The company that armed every competitor....and dominated them all

Welcome to Legacy Beyond Profits, where we explore what it really means to build a business that leaves a mark for the right reasons.

February 1991. Robin Saxby stood in a converted turkey barn outside Cambridge, England, facing twelve engineers and a decision that would define the future of computing. Around him, the semiconductor industry ran on vertical integration—Intel designed chips AND manufactured them in billion-dollar fabs. Merge with an established chip manufacturer? Build fabrication plants? The path seemed obvious.

Saxby chose the opposite. ARM would manufacture nothing. The company would design processor architectures, document them meticulously, then license blueprints to anyone willing to pay. Competitors would build the actual chips while ARM collected royalties—forever. His venture capital friends called it commercial suicide: "Joint ventures never work." Industry analysts predicted failure within 18 months. Thirty-four years later, ARM controls 99% of smartphone processors, has shipped 300+ billion chips, and generates 97.7% gross margins without owning a single fabrication plant.

From manufacturing power to architectural control

1.Design vs. Manufacture: The counterintuitive bet that neither was necessary

When Texas Instruments approached ARM in 1993, they needed low-power mobile processors but faced $24 million development costs and 18-24 month timelines. ARM offered an alternative: license our reference design for $3 million upfront plus 1.5% royalties per chip sold.

ARM had spent three years optimizing their architecture. For them, the same design could be sold to dozens of companies simultaneously. ARM invested 44.5% of revenue in R&D, amortizing those costs across hundreds of licensees. Samsung paid $5 million for the same architecture Apple licensed. Every customer strengthened the platform while ARM avoided taking sides.

2.The Switzerland Strategy: Profiting from every war

Compare the margins: ARM at 97.7%, Intel at 41%, AMD at 47%. ARM's profits dwarf competitors who actually manufacture chips.

ARM-based chips run applications faster than competing architectures not because transistors are superior but because three decades of compiler optimizations extract more performance from the same silicon. Apple designs custom processors for iPhones. Google created Tensor. Amazon built Graviton. Microsoft developed Azure Cobalt. Every custom chip licenses ARM's instruction set and pays royalties. Companies spending billions to escape ARM's platform cannot actually escape.

3.When your defining advantage arrives by mistake

April 1985. ARM's development team connected power meters to their first working silicon prototype. The needle didn't move. The processor appeared to consume zero power.

Investigation revealed a wiring fault: the chip wasn't receiving external power. ARM's processor was running entirely on leakage current, approximately 0.1 watts, orders of magnitude below competing designs. The team had optimized for simplicity and cost, not battery efficiency. As co-founder John Biggs admitted: "The low-power thing was a happy accident, which of course became important for handheld devices with limited battery life, but it wasn't initially the prime concern."

This accidental efficiency became ARM's defining characteristic. Nokia's 6110 ran on ARM processors enabling multi-day battery life. Smartphone manufacturers specified ARM because battery life was non-negotiable. Software developers optimized for ARM because that's where the market existed. By the time Intel recognized mobile computing's importance, ARM's 22-million-developer community made competition futile.

4.How commodity hardware generates software margins

ARM maintained strict neutrality by selling identical technology to bitter rivals. Apple and Samsung licensed the same architecture for competing smartphones. When hyperscalers designed custom ARM processors to escape vendor dependencies, their engineering investments strengthened ARM's position.

By 2025, barriers to entry had become insurmountable through network effects. Twenty-two million developers worked in ARM environments. University programs taught ARM because free resources existed. Startups specified ARM because their teams knew no alternatives. The switching costs made changing architectures economically irrational.

How ARM built a $133B platform by manufacturing nothing

Robin Saxby's first executive decision at ARM required immediate action. The company had twelve engineers in a cramped office in a converted agricultural building. Financial resources: $3 million from Apple, development tools from VLSI, and Acorn's intellectual property.

Saxby interviewed the team at a pub, arriving ten minutes early. When team members arrived on time, he opened with: "You're four minutes late. Another minute and I'd have gone." The message wasn't about punctuality but precision. ARM couldn't compete through capital intensity. Competitive advantage would come from disciplined execution.

He faced three options. Merge with an established semiconductor manufacturer. Partner exclusively with Apple. Or pursue pure IP licensing.

Industry advisors pushed merger or partnership. One venture capital contact stated: "Joint ventures never work." The pure licensing model seemed riskier. Why would semiconductor giants pay for designs they could develop internally?

