TWENTY YEARS AGO Britain's computer industry looked like a has-been. The global IT industry, desperate for a standard platform for its software, had gravitated to the IBM PC architecture and most UK companies that had built computers to their own designs during the 1980s had either disappeared or were on their way out.
It seemed that the only way to survive, unless you were Apple with a large home market and an eye for design and marketing, was to build PC clones.
The Cambridge company Acorn, often called the British Apple because of a tendency to go its own way, had been in trouble as early as 1984 and would be gone before the turn of the century.
Yet out of this mess was born one of the most successful UK start-ups of the computer age. ARM, which celebrates its 20th anniversary this weekend on 27 November, first set up shop in a converted barn outside Cambridge to exploit Acorn's single greatest asset, the intellectual property bound up in its home-grown Acorn - now Advanced - Risc Machine processors.
The early signs were not good. The launch came at the start of a recession and ARM's funding, just $250,000 from VLSI and $1.5 million from Apple, was small change by the standards of the semiconductor industry. The company's first task was to tailor a processor chip for Apple's ground-breaking pen-driven Newton handheld, which promptly flopped.
Moreover ARM seemed superficially to epitomise the contempt for marketing that had bedevilled British engineering for decades. The original team consisted of 12 engineers with no commercial experience led by a former engineer who none of them knew.
Robin Saxby, who got his electronics degree at Liverpool University, was something of an outsider in elitist Cambridge, which was probably not a bad thing. He had done the rounds of the industry, working for Plessey, Pye, and Rank Bush Murphy, and getting his hands dirty, as some in Cambridge would say, as a salesman for Motorola.
It was Saxby who gave the company its global vision and the innovative licensing model under which it sold not physical silicon but designs for other companies to manufacture. This enabled ARM to compete in an industry in which company revenues - and capital requirements - could exceed the GDP of a small country.
Very early, and very cheekily, Saxby declared an ambition for the ARM architecture to become a global standard.
Hiring Saxby was ARM's second stroke of luck. Its first had come way back in the 1980s when Acorn, designing a cheap processor for the home computer market, used a RISC architecture because it needed the fewest transistors. The fact that this also meant that it drew less power than CISC architectures like Intel's x86 chips was not seen as important at the time - power consumption did not become a serious issue on desktops and servers until the turn of the millennium.
But it was already a very big issue in embedded microcontrollers and portable devices. Remarkably, an ARM analysis of market opportunities, dated three weeks after the founding of the company, does not mention the mobile phone, though it does refer to "portables". Back in 1990 only around 12 million people globally owned a mobile handset.
Two crucial events came in 1998, when ARM still had a smaller market share than its direct rivals MIPs and Hitachi. First Nokia launched its 6110 mobile phone, which used a Texas Instruments chip that married an ARM core to a digital signal processor, creating a breakthrough single chip baseband processor. ARM's chief technology officer Mike Muller recalled later that the 6110 produced a flood of inquiries from chipmakers that had never talked to ARM before.
Then ARM went public, freeing itself of the differing agendas of its backers and allowing it to present its products as a neutral platform for licensees who were competing among themselves. It encouraged manufacturers to use its cores in system-on-a-chip (SoC) designs by tailoring licences for the purpose and freely disclosing the architecture of the AMBA bus used to connect peripheral modules. It also kept a tight control over its instruction set, preventing the kind of fragmentation that hampered some rivals.
The industry began to converge on ARM chips as computing went mobile for much the same reasons that it had plumped for Intel x86 chips for personal computers when computing hit the desktop a decade earlier: nobody wanted a multiplicity of hardware platforms. ARM was becoming a de facto standard. Saxby, almost incredibly, had achieved his ambition.
ARM's success does not rest entirely on mobile phones. Intel likes to boast that it can offer an x86 chip to suit every class of computer from a netbook to a server. ARM offers a similar diversity of smaller devices, ranging up from microcontrollers the size of a pinhead. Licensees can maximise power efficiency by choosing a design that precisely fits the needs of an application.
ARM was bound sooner or later to go into direct competition with Intel as laptops got smaller and phones became more like portable computers, and the ARM-based SoCs in Apple's Ipads have helped raise its credibility in larger formats. Not that anyone should have been surprised. ARM chips were running graphical interfaces on Acorn's Archimedes desktops when Intel PCs were still tootling with text screens. Now multicore ARM chips are being touted as server processors and Intel is fine-tuning its x86 processors for traditional ARM markets.
ARM is still a minnow by comparison with Intel, which will spend more on R&D alone this year than its rival's $6 billion market capitalisation.
But ARM licensees - including giants like Samsung, Toshiba, TI, Freescale and Qualcomm - are collectively more than big enough to act as a counterweight to Intel and have bolstered AMD as its principal rival. The little company from Cambridge has helped shift computing's centre of gravity from Silicon Valley.
ARM has spread itself across the world and is a UK company only in that it retains strong roots here. But it shows what can be done by a small company from a relatively small country in a recession. µ
It's time for our regular two-step through the Google news
Bug bounty offer: accepted