M-Series Chip

The M-Series chip is the central computing component in newer Mac computers and iPads. It comes from Apple itself, not from a supplier as it used to. Until 2020, Apple bought the main processors for its computers from the manufacturer Intel. Since then, the company designs them itself and has them manufactured in Taiwan. What’s special: on this one component sit not only the computing cores, but also the graphics unit, a part for AI calculations, and the working memory. The models are named in sequence M1, M2, M3, and M4, each with more powerful variants such as Pro, Max, and Ultra.

What triggered Apple’s shift away from Intel

For decades, the rule was: whoever builds computers buys the processor. Apple broke this rule and saved a lot of money in the process. Above all, though, the company can now align hardware and software with each other. The operating system knows exactly which components are in the chip and uses them in a targeted way.

The practical benefit showed up immediately in battery life. A MacBook with an M-chip often lasts 15 to 18 hours, while earlier models with Intel processors rarely made it past eight. The reason lies in the design: M-chips originate from smartphone technology, which was designed from the outset for low power consumption. At the same time, the computing power remained high enough for video editing or programming.

For the industry, this was a signal. Other manufacturers such as Qualcomm and Google are now also developing their own chips based on a similar principle. Amazon and Microsoft, too, build their own processors for their data centers. As a result, the market for standard processors has come under pressure.

Everything on one piece of silicon

Classic computers distribute their tasks across separate components. The main processor sits at one spot on the board, the graphics card at another, and the working memory in its own slots. Data has to travel back and forth between these stations. These paths cost time and power.

The M-Series chip puts everything on one shared component, a so-called system on a chip. You can picture it like a workshop where all the craftsmen sit at one table instead of working in different buildings. The working memory belongs to all the units at the same time. Apple calls this Unified Memory. An image that the graphics unit has generated doesn’t need to be copied first for the computing cores to keep working with it.

A dedicated area of the chip is called the Neural Engine. It specializes in the type of computation that AI models need: very many simple multiplications at the same time. However, this approach comes at a price. The working memory is soldered in place and cannot be upgraded afterward. Anyone who buys too little must later replace the entire computer.

AI models on your own laptop

The chip becomes most visible when buying a Mac or iPad. The designations M4 or M4 Max stand for the performance tier. Anyone who just writes and browses the web barely notices a difference between the tiers. With video editing or large amounts of data, the difference is clear.

In tech news, the M-Series comes up mainly in connection with AI. Because the memory is shared, a MacBook with a lot of memory can run fairly large language models directly, locally. Local means: without an internet connection and without data going to a foreign server. This is exactly what’s difficult with many ordinary laptops, because there the graphics memory is scarce.

A common misconception is that M-chips are meant for training AI models. For that, one still uses data centers with specialized graphics cards, mostly from Nvidia. The M-Series chip plays to its strength in running finished models, not in training them. For developers, it is therefore primarily a convenient testing device.