In June 2020 Apple announced it would stop buying processors from Intel. Five months later it unveiled laptops with its own chip, made by TSMC, that lasted hours longer on a charge than the Intel machines they replaced. In my view that episode told us less about Apple than about Intel: a customer that big left, and the product it moved to was simply better.
Two years, announced on stage
On 22 June 2020, at WWDC, Apple said it would ship the first Mac with Apple silicon by the end of the year and complete the transition in about two years. It promised industry-leading performance per watt, and it said a translation layer called Rosetta 2 would run existing Mac apps that had not been updated yet.
The finish line moved a little. The Mac Pro, the last Intel Mac, went to Apple silicon with the announcement of 5 June 2023, and Apple's own release says that completes the transition. That is just under three years to change the architecture of an entire product line while the old software kept running. Few companies could even attempt it.
Rosetta 2 was not magic, and AnandTech measured it honestly: translated code averaged roughly 70 to 80 percent of native speed in its SPEC 2017 floating-point tests, and the gcc test in the integer suite managed 49.87 percent. Still, AnandTech called the results outstanding for a translation layer, and I agree that running years of x86 software on a brand new architecture without user effort is a serious piece of engineering.
M1, and who built it
Apple unveiled M1 on 10 November 2020: a system on a chip on a 5-nanometer process with 16 billion transistors. Apple's release does not name its foundry. A TechInsights die analysis does: the M1 is made on TSMC's N5 FinFET process. Apple designs it, TSMC manufactures it, and in my view that division of labour is exactly what Intel could not match.
AnandTech tested the Mac mini a week later. The performance cores run at 3.2 GHz, and a single-threaded gcc compile added only about 6.3 W at the wall over idle. Under heavy multithreaded loads the whole machine drew between 18 and 27 W above idle, which AnandTech estimated as a 20 to 24 W chip. Against Intel's newest laptop part at the time, the 28 W Core i7-1185G7, the single-core result matched it at worst, and in multithreaded work AnandTech called Apple's lead indisputable, at lower power.
It was not a clean sweep, and I would rather say so. AMD's Ryzen 4000 chips still beat M1 in multithreaded integer work, though at higher power. Against Intel, the verdict in that review was one-sided.
The battery and the heat
Tom's Guide ran its standard web-browsing battery test at 150 nits. The M1 MacBook Air lasted 14 hours 41 minutes; the 2020 Intel MacBook Air lasted 9 hours 31. The M1 MacBook Pro lasted 16 hours 32; the 2020 Intel 13-inch Pro lasted 10 hours 21. That is about five hours more in the Air and just over six in the Pro, from the same company in the same year.
Heat tells the same story from the other direction. In July 2018 a YouTuber, Dave Lee, showed a Core i9 MacBook Pro slowing down under load. Apple acknowledged it across all 2018 MacBook Pro models, blamed a missing digital key in the firmware that affected thermal management, and shipped a macOS update. I will take Apple's word that it was firmware. But a laptop where one missing key can drag a flagship chip's clock speed down under heavy load tells me how little margin the design had. That is my reading, not Apple's.
Why Apple really left
Bloomberg reported as early as April 2018 that Apple planned to move Macs to its own chips from 2020. The reasons Apple gave in public were performance per watt and graphics. A different account came from François Piednoël, a former Intel engineer, in an interview with PC Gamer reported by MacRumors in June 2020. He said the quality assurance of Skylake was "abnormally bad", and that Apple became the number one filer of problems in that architecture. I cannot verify that, and Apple has not said it. It is one engineer's account, and I present it as that.
What can be checked is what Intel said a month after WWDC. On 23 July 2020 it announced its 7nm CPU timing was slipping about six months, with yields about twelve months behind its internal target. Its 7nm PC chips moved to late 2022 or early 2023. Its CEO said Intel would use someone else's process technology if it had to, and the shares fell about 10 percent in after-hours trading.
To me, no reasonable reading of that timeline makes Apple's decision look early. It looks like the only sensible option.
The M-series since
Apple has kept the cadence. The M2 arrived on 6 June 2022 on a second-generation 5-nanometer process with 20 billion transistors, and Apple claimed nearly twice the CPU performance of the latest 10-core PC laptop chip at the same power. The M3, announced 30 October 2023, was what Apple called the first personal computer chip on 3 nanometers; TechInsights identifies it as a TSMC 3nm part. The M4 followed on 7 May 2024, debuting in the iPad Pro, and Apple said it matches M2 performance at half the power. In October 2025 the M5 reached the 14-inch MacBook Pro with Apple quoting up to 24 hours of battery, and the M5 Pro and M5 Max followed on 3 March 2026.
Those generational figures are Apple's own claims, not independent tests, so treat them accordingly. The independent evidence is the M1 review, and it was enough.
What it did to the rest of the PC industry
On 20 May 2024 Microsoft launched Copilot+ PCs: Surface Laptop and Surface Pro machines powered by Qualcomm's Snapdragon X Elite and X Plus, on sale from 18 June. Look at what Microsoft chose to compare against. Microsoft said the new Surface Laptop was faster than the latest MacBook Air, based on its own multithreaded Cinebench 2024 test of a pre-release unit against the 13-inch MacBook Air with M3. Four years after WWDC, the yardstick for a Windows laptop was a MacBook Air. Windows on Arm needs a translation layer too, and Microsoft calls its version Prism.
Then there is Intel itself. PCWorld reported in June 2024 that the compute and controller tiles of Intel's Lunar Lake laptop chip are made at TSMC, and at Computex Intel's then-CEO, Pat Gelsinger, said Lunar Lake picked TSMC as the right technology at that point in time. The company that sold the world its processors for decades sent its new laptop chip to the same foundry that builds Apple's.
Apple did not out-argue Intel; it took Intel out of the equation and let its own designers and TSMC's factories do the talking.
What I take from it
Control of the chip is control of the product. Apple decided what its laptops needed, designed for it, and bought the best manufacturing available. Everyone else was waiting for a supplier.
For anyone buying or supporting machines, the lesson is plain: judge a laptop by its battery test and its sustained performance, not by a processor brand on a sticker. And a supplier that cannot keep its roadmap loses its biggest customers, however long the relationship lasted.