Ngomhlaka-11 ngesikhathi sendawo e-United States, i-IBM yamemezela ukuthi isifinyelele esivumelwaneni seminyaka eminingi esibiza u-$240 million (qaphela: cishe ama-yuan ayizigidi eziyizinkulungwane ezingu-1.622 ngokwezinga lokushintshisana lamanje) ne-Together AI, inkundla yamafu ezalwa nge-AI.
I-IBM izofaka iqoqo le-NVIDIA (英伟达) HGX B300 ku-IBM Cloud yayo, lisekele izinsizakalo ze-Together AI zokwenza i-inference yamamodeli avulekile. Lolu hlelo kulindeleke ukuthi luqale ukusetshenziswa ngekota yokuqala ka-2027.
I-IBM ithe lokhu kusetshenziswa kuyisigcawu sokuqala esikhulu esakhelwe i-inference kuphela ku-IBM Cloud, sisebenzisa amasistimu e-NVIDIA HGX B300 nobuchwepheshe be-Spectrum-X Ethernet.
Isikhulu esiphezulu se-Together AI, u-Vipul Ved Prakash, sithe:
Amabhizinisi afuna ukusebenza kwamamodeli asezingeni eliphezulu, kodwa awafuni ukuthwala izindleko eziphezulu zamamodeli avaliwe. Lokhu kungenzeka kuphela uma ingqalasizinda engaphansi ishesha, ithembekile futhi ingasetshenziswa ngezinga elikhulu.
Ukubambisana ne-IBM ne-NVIDIA kusibekele isisekelo salokho. Leli qoqo lisenza sikwazi ukuhlinzeka izinkampani eziningi ngamandla e-inference asezingeni lokukhiqiza ngokushesha, okuyisinyathelo esibalulekile emzamweni wethu wokwenza i-AI evulekile ibe yinketho yokuqala yamabhizinisi.
I-Together AI isanda kuqeda umzuliswano woxhaso wo-Series C ka-$800 million, ngenani elilinganiselwa ku-$8.3 billion, futhi amandla ayo okucubungula ama-token nyanga zonke afinyelela ku-400T.