Apple is deviating from its previous chip strategy with the M6: From M1 to M5, there were several variants of each Apple Silicon model, but this is supposedly not planned for the M6. According to a new Bloomberg report, Apple has already started the “tape-out” of several M7 chips, a process that usually serves to finalize the chip design. According to the information, there will only be one standard M6 processor, with M6 Pro, M6 Max, or even M6 Ultra being omitted.
This would mean the M6 would only land in an entry-level MacBook Pro and possibly a MacBook Air, possibly also in an iMac and possibly in a Mac mini model – although the latter compact computer has always been equipped with a standard chip and a Pro variant recently. The reason, according to Bloomberg, is that Apple is planning major changes to the integrated AI accelerators for the M7, which apparently do not fit into the M6.
Very much RAM addressable
The first M7 is planned for the first half of 2027, according to Bloomberg. By the end of the year, M7 Pro and M7 Max will follow, and an Ultra version will follow in 2028 (which the company has sometimes omitted in recent years). With the new M7 Ultra, Apple will be able to address up to 1.5 TB for the first time since the Mac Pro with an Intel chip in 2019.
The central innovation with the M7 will be the significantly accelerated local AI capabilities. Bloomberg speaks of “significant improvements” in the integrated neural processing units (NPUs), which are traditionally called Neural Engines at Apple. Apple is apparently approaching Nvidia's Blackwell chips with this. Internally, Apple is also already working on the M8, which is also said to come with high(er)-end variants. From 2028 onwards, 1.4-nanometer manufacturing processes are planned for this. AI is no longer just “another feature” at Apple, according to Bloomberg, but of central importance.
Local AI in the fast lane
Currently, there is still a clear separation between local and cloud AI. The latter also caused the massive inflation in DRAM and NAND memory prices. Apple has been relying on local AI systems for several years, for example through the clustering of multiple Macs, but also through the use of individual machines with tools like Ollama or LM Studio.
The local use of large language models has many advantages. For example, there are no token costs, which have become rapidly pricier in recent months, and you don't send your data to the cloud for training or other purposes. Local models are also catching up significantly technically, as combinations like Hermes and Qwen show. In the foreseeable future, AI data centers may therefore be needed primarily for training and no longer for inference, which is so expensive today. Apple seems to be betting heavily on this trend.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
(bsc)
Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.
This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.