Kwenzekeni
I-Rhodium Group ishicilele umbiko ohlaziya ukusebenza kwezimali konjiniyela abakhulu be-AI base-China abayisikhombisa, okuhlanganisa i-DeepSeek, Moonshot AI, Z.ai, MiniMax, Alibaba, ByteDance, ne-Kuaishou. Umbiko ulinganisela ukuthi lezi zinkampani zithole inzuzo ehlangene yonyaka ephindaphindayo (i-ARR) elinganiselwa ku-$10.7 billion phakathi kukaMashi no-Agasti. Uma kuqhathaniswa, i-OpenAI kanye ne-Anthropic zibike i-ARR ehlanganisiwe eyedlula i-US$100 billion ngesikhathi esifanayo. Umbiko uqokomisa umehluko omkhulu wokulinganisa wemali engenayo, uphawula ukuthi amafemu aseShayina afana ne-DeepSeek kanye ne-Moonshot banokuphindaphinda kwe-valuation-ARR okungu-163x no-50x, ngokulandelanayo, uma kuqhathaniswa no-34x we-OpenAI kanye no-21x we-ZXZAQAIU3QX Yize imali engenayo ikhula ngokushesha, njengoba i-Z.ai ibika ukukhuphuka okungu-400% konyaka nonyaka, umbiko uphakamisa ukuthi amasu esisindo esivulekile enza kube nzima imizamo yokwenza imali konjiniyela base-China.
I-Rhodium Group, inkampani yabacwaningi ezinze e-US, ikhiphe umbiko ngoLwesine ochaza ngezibalo zezezimali zabathuthukisi abakhulu be-AI base-China abayisikhombisa. Ucwaningo lubale ukuthi lezi zinkampani, ezihlanganisa i-DeepSeek, Moonshot AI, Z.ai, MiniMax, i-Alibaba Group Holding, i-ByteDance, ne-Kuaishou Technology, zenze imali engenayo yonyaka ehlanganisiwe (ARR) elinganiselwa ku-$10.7 billion kusukela ngoMashi kuya ku-Agasti.
Lesi sibalo simele cishe u-10% we-ARR ehlanganisiwe ebikwe yi-OpenAI kanye ne-Anthropic, edlule i-US$100 billion ngesikhathi esifanayo. Ngokuqondile, i-OpenAI ifinyelele ku-US$40 billion ku-ARR ngo-August, kuyilapho i-Anthropic yayiku-$65 billion ngoJulayi. Phakathi kwabathuthukisi baseShayina, i-ByteDance ihole iqembu nge-US $ 4 billion ku-ARR kusukela ngoJulayi, ilandelwa i-Alibaba ngamadola ayizigidi eziyizinkulungwane ezingu-2.4 ngo-Agasti.
Umbiko ugqamisa umehluko omkhulu ekubaleni okuhlobene nemali engenayo. U-Rhodium uphawule ukuthi ukuphindaphinda kwe-valuation-to-ARR kwe-DeepSeek ne-Moonshot kumi izikhathi ezingu-163 nezingu-50, ngokulandelana. Uma kuqhathaniswa, ukuphindaphinda kwe-OpenAI bekuyizikhathi ezingama-34 kuthi i-Anthropic kube izikhathi ezingama-21. Umbiko uchaze lezi zilinganiso zase-China 'njengezimba eqolo' uma kuqhathaniswa nemithombo yazo yamanje yemali engenayo, nakuba zihlala ziphansi kakhulu ngokwemibandela ephelele kunozakwabo base-US.
Ngaphandle kwegebe lemali engenayo, izinkampani ze-AI zaseShayina zithola ukukhula ngokushesha. I-Z.ai ibike ukuthi inyuke ngo-400% unyaka nonyaka emalini engenayo engxenyeni yokuqala iye kuma-yuan ayizigidi ezingu-953.9, kanti imali engenayo ye-MiniMax inyuke ngo-283% yaya ku-US$116.6 million. Kodwa-ke, inhloko ye-Macquarie Group yase-Asia ye-inthanethi kanye nocwaningo lwesoftware, u-Ellie Jiang, utshele iSouth China Morning Post ukuthi i-Z.ai kanye ne-MiniMax zingaqhubeka zenza ukulahlekelwa kuze kube u-2030 ngenxa yezindleko eziphezulu zamandla ekhompyutha adingekayo ukuze kugcinwe ukuncintisana kwezobuchwepheshe.
