news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

it home reported on september 14 that oracle founder larry ellison revealed an anecdote in the company's latest earnings call. he said that he and tesla ceo elon musk had personally asked nvidia ceo huang renxun to purchase the latest ai gpu.

at nobu restaurant in palo alto, ellison and musk pleaded with huang to supply them with nvidia’s latest gpus.we begged him, please sell it to us, no, sell it to us more, the more the better, we all need it, please” ellison recalled, “eventually it worked.”

the dinner was apparently a success, and ellison and oracle got the gpus they wanted. the company recently announced that it would build a supercomputing cluster called zettascale ai.composed of 131,072 nvidia gb200 nvl72 blackwell gpus, with ai performance up to 2.4 zettaflops, more powerful than musk and xai’s memphis supercomputing cluster, which currently has 100,000 nvidia h100 ai gpus.

oracle's ai plans require huge power support, so the company has obtained permission to build three modular nuclear reactors near its facilities. however, considering that the deployment of nuclear reactors may take years, oracle can temporarily follow musk's example and use large mobile generators to increase local power supply.

oracle is increasing its investment in ai. ellison said that in the next three years, the training cost of cutting-edge ai models will reach $100 billion (it home note: currently about 710.344 billion yuan), and oracle cloud infrastructure (oci) hopes to become one of the leaders in ai processing. ellison said in a conference call: "in this field, there will definitely be people who will do better than others. now there are many people trying. this is a race, and it is very important to get ahead."