Apple shipped the first M1 Macs in November 2020. The MacBook Air, the MacBook Pro 13, and the Mac mini, all with the same chip. The benchmark numbers that started appearing in reviews over the following weeks were the kind that made you check your testing methodology twice.
The M1 was not just competitive with Intel chips of similar power consumption. In many tests, especially on workloads that could use the chip efficiently, it was outperforming Intel chips that drew several times more power. A passively cooled MacBook Air, which had no fan at all, was matching or beating laptops with active cooling and substantially higher power budgets.
The benchmark performance had several technical roots. Apple had been refining the same fundamental architecture for a decade in iPhone and iPad chips. The unified memory architecture, with the CPU and GPU sharing the same memory pool, eliminated entire categories of overhead that x86 systems had to manage. The neural engine and other accelerators were on-die rather than separate cards, with all the latency and bandwidth advantages that implied.
But the deeper reason the M1 worked was that Apple was designing for its own workloads. Most of what people do on a MacBook Air is web browsing, email, document editing, light photo work, and video calls. The M1 was tuned for those workloads in ways that a general-purpose chip aimed at the broader PC market could not match. When you do not need to be a good generalist chip, you can be a great specialist chip.
The reaction across the industry took different forms. Microsoft and the Windows-on-ARM ecosystem suddenly had a much clearer reference point for what was possible if you committed seriously to the architecture. Intel was left with a public competitive performance gap that it had not had to address in laptop computing for many years. The PC manufacturers that built laptops around Intel chips faced the awkward question of why their products were worse than a fanless Apple laptop.
The M1 was the first chip in a planned line. M1 Pro, M1 Max, and M1 Ultra would follow over the next eighteen months, each pushing the architecture further into territory that had previously belonged to higher-end workstation chips. The transition Apple had announced in June was working better than even the optimists had predicted.