Imininingwane yomthombo: scmp.com ↗
Kungani kubalulekile
Lo mbiko uhlinzeka ngophawu lwezezimali olubambekayo mayelana nesimo sokuncintisana phakathi kwezimboni zase-US kanye ne-Chinese AI. Iveza ukuthi naphezu kokutshalwa kwezimali okunolaka nokulinganisa okuphezulu, izinkampani ze-AI zase-China okwamanje zenza ingxenye encane yemali engenayo yozakwabo base-US. Leli gebe liphakamisa imibuzo mayelana nokusimama kokulinganisa kwamanje kanye nempumelelo yamasu esisindo esivulekile ekukhiqizeni imali eqondile. Kubatshalizimali nabakhi bezinqubomgomo, lezi zibalo zinikeza isithombe esicacile seqiniso lezomnotho elilandela umjaho wezobuchwepheshe, ezigqamisa ukuthi ngenkathi amamodeli aseShayina enciphisa igebe lokusebenza, amamodeli amabhizinisi awasekelayo asakhula futhi abhekene nezinselele ezihlukile zokwenza imali uma kuqhathaniswa nemithombo evaliwe, amamodeli asindayo okubhaliselwe we-OpenAI kanye ne-Anthropic.
Lo mbiko ugcizelela inselelo enkulu embonini ye-AI yaseChina: ukuguqula ukulingana kwezobuchwepheshe kube amamodeli webhizinisi esimeme. Nakuba abathuthukisi baseShayina beye banciphisa ngokushesha igebe lokusebenza nozakwabo base-US, ukukhiqiza imali yabo kusasele ngemuva kakhulu. Lo mehluko uphakamisa ukuthi umjikelezo wamanje wokutshalwa kwezimali e-Chinese AI uqhutshwa kakhulu yisu lokuma kanye nokutholwa kwesabelo semakethe kunenzuzo esheshayo.
Into esemqoka efaka inkimbi ukwenza imali ukutholwa okusabalele kwamamodeli omthombo ovulekile kanye nesisindo esivulekile onjiniyela base-China. U-Rhodium uphawule ukuthi leli su lenza kube nzima kubathuthukisi boqobo ukuthi bathwebule imali engenayo, njengoba abahlinzeki bamafu benkampani yangaphandle bengakwazi ukusebenzisa lawa mamodeli futhi badayise ukufinyelela ngaphandle kokukhokha abadali. Lokhu kuhlukile kumthombo ovaliwe, amamodeli asuselwe ekubhaliseni we-OpenAI kanye ne-Anthropic, avumela ukuthwebula imali engenayo eqondile.
Idatha yezezimali ihlinzeka ngophawu olubalulekile lwabatshalizimali abahlola iphrofayili yobungozi yamasheya e-Chinese AI. Ukuphindaphindeka okuphezulu kokulinganisa kwenzuzo kubonisa ukuthi abatshalizimali banamanani ekukhuleni okuphawulekayo okuzayo noma inani lesu elingakaboniswa enzuzweni yamanje. Lokhu kudala ubungozi obungaba khona uma ukukhula kuhamba kancane noma uma izindleko zokugcina ubuholi bezobuchwepheshe ziqhubeka nokudlula ukukhula kwemali engenayo.
