Apple Deploys M5 Chips for Private Cloud Compute
Apple is now using its M5-series chips in the Private Cloud Compute servers that power its "Apple Intelligence" features. This infrastructure, revealed in recent software releases, is designed to process sensitive user queries in a secure, Apple-controlled environment, reinforcing the company's focus on privacy-by-design for its AI services.
- The M5 chip is built on a third-generation 3-nanometer process and features a 10-core CPU, a 10-core GPU, and a 16-core Neural Engine. Apple claims the CPU delivers up to 15% faster multithreaded performance and over 4x the peak GPU compute performance for AI compared to the M4 chip. - This move represents a significant upgrade, as Private Cloud Compute servers have historically used M2 Ultra chips, with Apple largely skipping the M3 and M4 generations for this infrastructure. - The Private Cloud Compute architecture is designed to be "stateless," meaning user data is only used to fulfill a request, is never stored or made accessible to Apple, and is deleted after processing. Apple allows independent experts to inspect the server code to verify this privacy promise. - The hardware upgrade is accompanied by a new "Private Cloud Compute Agent Worker" software component. This system runs a specialized version of iOS with a new "agentic architecture" designed to serve AI requests, and code to interface with it is present in iOS 26.4. - These M5-powered servers are being assembled at a factory in Houston, Texas, as part of Apple's $600 billion investment in domestic infrastructure. - This M5 deployment is considered an intermediate step before Apple deploys fully custom-designed AI server chips. Analyst Ming-Chi Kuo reports that those dedicated chips are slated for mass production in the second half of 2026 for deployment in 2027.
