The Silicon You're Using Was Built By Women You've Never Heard Of

Okay, let's talk silicon.

Not the think-piece kind where someone writes "we need more women in STEM" and calls it a day. I want to talk about the actual hardware — the ISAs, the VLSI methodologies, the architectural decisions — and who was in the room when they were made.

Because here's what my years in hardware QA taught me: attribution matters. When a thermal design fails, you trace it back to the engineer who signed off on the thermal pad spec. When a power delivery rail browns out under load, you find the engineer who signed off on the VRM margins. The work has a name. The failure has a name.

The same rule applies to breakthroughs. And right now, the industry has a serious attribution debt.

The Chip In Your Phone Has a Specific Origin Story

There's a strong chance you're running something ARM-based right now. Your phone, almost certainly. Your tablet, likely. Your laptop, if it's a recent Mac or a Snapdragon X machine. ARM is everywhere in consumer silicon.

The ARM instruction set architecture — the actual specification those chips implement — was designed primarily by Sophie Wilson.

Let that sit for a second.

Wilson, working at Acorn Computers in the 1980s alongside Steve Furber, designed the original ARM ISA from scratch. She wrote the instruction set. She wrote the first assembler. She ran the simulation that validated the design before a single physical chip was taped out. The architecture was so clean and so efficient that it became the dominant blueprint for mobile processors across the following four decades.

If you've praised Apple Silicon for its thermal efficiency and performance metrics, or applauded the battery life on any ARM-based laptop — you are downstream of Sophie Wilson's engineering work. Not metaphorically. Directly. The instruction set she designed is the one being executed.

Ask ten people to name the key architects behind ARM. Her name rarely comes first. (You're more likely to hear "ARM Holdings," which is a company, not a person.)

That's the attribution debt.

The Methodology Behind Modern Chips

Go back further. Before ARM, before the chip design tools we take for granted, building custom silicon required engineers doing manual layout — placing individual transistors and routing connections by hand on mylar sheets.

Lynn Conway changed that.

Close up of an engineer's hands working on complex mylar sheet routing diagrams for chip design in the 1970s.

Working at Xerox PARC in the 1970s, Conway co-developed the Mead-Conway VLSI design methodology — a set of scalable rules and abstractions that made chip design teachable and accessible to a far broader pool of engineers. This wasn't incremental. The textbook she co-authored with Carver Mead, Introduction to VLSI Systems (1980), became foundational curriculum for a generation of chip designers and is widely credited with enabling the proliferation of custom ASIC development.

Much of modern chip design pedagogy and tooling traces its lineage back to methodologies Conway helped establish — not all of it, but enough that her absence from the history most engineers receive is a real gap.

Conway did this while also navigating a career reconstruction after being fired by IBM in 1968 for transitioning — the company erased her from their records. She rebuilt under a different name, produced paradigm-changing work, and it took decades for the industry to formally recognize her contributions.

If the tech industry wants to talk about systemic barriers, that's a pretty good place to start.

The Turnaround Nobody Explains Correctly

More recently: Lisa Su saved AMD.

I'm not being hyperbolic. When Su became CEO in 2014, AMD was genuinely in crisis — burning cash, losing market share, their CPU architecture was being dismantled by Intel's Core series. The company had already spun off its manufacturing operations to GlobalFoundries. There were credible questions about whether AMD would survive the decade.

What followed was the Zen architecture — a clean-sheet CPU redesign that Su greenlit and pushed through. Zen launched in 2017 and was competitive at its price tier. Zen 2 in 2019 was dominant in most workloads and closed the gap with Intel across the stack. By 2021, AMD was winning in enough benchmark categories that Intel was forced into a significant architectural response.

The actual engineering was done by a large team — Jim Keller did significant architecture work on Zen 1 before departing — but Su made the strategic calls: commit to the clean-sheet redesign, target the segments where AMD could win on price/performance, stay with TSMC's external foundry model. That last call mattered. It kept AMD on leading-edge process nodes while Intel's internal fabs hit repeated delays on 10nm.