Lo mbiko uphinde ugqamise ukuvela kwezinhlobo zebhizinisi le-AI, ikakhulukazi mayelana namasu esisindo esivulekile. Njengoba amafemu aseShayina afana ne-Moonshot kanye ne-Alibaba befuna izivumelwano zokwabelana ngemali engenayo nabasebenzisi abakhulu, imboni ingase ibone ushintsho endleleni i-AI yomthombo ovulekile eyenza ngayo imali. Lokhu kungaba nomthelela obanzi emakethe ye-AI yomhlaba wonke, okungenzeka kube nomthelela ekutheni abanye onjiniyela babhekana kanjani nokukhishwa kwesisindo esivulekile.
I-Interactive Mechanism: Indlela Esebenza Ngayo Ngempela
Hlola ubuchwepheshe obuyisisekelo ngemuva kwalokhu kuthuthukiswa ngokuhlanganyela.
Which component of an AI application is the machine-learning model itself?
Ongakubuka ngokulandelayo
Abatshalizimali kufanele baqaphe ukuthi izinkampani ze-AI zaseShayina zingasebenzisa ngempumelelo izivumelwano zokwabelana ngenzuzo zamamodeli azo anesisindo esivulekile, njengoba kuxoxiwe ngabakwaRhodium. Ukwengeza, ama-IPO azayo we-Moonshot AI nokufakwa kuhlu okungaba khona kwabanye abadlali abakhulu kuzohlola imakethe yalokhu kulinganisa okuningi. Ikhono lalezi zinkampani lokuguqula ukutholwa kwabasebenzisi ngokushesha kube yinzuzo esimeme ngaphandle kokufaka izindleko zekhompiyutha ezingasimame kuzoba inkomba eyinhloko yokusebenza kwesikhathi eside komkhakha we-AI waseChina.
Ukuqaliswa kwezivumelwano zokwabelana ngenzuzo zamamodeli anesisindo esivulekile kuzoba inkomba ebalulekile yokuthi izinkampani ze-AI zaseShayina zingakwazi yini ukuthuthukisa ukusebenza kahle kwazo zokwenza imali. U-Rhodium uphawule ukuthi i-Moonshot isezingxoxweni ne-Microsoft, Amazon, kanye ne-Google, namasheya angaba khona afinyelela ku-30%. Impumelelo yalezi zingxoxo ingaba yisibonelo kwabanye onjiniyela.
Iminikelo yomphakathi yokuqala ezayo (ama-IPO) yezinkampani zase-China ze-AI, ezifana nokugcwalisa okuyimfihlo kwe-Moonshot AI kwe-IPO yase-Hong Kong kanye nomjikelezo woxhaso wangaphambi kwe-IPO we-DeepSeek we-DeepSeek, kuzohlola ukuzethemba kwemakethe kulokhu kulinganisa okuphezulu. Ukwamukelwa kwalezi zinhlu kuzonikeza ukuqonda kokuthi abatshalizimali bazimisele yini ukugcina ukuphindaphinda okuphezulu lapho bebhekene nemali engenayo elinganiselwe yamanje.
Ikhono lezinkampani ze-AI zaseShayina zokuphatha izindleko zekhompiyutha kuyilapho zigcina ukuncintisana kwezobuchwepheshe zizobaluleka. Njengoba kuphawulwe i-Macquarie Group, izindleko eziphezulu zamandla ekhompyutha zingagcina izinkampani ezifana ne-Z.ai ne-MiniMax zibomvu kuze kube ngu-2030. Ukuqapha izakhiwo zazo zezindleko kanye nokuthuthukiswa kokusebenza kahle kuzobaluleka ekuhloleni ukusebenza kwesikhathi eside.
Amandla okuncintisana phakathi konjiniyela be-AI base-US nabaShayina azoqhubeka nokuvela, izinhlangothi zombili ziqhudelana ngentengo, ukusebenza, nokutholwa komthombo ovulekile. Okutholwe yilo mbiko kusikisela ukuthi nakuba igebe lokusebenza liyancipha, igebe lezomnotho lisabalulekile. Intuthuko yesikhathi esizayo kumasu entengo nokukhishwa kwamamodeli kuzoba ukhiye wokuqonda ukuthi lokhu kuguquguquka kudlala kanjani.