news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

according to media reports, on september 3, openai, the developer of the generative artificial intelligence chatbot chatgpt, has become a major customer of tsmc's nominal 1.6 nanometer manufacturing process together with apple.

according to reports, apple has booked the first batch of a16 process production, and openai has also booked production. the a16 process is the most advanced manufacturing process in tsmc's circuit diagram so far, and is expected to go into production in two years (2h26).

although apple will be a major customer, openai has so far remained a pure software company. earlier this year, it was reported that openai ceo sam altman was eager to enter the chip manufacturing industry.

citing unnamed industry insiders, the media reported that openai had been negotiating with tsmc, hoping that tsmc would build a dedicated fab for it. however, the report said that the plan has been shelved. the report also said that openai is seeking to work with broadcom and marvell to design its own asic chips.

the openai chip, developed by broadcom and marvell, will first go into production on tsmc’s 3nm manufacturing process and then on the a16.

apple is believed to have a close relationship with openai. apple's personal intelligence system apple intelligence was launched in june 2024 and is believed to include chatgpt. it aims to enhance the capabilities of iphones, ipads and mac computers by combining generative ai models with personal environments. it covers areas such as writing and image creation.

apple is also rumored to be in talks for an investment that would value openai at $100 billion.

the a16 manufacturing process technology will be based on nanosheets and include tsmc's backside power delivery system, called super power rails (spr). this can increase logic density and performance by dedicating frontside routing resources to signals. tsmc claims that backside power reduces ir drop.

compared with the n2p process, the performance of a16 is improved by 8%~10% at the same vdd, the power consumption is reduced by 15%~20% at the same performance, and the chip density is increased by 7%~10%.