloud infrastructure our for Internet we

and before, Gen transformations allowed foundation for efficiently for us Secondly, industries. have digital more facilitating AI our intelligent than and traditional AI to customers models solutions build

both driving see growth are XXXX So year. revenue overall, in cloud over for of X factors revenue. growth our should these last we And cloud accelerate

AI Additionally, maintaining profitability we are for in pretty our confident Cloud.

gross the consistently For Enterprise able to Cloud, should margins for cloud improve businesses. we be [indiscernible]

stage As AI the LLM market for at and still is very of early development. a Gen businesses,

a enterprise the we pretty market dynamic strategy quickly customers. hold pricing expand penetration more and our So should to educate into

So should traditional have new we a margins businesses. Thank cloud the normalized higher over term, than business long believe the you. the

question next from Chong come Thomas of Our will today Jefferies.

I time about ask benefits for on search how And super should traffic the May pace And AI of XC generative product traffic exposure shared see which the been? AI app? growth the metrics or How search? be our key new has any spent? can when development we the traffic

Yes, products we are reconstructing all generative of AI. XC our using

making powerful. Gen models are all think products I of and AI our foundation more

user been encouraging. Over the feedback this past few the initial months, we made significant strides in kind of and efforts has

of enabled Baidu broader including introduction and AI to the answer search, complex, range queries. has questions, a open-ended more of For comparative Gen

we provide clear bot, and interactive more ERNIE way. direct can With a in answers

search During providing links, by ERNIE bot. the just some and some lending content past months, few results instead more of were and generated more

new questions. are result, asking and more a users Baidu As frequently with engaging

and content more be images. or creation, are Baidu to text for it more example, coming For users

new New for helped Chinese of but greeting users the During generate year of users is large for a kind not Year this their Baidu typical Baidu on search case personalized holiday, messages rely need. we use and for number e-card a loved This see create engine, ones.

intent, that then complicated multiround user addressed Going answers user search use clarify to language. through to needs so forward, bot we natural use generic and for increasingly conversation be ERNIE queries can will

of While bot to these initiatives search. have using early resulted reconstruct Baidu in an enhanced the search still experience, ERNIE we in stages are

on based will for larger iterate AI ready and features enabled a We upgrades And before Gen for rollout. will new test experiences follow testing continue playbook fine-tuning our scale to we and typical are user feedback.

complement of traditional time and range on for an to a create and for spent bot will believe users to one-stop has Ultimately, attracted And transformed addition content wide And Baidu. we were Gen search. Wenku. users tablets to a from features find engagement in a copilot retention user application new for think Gen format. by ERNIE about Overall, shop AI to of as increasing XX% Wenku AI search, to users documents year-to-date, I paying in acting

that same through helping in third-party active at ERNIE the build believe And We emphasize early we new learning. using of reconstruct the on to want success I app apps attracting our enterprises build through apps or time, its are on Baidu apps. are and and wide [indiscernible] and stages the of whether adoption, depend ones. we to

will come today from Lee Our next question James Mizuho. of

for ERNIE's Does similar map technology sort can or or Can feature, AI maybe or up milestone talk include road launching metrics you maybe to guys it any to? AI key Is engine. you multimodal XXXX. store an of speak about there maybe potentially opening [indiscernible]

that Are managing part second how on the we you about puts process? AI, should ways any question to is cost more and Gen talked about there levers think Obviously, take maybe some running going inferencing on the And optimize of efficient. to this make it costs forward? additional to

hand the next us advance we chips level. to have the EBX on should enable Yes, to

adjust for tell increasing ERNIE reliability we and we application-driven could multi-modalities, users and earlier, on. will and our that models. be mentioned building [ ] us I our where approach so an to take As intense should It and improve customers agents

foundation to bring ranking It's important inference model note mentioned operations. that have right, ERNIE is large-scale rather earlier price we [indiscernible] we that high-performance publication. is making high are important I And you're than very focusing cost. to continuously value affordable users for been model customers research issue, using achieving a on critical to simply and reducing real

interest of of of the about March Now X% XXXX version. EBX.X is cost

and their customers, to We on efficiency for many more they develop speed. By cost ERNIE. increasingly iterate that enterprises with willing doing test, and balance understand tend to more and so, applications are

a customers smaller also for language and offer leverage performance. help our of our we model experts So MOE that's best mixture

end-to-end developed increasingly there room our [indiscernible] And customers. still we are -- of the to make most adoption With approach, help Internally, number we believe our apps further on monitoring the developed the this will ERNIE. and of models closely cost by affordable drive of our is to reduce ERNIE. ample them

now, from right queries app handling now ERNIE in from cost. API But been million than costs apps I per sizes over ERNIE And as mentioned XX cost have internal As ERNIE application for different external external day. larger still will been it at as of the Baidu it just and end We're this useful have -- more journey, more become more or powerful, use users apps and third-party through ERNIE apps. increasingly beginning through increasing only more be rapidly. smarter previously,

This to us ecosystem will enable around cultivate an ERNIE.

this end Thank will significant they users, and actively by models inferencing As you. revenue us. app generate used for are also

Charlene Our next HSBC. today come Liu will question from of

understand versus may drivers help you we API can how ERNIE has will also and apply underlying applications us Lastly, very enterprises be? that gone earnings help with evening, Good compared to assume the the question again the unlikely using integration build us opportunity quarter other and last to and you thank How are enterprise us to I using that adoption related have number peers? the ERNIE. models kindly you. please Thank what who does elements you and its for of management, whether ERNIE share Can a latest enterprises can understand be that

question, Great Charlene.

