Shiin waxay si taxaddar leh u bilowday inay dib u furto marin u helka hawlgallada horumarsan ee AI ee Nvidia, iyadoo siinaysa ogolaansho koox gaar ah oo ka mid ah shirkadaha tiknoolajiyada ugu waaweyn. Shirkadahaan waxaa ka mid ah ByteDance, Alibaba, iyo Tencent, kuwaas oo la filayo inay wadajir u iibsadaan in ka badan 400, 000 oo ka mid ah H200 accelerators-ka Nvidia, sida uu sheegay Reuters. Shirkado kale oo tiknoolaji ah ayaa laga yaabaa inay raadsadaan ogolaansho mustaqbalka. [Related: Warar dheeraad ah oo Nvidia ah iyo aragtiyo] Talaabadan ay Beijing qaadday waxay muujineysaa isbeddel iyadoo isku dayeysa inay isku dheelitirto baahida degdega ah ee xisaabinta AI-ga iyadoo la eegayo yoolkeeda muddada fog ee horumarinta nidaamka gudaha semiconductor-ka. “Tan waxay fureysaa sahamin la saadaalin karo oo ku saabsan sahayda accelerators-ka waxqabadka sare leh, taasoo suurto gal ka dhigaysa in shirkadaha waaweyn ee Shiinaha ay qorsheeyaan in ay sameystaan gabdho GPU oo tayo sare leh, shabagyo bandwidth oo sarreeya, iyo qorshayaal tababarka moodeelada oo dhaqsaha badan, ” ayuu yiri Charlie Dai, Vice President iyo khabiir weyn oo falanqeyn ah oo ka tirsan Forrester. “Hyper-scale-ka ayaa sii wadi kara inay dardar geliyaan dhismaha kaabayaal AI ah iyo inay wax ka beddelaan awoodooda tababarka oo ah petaflops badan, halkii ay si keli ah ugu tiirsanaan lahaayeen beddelaad gudaha ah oo aan joogto ahayn. ” Saameynta waxqabadka shaqooyinka AI Oggolaanshahan waxaa laga yaabaa inay ku qasabto hyper-scale-yada Shiinaha inay dib uga fiirsadaan qaybinta shaqooyinka AI. “Waxaan filaynaa in shirkadaha ugu waaweyn ee Shiinaha ay qaataan istaraatiijiyad laba waddo ah oo waxqabad: inay mudnaanta siiyaan H200s si loo tababaro moodeelada waaweyn ee aasaasiga ah, iyagoo inta badan adeegsanaya chips-yada gudaha si loo qabto hawlgallada inference iyo tababaro yaryar, ” ayuu yiri Galen Zeng, maareeyaha cilmi-baarista sare ee cilmi-baarista semiconductor-ka ee IDC Asia/Pacific. Shiinaha ayaa sidoo kale xooga saaraya inay kor u qaadaan awoodoodda gudaha ee naqshadeynta iyo soosaarka chips-yada tababarka, iyadoo ujeedadu tahay in la yareeyo ku tiirsanaanta muddo dheer ah ee alaab-qeybiyeyaasha dibadda, Zeng ayaa intaas ku daray. Waxaa dhici kara in dhibaatooyinka ay ka soo baxaan haddii maamulka ay ku khasbaan in chips-yada la keeno dibadda in la isticmaalo si wada jir ah Accelerators-yada gudaha loo soo saaray.

Reuters ayaa sheegay inay suurtogal tahay inay jirto amarka noocaas ah. “Xaqiijinta in la isku daro chips-yada ayaa abuuraysa deegaan kumbuyuutar ah oo kala duwan, taas oo si weyn u kordhinaysa kakanaanta nidaamka, ” ayuu sharaxay Zeng. “Cadaadiska waxqabadka aan muuqan iyo la’aanta isku ekaanta habka isgaarsiinta ee u dhexeysa qaab-dhismeedka chips-yada kala duwan waxay kordhin kartaa culeyska howlgallada iyo dayactirka, waxayna dhalin kartaa qaar ka mid ah dib u dhac ku yimaada shabakadda. ” Si kastaba ha ahaatee, Zeng waxa uu sheegay in ogolaanshakan aysan yareyn doonin farqiga u dhexeeya hyper-scale-yada Shiinaha iyo kuwa Mareykanka, isagoo tilmaamay in H200 uu ka dambeeyo hal jiil oo ka mid ah qaab-dhismeedka Nvidia ee Blackwell, sidoo kalena tirada la ogolaaday ay aad uga hooseyso baahida guud ee Shiinaha. Saameynta shirkadaha caalamiga ah Hoggaamiyeyaasha IT-ga iyo shabakadaha shirkadaha caalamiga ah, horumarintan waxay ku dareysaa arrin kale qorshayaasha horudhaca ah ee dhismaha kaabayaal AI ee muddada fog. Iibsashada sii kordhaysa ee chips-yada Nvidia H200 ayaa ka caawin karta shirkadda inay kordhiso wax soo saarka, iyadoo sidoo kale suurto gal ka dhigeysa in qiimaha la dhimo kuwa caalamiga ah ee wax ka shaqeysta kaabayaal AI oo ku saleysan H200, ayuu yiri Neil Shah, Vice President cilmi-baaris ee Counterpoint Research.