Report: Apple commissions Foxconn to build new M4 AI servers
It's not just Macs for end customers that are getting new chips. AI data centers will also be regularly overhauled in the future.
M4 chips: There is still no Ultra – but the M2 is long outdated.
(Image: Apple)
Apple intends to use the new M4 SoCs in its Apple Intelligence infrastructure in the near future. The Japanese business news agency Nikkei reports that servers currently still using the M2 Ultra are to be replaced. Apple's in-house supplier Foxconn has been commissioned to manufacture new servers in Taiwan. However, this could clash with the company's other orders.
Necessary for private cloud compute
Like many other cloud companies, Apple has so far relied primarily on standard technology for iCloud servers rather than its own computers, buying capacity from Google (Cloud), Amazon (AWS) and now even Microsoft (Azure), among others – in addition to its own data centers. Apple Intelligence is bringing about a fundamental change here.
Videos by heise
As part of the Private Cloud Compute (PCC) concept, AI workloads are outsourced to servers at Apple, but this is encrypted and cannot be viewed by Apple. For this to work, the remote station must have Apple's security measures such as the Secure Element. This also makes Apple SoCs mandatory. Accordingly, the company is currently investing heavily in building its own facilities.
Data centers must become faster
According to Nikkei, which spoke to TrendForce analyst Frank Kung among others, the M4 is intended to help make the infrastructure more performant. It was not specified which SoCs will be used. Apple currently uses the M4, M4 Pro and M4 Max for the Mac; an M4 Ultra (for Mac Studio and Mac Pro) is not yet officially on the market. Other analysts have previously claimed that the M4 will be used as part of PCC.
Apple tries to process as many Apple Intelligence requests as possible locally on the device –, i.e. an iPhone from A18 Pro, an iPad from M1 or an Apple Silicon Mac –. However, if the workloads become too complex, they are outsourced to the cloud. Users only notice this because the AI system then requires an internet connection. Apple is carrying out a slow rollout of Apple Intelligence. It is currently only available in English and the EU rollout is scheduled for April 2025. This may also have to do with infrastructure issues. Incidentally, it is still unclear whether Foxconn can accept Apple's order to the desired extent: according to Nikkei, the company also has full order books in Taiwan due to Nvidia AI servers.
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)
