#923 1990 · ARM (Advanced RISC Machines Ltd.) · Semiconductor design
ARM gave up building chips and licensed its blueprints to would-be competitors instead
问题
A small chip-architecture spinout had no fab and no marketing budget to out-manufacture or out-market Intel
背景
Acorn Computers' engineers had designed the ARM1, an efficient RISC processor, in 1985, but Acorn's parent Olivetti lacked the capital to commercialize it at scale. In November 1990, Acorn, Apple and VLSI Technology spun the architecture out as a joint venture, Advanced RISC Machines Ltd., with Apple investing to secure a low-power chip for its upcoming Newton handheld. Building and selling ARM's own branded chips the way Intel did would have required fabrication plants costing hundreds of millions of dollars that the tiny new company simply did not have, and even with a fab, ARM would have been fighting Intel's manufacturing scale and x86 software ecosystem directly.
The Newton, ARM's first real commercial showcase, launched in 1993 and sold poorly, and CEO Robin Saxby concluded that ARM could not survive as a company betting its future on the success of any single product or partner. It needed a way to earn revenue from the architecture regardless of which specific devices succeeded or failed in the market.
换别人会怎么做
The available path for a new chip-architecture company was to raise enough capital to build or contract fabrication capacity and compete as a branded chip manufacturer, the model every incumbent from Intel to Motorola already used and had scale advantages in.
他们看到了什么
ARM's asset was not a chip, it was a reusable blueprint whose value did not depend on ARM building it. Licensing it for a royalty turned potential manufacturing rivals into paying customers instead.
那一手
Rather than manufacture chips, ARM licensed its processor designs to semiconductor companies for an upfront fee plus a small royalty on every chip a licensee actually shipped, letting Texas Instruments, Samsung, Qualcomm and dozens of others build their own ARM-based chips, customized for their own products, while ARM itself never had to build a fab or bet its survival on any one customer's success.
为什么管用
The mechanism works because it decouples ARM's revenue from its own manufacturing or marketing execution entirely — ARM earns whether Apple's Newton fails or a licensee's phone chip succeeds, as long as some ARM-based product somewhere sells, which diversifies risk across an entire industry's worth of bets instead of concentrating it on ARM's own product decisions. It also lets licensees customize the architecture for their own markets, so ARM chips end up everywhere from washing machines to smartphones without ARM having to design or sell into any of those markets itself.
值了多少
Over 250 billion ARM-based chips have shipped since, powering most of the world's smartphones without ARM owning a single fab.
什么时候会失灵
Pure IP licensing only works where the design itself, not proprietary manufacturing know-how, is the scarce and valuable asset, and where licensees have both the capital and expertise to fabricate competently on their own. It also concentrates ARM's fate on the health of its licensees' businesses collectively — a broad enough downturn in the chip industry hits ARM's royalties even though ARM never carried any of the manufacturing capital risk itself, and it leaves ARM dependent on licensees continuing to value neutrality over building competing in-house architectures.
后来呢
ARM's pure-licensing model became the template for the modern fabless-and-IP-licensing structure of the chip industry, proving that owning an architecture's design rights could be more durable and scalable than owning the factories that build it, a lesson later echoed by companies from Imagination Technologies to RISC-V's open-license approach.
资料来源
- [1]The Official History of ArmArm Newsroom, 2023newsroom.arm.com