U-CEO we-NVIDIA uJensen Huang ubhale kubhulogi esemthethweni kusihlwa namuhla (ngomhla ka-17) wathi i-OpenAI ithembise ukufaka ngobukhulu ingqalasizinda ye-AI ye-NVIDIA ngaphambi kuka-2030.
Izibopho zokufakwa kwezinhlelo ezikhona nezihleliwe ze-OpenAI zihlanganisa amandla ekhompyutha e-NVIDIA angaba ngu-12GW (ama-gigawatt); uma i-NVIDIA ikhulisa ukubambisana kwe-PORTS-Pike ngaphezu kwesigaba sokuqala esingu-4.25GW, inani lingakhula lifinyelele cishe ku-16GW.
Ngokusekelwe kulesi sikali, ngo-2030 amathuba ebhizinisi ahambisana namandla ekhompyutha e-NVIDIA azoba cishe ama-dollar aseMelika ayizigidi eziyizinkulungwane ezingu-600 (qaphela: cishe ama-yuan ayizigidi eziyizinkulungwane ezingu-4.05 ngesilinganiso samanje sokushintshisana).
UJensen Huang uthe ukwakha amafekthri e-AI kudinga uhlelo oluphelele lwezinsiza ezibalulekile, oluhlanganisa ama-chip athuthukile, ukupakishwa, inkumbulo namanethiwekhi, kanye nomhlaba, ugesi negobolondo lesakhiwo. Esikhathini esedlule, i-NVIDIA ibisebenzisa isikali sayo, ukubikezela kwesidingo sesikhathi eside kanye nobudlelwano bokubambisana nochungechunge lokuhlinzeka ukuze iqinisekise izinsiza ezibalulekile zama-semiconductor. Manje isilungiselela ukusebenzisa indlela efanayo emhlabeni, kugesi nasegobolondweni lesakhiwo (LPS), iqinisekise ngokukhethekile umthamo odingekayo wamafekthri e-AI asebenzisa amandla ekhompyutha e-NVIDIA.
Ngesikhathi esifanayo, i-NVIDIA imemezele ukubambisana ne-SB Energy ukuze iqinisekise umthamo we-LPS epaki yezobuchwepheshe i-PORTS-Pike ePortsmouth, e-Ohio, e-United States, ozosetshenziselwa ukufaka amandla ekhompyutha e-NVIDIA. I-OpenAI izoba ngumqashi wendawo.
UJensen Huang udalule ukuthi i-OpenAI izokwakha futhi iqhube ifekthri ye-AI esezingeni lomhlaba e-PORTS-Pike. Le fekthri izosebenzisa inkundla ephelele yefekthri ye-AI i-DSX ye-NVIDIA, ehlanganisa ama-GPU, ama-CPU, amanethiwekhi kanye nesofthiwe yengqalasizinda.
Ukufakwa kwesigaba sokuqala kulindeleke ukuthi kuhlinzeke ngomthamo wefekthri ye-AI ongu-4.25GW. Ngaso sonke isizukulwane sohlelo lwefekthri ye-AI ye-NVIDIA esifakwa e-PORTS-Pike, kulindeleke ukuthi kuhambisane nama-GPU e-NVIDIA angaba yizigidi ezingu-1.5, okuzoletha i-NVIDIA imali engaba ama-dollar aseMelika ayizigidi eziyizinkulungwane ezingu-150 kuya kwezingu-200 (cishe ama-yuan ayizigidi eziyizinkulungwane ezingu-1.01 kuya ku-1.35 ngesilinganiso samanje sokushintshisana). Phakathi nomjikelezo weminyaka engu-20, leli paki lingakwazi ukusekela ukuthuthukiswa kwezinhlelo izikhathi eziningi.
Umqondo oyinhloko wezomnotho lapha ukuthi ukubambisana kwe-LPS kuqinisekisa isizinda sefekthri ye-AI esingasetshenziswa isikhathi eside, kuyilapho imishini yekhompyutha ye-NVIDIA engaphakathi ingathuthukiswa njalo. Uhlelo olusha ngalunye lungaletha ukukhiqiza okuphezulu, ubuhlakani obengeziwe kanye nezinzuzo ezingcono kwezomnotho.
Esikhathini esizayo, i-NVIDIA ingase futhi ikhulise isikali sokubambisana kwe-PORTS-Pike ngaphezu kwesigaba sokuqala esingu-4.25GW, iqinisekise umthamo osele ongu-3.75GW.