OpenAI wants its AI smartphone to go into production in 2027, according to Ming-Chi Kuo

OpenAI seeks to accelerate its plans to develop a ‘smartphone’ propio with artificial intelligence (AI) so that it can be ready as soon as possible and enter production in 2027.

Analyst Ming-Chi Kuo, who previewed the conversations that OpenAI was having with industry suppliers to develop its own smartphone, has now advanced details about the processor that will power it.

According to him, the supplier that seems to be best positioned to develop the processor is MediaTek, which would work with a custom version based on Dimensity 9600 –which has not yet been announced–, the semiconductor firm’s 2nm chipset that is expected to arrive at the end of the year, as stated in an article shared in X.

[Industry Survey Update]OpenAI may be accelerating the development of its first AI agent mobile phone, targeting mass production as soon as 1H27. The reasons may include favorable IPO narrative at the end of the year, accelerated competition in AI agent mobile phones, etc. At present, MediaTek is more likely to obtain the processor order exclusively. The machine is expected to use a customized version based on Dimensity 9600 and be produced by TSMC N2P in 2H26. ISP… https://t.co/EYEtbWf16X

— Ming-Chi Kuo (@mingchikuo) May 5, 2026

The ‘chip’ for OpenAI would have an advanced ISP to help the AI ​​perceive the real world, as well as until NPU and technologies such as LPDDR6 RAM and UFS 5.0 storage which will help avoid memory bottlenecks. For security, the analyst points to the use of pKVM and online hash.

The first information suggested that mass production would begin in 2028, but now the analyst cites OpenAI’s rush to have your smartphone ready as soon as possibleto the point that it would go into production in the first half of 2027. The goal, he says, is to ship about 30 million units in 2027 and 2028.

The ‘smartphone’ will offer a different experience from what mobile phones currently offer, since it will not depend on applications and instead will integrate an artificial intelligence agentwhich will run both on the device and in the cloud, depending on the needs of the requested task.