Kwenzekeni
I-International Business Times ibika ukuthi i-Microsoft ingemuva kakhulu kwezinhloso ezihlongozwayo zokusatshalaliswa kwe-AI chips kuzo zonke izikhungo zayo zedatha zomhlaba wonke. Lesi sihloko sithi uphenyo lwe-Guardian, olucaphuna imithombo ye-Microsoft kanye nemibhalo yangaphakathi, lubeke inani elifakiwe lenkampani kuma-chips we-AI ayizigidi ezingu-2.2, kuyilapho abanye abahlaziyi beveze ukuthi lelo nani liphindwe kabili. Umthombo awubahlonzi labo bahlaziyi noma unikeze impendulo yomphakathi engu-Microsoft eqinisekisa noma ephikisa isibalo.
I-International Business Times ibika ukuthi uphenyo lwe-Guardian luthole ukuthi i-Microsoft ingemuva kwemigomo yayo yokusatshalaliswa kwe-AI chips kuzo zonke izikhungo zayo zedatha zomhlaba wonke. Ngokusho kwe-IBTimes, imithombo ye-Microsoft itshele i-Guardian ukuthi inani lenkampani eliphelele lishintshile ngonyaka odlule naphezu kwezigidigidi zamaRandi ekusetshenzisweni konyaka kwesikhungo sedatha. I-athikili ihlanganisa isilinganiso esimaphakathi kumadokhumenti angaphakathi ashiwo umbiko we-Guardian: cishe ama-AI chips ayizigidi ezingu-2.2 afakiwe. I-IBTimes ithi ingqikithi ingaphansi kwengxenye yalokho abanye abahlaziyi ababekubikezele ngesiteji sokukhishwa kwe-Microsoft. Umthombo awusho amagama alabo bahlaziyi, usho inani eliphezulu eliqageliwe, noma uchaze indlela esetshenziswa ukubala ama-chips.
Umbiko wethula ukushoda njengengxenye yenkinga ebanzi yamakhono. I-IBTimes ithi isidingo sekhompiyutha ye-AI siyaqhubeka nokudlula ukunikezwa okuvela eqenjini elincane labakhiqizi abaphethe imakethe ye-AI-chip yomhlaba. Ixhuma isibalo se-chip esibikiwe se-Microsoft nemibuzo mayelana nokuthi inkampani ingakwazi yini ukuhlangabezana nezibopho zomthamo wekhompuyutha oxhumene nokutshalwa kwezimali kwesikhungo sedatha. Kodwa-ke, umthombo awucacisi ukuthi yiziphi izikhungo zedatha ze-Microsoft ezithintekayo, yiziphi abakhiqizi bama-chip noma izinhlobo zama-chip ezifakiwe, noma ukuthi isibalo sivala ama-chip asetshenzisiwe kuphela noma nezinto zokusebenza ezi-odiwe, ezilethiwe noma ezigciniwe. Inhloso ye-Microsoft uqobo, umnqamulajuqu ofanelekile kanye nemibandela yanoma yikuphi ukuzibophezela kwamandla nakho akunikeziwe.
Isihloko sibe sinweba ingxoxo ngale kokusatshalaliswa okubikiwe kwe-Microsoft. Ithi ukwakhiwa kwesikhungo sedatha kubhekene nokuphikiswa kwendawo, ikakhulukazi ngenxa yengcindezi engase ibe khona emanzini, futhi ithi iNew York imise ukwakhiwa kwezikhungo ezilinganiselwa ku-50 megawatts noma ngaphezulu. Iphinde ibike ukuthi amakhasimende ebhizinisi abhekane nezindleko eziphakeme njengoba abahlinzeki be-AI basebenzisa amanani ethokheni asekelwe ekusetshenzisweni. Ngokwehlukana, i-IBTimes icaphuna i-J.P. Morgan Global Research projection ukuthi amanani e-DRAM, ingxenye yenkumbulo esetshenziswa kumthwalo wedatha we-AI wedatha, ingakhuphuka ngaphezu kuka-400% ekupheleni kuka-2026. Lezi zimangalo ezizungezile zethulwa njengomongo; umthombo awuqinisekisi ukuthi enye yazo ibangele ngokuqondile ukushoda kwe-chip okubikiwe kwe-Microsoft.