China gained million have else one and no cloud XX as XX,XXX different volume So prices spreading costs marching we right, our our believe daily verticals, December increase API a ERNIE has sizes, API a just about ERNIE different and in basis, mentioned, over received API XXX% from caused customers high of such exceeded so quarter-over-quarter. in and The requests. Robin many on a of

for reasons. choosers enterprises The following primarily the

we cleaning primarily Firstly, cost-efficient infrastructure of for strong optimization. the modal China, of ability have most and inferencing our AI because in end-to-end

the large AI for and language enhancing public landscape and Gen models cloud reshaping inventory competitive before, advantage. competitive mentioned Just our China's

in ability utilization very cloud extensive strong high infrastructure. AI has our continuously GPU enhanced with So reach our managing GPU-centric

build So at and native enterprises AI our develop on low their help run cloud. models apps a and cost can we as result,

have attracted Secondly, family models EB the customers cloud. many of to our

Over past customers. the have ERNIE consistently few we from the performance, positive months, receiving feedback enhanced

We also to structures. accommodated models needs sizes different cost in customers' better offer ERNIE regarding

AI China native we one-stop development. is the were to a a application as service model first a LLM and company Thirdly, in shop offering, launch for which

most or easy applications models also enterprises makes it for fine-tune [indiscernible] LLMs. help to easily So use our providing trim and on enterprises cloud. develop toolkit to their We're

product well. app different So data, models can power own can the to as by use with development. them proprietary MOE applications support can their help directly approach incorporating to ERNIE cost different the We build and in purpose the effectively features API to also toolkits, their they using customers adopting

on the result, a ]. as efforts focus rather enterprises pinpoints than on their So program customers' can identifying expanding [

helped a initiatives these establish and LLMs. of Gen all us have So AI first-mover advantage in

and user your accumulated AI For customer aimed mass ] platform last at on our generated and to attracting use cloud. native be question, as users, applications more will insights [ customers substantial develop our

will these further the toolkits. us to also So insights refine allow

and become will tools They tune apps. to friendly and with effortlessly stay fine create our inclined As increasingly us. help more models be customer enterprises

crucial models. noting create is that to current for is [indiscernible] it their worth for customers language chosen large of Additionally, the at models, it employing pumps stage

had it since because them using another best So switch to portfolio. a have becomes will effort considerable -- building intelligent to their props for and large models. to prompts in And prompt accumulating they their they for model to invest reestablish language

result customer switching and as us help our costs usage increase a So the with increasing and will platform, of active customer adoption satisfaction retention.

next Xiaomeng from of America today Securities. come will Our of Zhuang Bank question

About impact the U.S. on after the on Is from recent alternative chip, wondering there AI development the chip what's update restriction chips. your the further any AI

the what the developing product concern, is in chip what the next what how with be monetization do years? peers, the AI Given few the up and versus peers difficult? model and keep to overseas become can differently company the Baidu achieved And may will overseas

development In monetization. the is model the minimal near for product impact inventions our or or term,

powerful X already less enables enhancing or us most next we have chip our quarter, earnings the chips. for for AI powerful last model X continue mentioned the to foundation reserve China inference, years. I it model As in requires And and

users chips are native the end Our many to available sufficient for reserve and and the application for market on us AI power customers.

And compromised. model the for in and the the the layer layer, innovation homegrown not to But will application the most room the with ample long user access experience we cutting-edge stack. software the Net-net, There's efficient not in be run, have GPU. may most framework layer.

AI with building applications advantage the support provides to using and over competitive and chips along Our architecture inferencing. And end-to-end with peers. domestic efficient advanced self-developed for AI. us way on best team our R&D This training our developers, embrace ERNIE, unique most a in Baidu enterprises strong will model efficient and will be X-layer for less

from today Fong question will come Ken of next Our UBS.

is on of [indiscernible] Kenneth. This behalf

question is, my have generation video video days, or technology. text recent So we seen numerous in developments to

do you in map it elaborate hold impacting forward? strategic road on for what this ERNIE. moving how So the ERNIE that for technology you broader AI implications and China Could your would industry envision development

does in and And video to these how currently ERNIE text perform you Furthermore, image, do task? text generation in what areas. text improvements to generation foresee

Robin. is This

future of an multi all, and integration area, as to important AGI. And for a is and or It this already continue development. audio the model First multi-modalities video to text of future. in will do the Baidu invested is must-have for model foundation such in direction so has

type enough the this scenarios. of is of make room we if are most and look accessible evolving large to Even still models, for offering for very for lot various real the more world of not stages. world of early models innovation. Secondly, the MOE problems are good to at foundation in applications. still huge agents There at are enterprises and powerful very Smaller-sized solve quickly. models, We in the market development all all language strive models plenty and language a

Baidu vast a a foundation autonomous in and model. pioneer thirdly, And the significant application is market driving, realm with leader. notably, One, of global potential is in visual which

generation transformer We video purposes. to for have self-driving and been diffusion train our models using

world. in intelligent, tasks on specific enabled driving We segmentation, world into resulting has road strides and to translate consistently have and the better and us safe This physical the autonomous images rule adaptable classification, videos and understanding made more thereby detection the object also technology. physical of the captured in

is real-world problems, most to Overall, you. this models invest we and to position. solve divest foundation ensure our leadership area to to powerful our continue strategy the Thank in

Conference ladies and XXXX attending do the Baidu's Fiscal will time, we conclude today's presentation. question-and-answer conclude you. for Thank thank And session and gentlemen, Earnings Fourth you this Quarter We Year at and Call.

You lines. may disconnect your now