The result: AMD's stock rose from around $2 in early 2015 to a peak above $160 in late 2021. The CPU market shifted from near-monopoly to genuine competition across the mainstream and enthusiast tiers. Those are hardware outcomes worth examining on their merits, separate from who was running the company.

That's not a diversity win. That's a hardware turnaround win. Su happened to be the person who executed it.

The Structural Problem (And It's Not What The Ads Say)

Here's where I get specific about the failure mode I actually observed in hardware QA: homogeneous teams have systematic blind spots.

Not because of anything philosophical — because of how design review actually works. When everyone in the meeting shares the same background, training, and reference experiences, they share the same assumptions about what "normal use" looks like. Those assumptions get baked into thermal envelopes, grip geometry, and how edge cases get handled.

I've seen products ship with grip-force sensors calibrated on a narrow demographic sample. Display white-point tuning optimized for a test panel with minimal diversity. Voice recognition that worked acceptably for some accents and degraded significantly for others. (MIT Media Lab's 2018 "Gender Shades" study documented commercial facial recognition error rates up to 34% on darker-skinned women versus under 1% on lighter-skinned men — the failure mode doesn't stop at face unlock.)

These aren't malicious decisions. They're the consequence of nobody in the review room flagging the problem because nobody in the room experienced it as a problem. This is what happens when engineering teams don't scrutinize their design assumptions hard enough before shipping.

The diversity argument for hardware isn't primarily a values argument. It's an engineering quality argument. More diverse review teams surface more failure modes before the product ships.

On the retention side: McKinsey's Women in the Workplace research has consistently identified what they call the "broken rung" — women are promoted from individual contributor to manager at lower rates than men with equivalent performance reviews, and this is where the pipeline numbers actually fall off. Women are not absent from CS and EE graduation pipelines; they represent roughly 20% of CS bachelor's degrees in the US currently (down from around 37% in the mid-1980s, per NCWIT data). The differential attrition is mid-career, right before the senior engineering and hiring-influence levels. That pattern points to structural retention problems, not supply.

The "pipeline problem" framing lets companies off the hook — we'd hire them if they existed. The data doesn't support that framing.

What International Women's Day Actually Calls For

Every year, the "Women in Tech" content cycle runs its course. Inspirational quotes. Lists of "women to watch." Articles about pipeline programs.

Here's my list, which is shorter:

Name the work correctly. When you describe ARM, include Sophie Wilson. When you teach VLSI design history, include Lynn Conway. When you cover the AMD turnaround, include the specific calls Lisa Su made and why they were right. Attribution isn't a courtesy — it's accuracy. The historical record has gaps that need correcting. (This goes beyond the usual tech industry mythology we all absorb without questioning.)

Fix the mid-career retention problem. This is where the numbers fall off. Parental leave policies, flexible work arrangements that don't quietly tank your performance review, promotion criteria that don't reward the kind of constant availability that's harder to maintain when you're the primary caregiver at home. Engineering orgs that have actually fixed this stop losing senior women.

Hold the hardware accountable. The next time a product ships with a face-unlock system that performs poorly on darker skin tones, or voice assistant accuracy that degrades by accent, trace it back: who was in the design review? Were those failure modes surfaced? The product is evidence of the team's blind spots.

The Verdict For Your Wallet

The silicon in your phone, laptop, and tablet runs on architectural work and methodology developed by women who didn't receive the credit their male counterparts would have received for equivalent contributions. That's documented history with direct implications for how we assess "who built tech."

The hardware industry produces better products when it stops losing engineering talent at the mid-career transition. That's self-interest dressed as principle, and I'll take it. The case for fixing retention doesn't require moral persuasion — it requires wanting fewer systematic failure modes in shipping products.

But also: know the names. Sophie Wilson. Lynn Conway. Lisa Su. Radia Perlman (spanning-tree protocol — the thing that keeps your ethernet network from collapsing under broadcast loops). Frances Allen (compiler optimization theory, IBM Fellow, first woman to win the Turing Award, in 2006). The work is documented. The attribution is overdue.

Stay wired.