Imininingwane yomthombo: ibtimes.com ↗
Kungani kubalulekile
Ukushoda okubikiwe kungase kuthinte ikhono le-Microsoft lokukhulisa amandla ekhompuyutha kumasevisi e-AI futhi ihlangabezane nezibopho ezihambisana nokutshalwa kwezimali kwayo kwesikhungo sedatha. Iphinde ikhombise ukuthi ukuhlinzekwa kwe-chip, izindleko zememori, ukusetshenziswa kukagesi kanye nokuphikiswa kwendawo ekwakhiweni kwesikhungo sedatha kungavimbela kanjani ukunwetshwa kwe-AI ngisho nalapho izinkampani zenza imali enkulu engqalasizinda. Isikali nomthelela webhizinisi wegebe elibikiwe kusalokhu kungaqinisekisiwe ngokuzimela.
Uma isibalo esibikiwe sinembile, ukunwetshwa kwengqalasizinda ye-Microsoft kungase kuqhubeke kancane kunalokho okulindelwe ngaphandle okusikisela. Lokho kubalulekile ngoba izinsiza ze-AI zincike ekufinyeleleni ikhompuyutha ekhethekile, futhi isisekelo esincane esifakiwe singakhawulela ukuthi inkampani ingeza ngokushesha kangakanani umthamo, iqeqeshe noma inikeze amamodeli, futhi isekele amakhasimende ebhizinisi. Umthombo awubali noma yiziphi izithiyo zesevisi eziwumphumela, izibopho ezigejiwe, ukubambezeleka noma imiphumela yemali engenayo, ngakho leyo miphumela akufanele ithathwe njengamaqiniso amisiwe. Izindaba ezisheshayo igebe elibikiwe phakathi kwesibalo sangaphakathi se-chip kanye nokuqagelwa komhlaziyi okungashiwongo.
Lo mbiko uphinde uveze ukuthi ingqalasizinda ye-AI ivinjwe imali engaphezulu kwezindleko zezinkampani. Ukutholakala kwe-chip, ukunikezwa kwenkumbulo okuthuthukisiwe, ukwakhiwa kwesikhungo sedatha, izidingo zikagesi namanzi, nezimvume zasendaweni zonke zingathonya ijubane lokuthumela. I-IBTimes icaphuna lezi zingcindezi ngenkathi ichaza ukumelana okubanzi nezindawo ezintsha kanye nokukhuphuka kwezindleko zokusebenzisa ibhizinisi. Lokho kuxhumana kuhambisana nokuqonda komphakathi umkhondo wezinyawo we-AI, kodwa i-athikili ayinikezi ubufakazi obanele ukuze kutholwe ukuthi isici ngasinye sinengxenye engakanani esikhundleni se-Microsoft noma ukuthi i-Microsoft ibhekene nezingqinamba ezingajwayelekile yini uma iqhathaniswa nezinye izinkampani ezinkulu ze-AI.
Kumabhizinisi nabakhi benqubomgomo, umbiko uphakamisa umbuzo osebenzayo mayelana nokuthi ukunwetshwa kwe-AI kufanele kukalwe kanjani. Ukutshalwa kwezimali okumenyezelwe kanye nomthamo ohleliwe wesikhungo sedatha akusho ukuthi uhumushela kuzinsiza zekhompuyutha ezifakiwe nezisebenzayo. Isilinganiso se-chip-count singase futhi sishiye ukusetshenziswa, ukunethiwekha, ukutholakala kwamandla, ukupholisa kanye nezingqinamba zesofthiwe. Umthombo awunikezeli ukuhlukaniswa kwalokhu okuguquguqukayo. Ngenxa yalokhu, isibalo esibikiwe sezigidi ezingu-2.2 siyinkomba engase ibe umphumela wenqubekelaphambili yokuthunyelwa, kodwa akusona ngokwaso isilinganiso esiphelele somthamo we-AI we-Microsoft noma isikhundla sokuncintisana.