Saxby committed to licensing. ARM would design processor instruction sets, then allow semiconductor manufacturers to build actual chips. The company would collect upfront fees ($1-10 million) plus ongoing royalties (1-2% of chip selling price). Every unit shipped would generate revenue forever. No fabrication costs. No inventory risk.

The crucial test came in 1993. Apple launched the Newton MessagePad. The device became late-night comedy material due to handwriting recognition failures. Apple sold 50,000 units in three months against projections of one million. Industry observers declared ARM's business model dead.

Saxby gathered the twelve engineers. "Newton's market performance doesn't invalidate our architecture," he told them, though privately he understood the stakes. Apple might terminate the partnership. "Our processor works exactly as designed. If the product fails, that's a software problem, not a silicon problem."

The team knew Saxby was right technically but wrong strategically. In semiconductors, perception often mattered more than reality. Then, six weeks later, Texas Instruments called.

TI engineers noticed something market analysts missed: ARM's processor ran efficiently despite Newton's software problems. TI faced $24 million per design in development costs with 18-24 month timelines. ARM offered proven architecture for $3 million upfront, 12-month time-to-market, plus royalties.

Texas Instruments became the third licensee. More importantly, TI's validation attracted Samsung. After four meetings, Samsung signed. Sharp followed for Nintendo processors. The pattern emerged: licensing ARM designs saved tens of millions in R&D while reducing development risk.

Jamie Urquhart, one of the original twelve engineers whom Saxby convinced to become VP of Marketing despite zero marketing experience, reflected: "The original strategic plan was to gain a semiconductor licensee in the three major regions: North America, Europe, and the Far East." The geographic diversification insulated ARM from regional downturns while expanding the partner network exponentially.

The Apple connection created ultimate validation through ironic reversal. In 1997, Steve Jobs returned to a nearly bankrupt Apple. The company was 90 days from declaring bankruptcy. Jobs needed liquidity beyond Microsoft's $150 million investment. Between 1997 and 1999, Apple liquidated its ARM stake, retaining 14.8% ownership and realizing $800-$1.1 billion.

That 36,600% return on the original $3 million gave Jobs runway to develop the iMac. The deeper irony emerged a decade later. In 2005, Apple approached Intel to manufacture processors for iPhone. Intel declined. Mobile chips seemed like low-margin distraction.

Apple turned to ARM architecture. Every iPhone since 2007 has run on ARM-based processors. Initially designed by Samsung, later by Apple's internal team, but always licensing ARM's instruction set. Intel's refusal became one of technology's most expensive strategic errors.

The compounding effects revealed themselves through ecosystem expansion. Nokia specified ARM processors, creating volume that attracted developers. Developers optimizing for ARM made the architecture more valuable to manufacturers. Increased volume reduced per-unit costs. Lower prices expanded markets. By 2012, ARM controlled 95% of chips in smartphones and tablets. Microsoft, Google, Amazon, and Meta all tried to escape and strengthened ARM instead.

By fiscal 2025, the metrics spoke definitively: $4.007 billion revenue, 97.7% gross margins, 99% smartphone market share, 300+ billion chips shipped, 22 million developers, 70% of global population using ARM-powered devices daily. The company that manufactured nothing created computing's most pervasive platform.

Robin Saxby, knighted in 2002, proved that in markets where technical capabilities commoditize, permanent advantages belong to whoever controls the invisible instruction set that entire industries cannot function without.

From strategy to legacy

Architectural licensing demolishes the assumption that competitive advantages require manufacturing control. While competitors hoarded advantages through capital intensity and vertical integration, Saxby made ARM's designs universally accessible, then controlled the only platform where that accessibility created value.

This pattern transcends semiconductors. AWS doesn't dominate cloud computing through better servers. They dominate because millions of developers built careers around AWS-specific services. Tony's Chocolonely transformed chocolate not through superior cocoa but through transparent supply chains that competitors cannot replicate without abandoning existing relationships.

Saxby's genius wasn't recognizing that software beats hardware. Every tech CEO understood that by 2000. His insight was simpler: in markets where everyone eventually builds similar products, permanent control belongs to whoever owns the language those products speak. ARM doesn't charge for silicon performance. They charge for three decades of developer fluency. That vocabulary cannot be unlearned.