The First ARM Processor

The history of the ARM architecture starts with a chip delivered to Acorn Computers in Cambridge, England, on 26 April 1985: the ARM1, whose instruction set was designed by Sophie Wilson and whose microarchitecture was designed by Steve Furber. The name then stood for Acorn RISC Machine. In Furber's account the chips arrived in the morning, and after a few board-level problems with the test card were fixed, they were running BBC BASIC by the afternoon. The Arm instruction set used today in everything from phones to data centres descends directly from this 32-bit RISC processor.

Why Acorn designed its own CPU

Acorn made the BBC Micro, Britain's educational home computer, which ran a 6502 at 2 MHz. Looking for a processor for its next machine, the company was not happy with anything on the market. According to Furber's Computer History Museum oral history, co-founder Hermann Hauser brought in papers describing the Berkeley RISC and Stanford MIPS processors, and the idea of a reduced instruction set — making the processor much simpler — seemed to address their problem.

A trip in October 1983 mattered. Wilson and Furber visited the team in Phoenix, Arizona, that was designing a successor to the 6502 (the 65C816), and found a small group working out of a suburban bungalow, with local school kids helping on some of the cell design in their summer holidays. Furber names the visit as one of the major points in putting ARM on a formal footing.

The birth of ARM1

The method was close to handmade. Furber wrote a reference model of the instruction set in BBC BASIC and broke it into block specifications; the whole development environment, he says, was about 850 lines of BBC BASIC. The final checks on the design were done at VLSI Technology's offices in Munich in January 1985. From serious start to working silicon took about 18 months.

The team's fear was that RISC ideas were widely published, so some big company would build a commercial RISC processor first and trample them. When they had working silicon in April 1985, nobody else had.

Was low power an accident?

The designers tell the low-power story, the trait that later suited ARM to portable devices, slightly differently.

Furber says the goal was to fit a cheap plastic package, which meant staying under about a watt. The tools of the time could not predict power accurately, so the team used generous margins and ended up under a tenth of a watt. He describes the result as somewhat serendipitous but "not accidental".

Wilson's version, as told to The Register, is more dramatic. When the first chip went onto a development board, the meter in series with the power supply read zero: a fault meant no current was reaching the supply lines, and the processor was running on leakage from the logic circuits. The low power for which ARM is valued, she said, was "a complete accident".

The Evaluation System, then Archimedes

ARM1 never became a product chip. In a press release dated 7 July 1986, Acorn announced an evaluation system for the ARM, which it said it had first announced in August of the previous year. One version attached to the BBC Micro and Master Series, another to IBM PCs and compatibles; the BBC version cost £4,500 excluding VAT. The release described the ARM as a single-chip 32-bit RISC in low-power CMOS, averaging 4 MIPS on an 8 MHz clock with 44 basic instructions, and listed applications from communications and microcomputers to expert systems and laser printer controllers. VLSI Technology would manufacture and market the chip set.

What reached homes and schools was the Acorn Archimedes, with the second-generation ARM2; the first models were released in June 1987. Furber explains that the Archimedes chip set was the ARM2 plus three support chips, MEMC, VIDC and IOC. ARM had never been conceived as a single chip but as one of the four needed to build that machine.

Arm as a company

Saxby's model, licensing designs instead of making chips, put ARM into the silicon of many companies. In its May 2026 results Arm said more than 350 billion chips had shipped. The Japanese connection is SoftBank: it bought Arm in 2016, and in the 2023 Nasdaq listing a wholly owned SoftBank Group subsidiary was the selling shareholder.

Apple moved the Mac to its own ARM-based chips with the M1 in 2020. For the contrast with x86 see ARM vs x86 in the glossary; for an instruction set opened to everyone, see RISC-V. The machines of the period are on the hardware timeline.