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
Ubufakazi obulandelayo obubaluleke kakhulu kungaba impendulo yomphakathi evela ku-Microsoft, imininingwane eyengeziwe mayelana nemibhalo yangaphakathi, kanye nokucaciswa kokuthi isibalo esiyizigidi ezingu-2.2 sabalwe kanjani. Buka futhi ukuze uthole ulwazi mayelana nengxubevange yama-chips, izindawo nomthamo wezikhungo zedatha ezithintekile, nokuthi ingabe izithiyo zokuhlinzeka noma izimvume nokubambezeleka kokwakha kuyizimbangela eziyinhloko. Umthombo uphinde ucaphune isibikezelo sika-JP Morgan sokunyuka okungaphezulu kwama-400% kumanani entengo ye-DRAM ekupheleni kuka-2026, kodwa lokho kuqagela akuhlolwa ngokuzimela lapha.
Iphuzu lokuqala lokuqinisekisa impendulo ye-Microsoft. Ukucaciswa komphakathi okuwusizo kungasho ukuthi isibalo esiyizigidi ezingu-2.2 sinembile yini, lokho inkampani ebala njenge-chip ye-AI efakiwe, nokuthi isamba siqhathaniswa kanjani nemigomo yayo yangaphakathi. Umthombo awukubiki ukuphika kwe-Microsoft, ukuqinisekiswa noma incazelo. Ukubika okwengeziwe okuvela ku-Guardian noma kwezinye izitolo kungase futhi kuhlonze imibhalo, izinsuku zabo, abantu abanikeze ulwazi kanye nesisekelo sokuqagela komhlaziyi. Kuze kube yileso sikhathi, inombolo emaphakathi kufanele ihlale ibonisiwe kunokuba yethulwe njengoba iqinisekiswe ngokuzimela.
Imininingwane yamandla izonquma ukubaluleka kokusebenza kombiko. Buka ukuze uthole ulwazi ngezinhlobo zezisheshisi ezihilelekile, ukusetshenziswa kwazo kuzo zonke izifunda, ukutholakala kwamandla esikhungo sedatha, amashejuli okwakha namazinga okusetshenziswa. Kuzophinde kube nendaba ukuthi ingqikithi ye-chip ye-Microsoft ishintshile yini ngenxa yokubambezeleka kokuthengwa kwempahla, izingqinamba zokukhiqiza, ukubambezeleka kwesayithi, ukulinganiselwa kwenethiwekhi noma ushintsho lwamabomu lwesu. Ayikho kulezi zincazelo esungulwe umthombo, okusho ukuthi akufanele ithathwe ekubalweni okubikiwe kuphela.
Izinkomba zokuphakela nezinqubomgomo zomphakathi zingenye indawo okufanele iqashwe. I-IBTimes ithi izintengo ze-DRAM zingakhuphuka ngaphezu kuka-400% ekupheleni kuka-2026, ngokusekelwe ku-JP Morgan Global Research projection, futhi ichaza ingcindezi evela ekuphikisaneni kwesikhungo sedatha nokukhathazeka kwamanzi. Lezo zimangalo zingathinta izindleko nokuba nokwenzeka kwengqalasizinda ye-AI, kodwa umthombo awunikezi isiqinisekiso esizimele noma ukuqagela kwesibikezelo esinemininingwane. Landela idatha yentengo elandelayo, ukudalulwa kwenkampani, izinqumo ezivumela kanye nobufakazi bokubambezeleka kwangempela kokwakhiwa noma ukuthunyelwa ngaphambi kokuthatha iziphetho ezibanzi mayelana nejubane lokwakhiwa kwe-AI.