How ARM's low-power chip architecture came to power every smartphone

Executive overview

Every smartphone on earth runs on ARM processors, yet ARM manufactures zero chips and often doesn't design them either. The company traces back to a BBC government education initiative and a rejected Intel order.

ARM's RISC architecture used 30,000 transistors where Intel used 130,000 — achieving better real-world performance at a fraction of the cost and power draw. When Apple needed a chip for the Newton PDA in 1990, they co-founded ARM as a joint venture with Acorn, seeding the company that would later save Apple from bankruptcy.

SoftBank's $32 billion acquisition in 2016 — bought in two weeks after a dinner conversation — became the template for the entire Vision Fund strategy.

The core insight: ARM succeeded not by competing in existing markets but by owning the architectural standard for the next computing wave before that wave arrived.

From BBC micro to RISC breakthrough

• The BBC Computer Literacy Project (1980) contracted Cambridge-based Acorn to put personal computers in UK schools; Acorn shipped 1.5 million BBC Micros.
• IBM's 1981 PC launch signalled a professional computing wave; Acorn needed a more powerful chip for their next machine, the Archimedes.
• Intel refused to supply Acorn with the 286 processor — one of the worst business decisions in computing history.
• Acorn engineer Sophie Wilson, drawing on a UC Berkeley RISC paper, designed a processor from scratch: the Acorn RISC Machine (ARM).
• The ARM chip had ~30,000 transistors vs Intel 286's ~130,000, yet outperformed it on the 80% of instructions executed most frequently.
• A side effect: so few transistors that the chip sometimes ran on power leaking from nearby components — ARM was accidentally ultra-low-power.

Why ARM didn't win the PC era

• By the time Archimedes launched (1987), DOS ran only on x86; every business application was written for Wintel.
• The compiler stack was tightly coupled to x86 — porting an OS to a different architecture was considered near-impossible.
• Acorn was acquired by Italian firm Olivetti and exited the PC business; the RISC technology would have died there.

The Apple Newton joint venture

• Acorn co-founder Herman Hauser foresaw mobile computing in 1988 and started Active Book Company to build PDAs; he asked Sophie Wilson's team to re-optimise the chip for battery life.
• Apple, under John Sculley, was pursuing the same PDA vision with the Newton project; their first chip candidate, AT&T's "Hobbit," was a failure.
• VLSI (Acorn's silicon fabricator) introduced Apple's Larry Tesler to the ARM chip; he recognised it was exactly what Newton needed.
• In 1990, Apple, Acorn/Olivetti, and VLSI spun ARM out into an independent company in six weeks. Apple invested $1.5M for 43% equity.
• 12 engineers set up in a converted barn in Cambridge; Robin Saxby (ex-Motorola) was brought in as CEO.
• The ARM 610 powered every Apple Newton. The Newton failed commercially — but ARM survived.

The business model that changed everything

• With Newton sales too small to sustain the company, CEO Robin Saxby pivoted to a three-layer licensing model:
  1. Upfront license fee for rights to the ARM instruction set architecture.
  2. Engineering services fee for embedding ARM teams with chip designers.
  3. Per-unit royalty on every device shipped containing ARM technology.
• This aligned ARM's financial incentives with partner success: the more units shipped, the more ARM earned.
• The model also enabled systems-on-a-chip (SoC) — ARM would help partners integrate processor, graphics, and other functions onto one piece of silicon, rather than requiring separate cards and buses.
• In 1993, ARM signed a landmark deal with Texas Instruments to power the Nokia 6110 — the first major consumer GSM phone. Mobile was on.

ARM's IPO and saving Apple

• By end of 1997: ARM was profitable with over £25M revenue.
• April 1998: dual-listed on London Stock Exchange and NASDAQ at £5.75/share; market cap £264M.
• Apple still held its ARM equity from the 1990 JV. As the ARM stock price soared during the tech bubble, Apple liquidated its stake over several years — generating $792M in profit.
• That $792M is widely considered what kept Apple solvent while Steve Jobs restructured the product line.
• Post-bubble: ARM recovered quickly; by 2010, ARM partners were shipping over 5 billion devices per year.

Scale and the mobile duopoly lock-in

• 130 billion ARM chips shipped to date.
• ARM holds ~96% market share across smartphones and embedded devices.
• iOS and Android both run on ARM's RISC architecture; re-architecting either for a new ISA would be an undertaking comparable to what Apple did moving Macs from PowerPC to Intel.
• Microsoft missed mobile partly because Windows and DOS were so deeply coupled to x86 CISC that porting was never viable during the smartphone ramp.
• Apple was uniquely positioned to make the iOS/ARM transition because they had already ported Mac OS twice (PowerPC → Intel) and carried Newton DNA.

SoftBank's $32 billion acquisition

• Summer 2016: ARM CEO Simon Segars attended a dinner at Masayoshi Son's Woodside estate; Masa asked him what ARM would do with unlimited capital.
• Days later, Masa summoned Segars and ARM Chairman Stuart Chambers (pulled off a yacht in Turkey) to a private restaurant in Marmaris.
• Masa offered $32 billion — a ~43% premium to market price — and gave them two weeks to decide.
• Board approval and announcement came within two weeks. ARM remained an independent division inside SoftBank.
• Masa funded the deal by selling $10B in Alibaba shares, $7B in Supercell shares, and taking a $9B loan. He timed the purchase to the post-Brexit pound weakness.
• The next year, SoftBank corporate sold a 25% stake in ARM to the Vision Fund at cost ($8B), establishing ARM as the Vision Fund's anchor and prototype investment.

What SoftBank ownership enabled and the open questions

• ARM headcount grew from ~4,000 to over 6,000 in two and a half years — impossible as a public company posting net losses.
• Operating margin fell from 52% to 24% as revenue grew from $1.6B to $1.8B; SoftBank is absorbing losses to fund growth investment.
• ARM is investing heavily in AI chips and autonomous vehicle silicon (competing in the space Intel entered with the $17B Mobileye acquisition).
• The original IOT thesis has shifted; the current investment focus is connected cars and AI inference at the edge.
• Bull case: ambient computing (wearables, sensors, connected devices everywhere) means the per-unit royalty base grows by an order of magnitude.
• Bear case: royalty revenue per unit shrinks as cheap sensors dominate shipments, and ARM's share of value creation remains small relative to device makers.
• A strategic re-IPO of the Vision Fund's ARM stake is the most likely exit, though the 12-year fund clock is running.

Why no one else could have bought ARM

• Selling to Apple, Qualcomm, or any chip partner would have triggered ecosystem panic and broken the neutral-Switzerland value of the ARM standard.
• Very few entities could write a $32B cheque without corrupting the asset they were buying.
• SoftBank's structure — willing to absorb years of losses in exchange for long-run upside — matched precisely what ARM needed to invest in the next computing wave.