Q-Semicon 2: An ode to Lisa Su

How did we get here?

In my last post in this series, I made the case that Jensen Huang, the CEO of Nvidia is uniquely positioned to calm the tensions in the global semiconductor supply chain today.

But first, a quick history lesson!

In supply chains, secondary sourcing is a basic and crucial strategy to mitigate against problems arising in the primary supply. So, how did the U.S. end up wholly reliant on Taiwan for advanced chip manufacturing in 2020?

It wasn’t always this way.

Andy Grove, Intel’s third employee and CEO from 1987-1998, is revered in technology circles. His books on management have inspired many of the valley elite, and he is credited as the Father of OKR style management. He was, by all accounts a legendary engineer.

Intel’s CEO today, Bob Swan, was previously its CFO, and has not held an engineering position.

Intel’s fate has always been closely tied to another company - Advanced Micro Devices, or AMD. AMD was created by a group of employees - led by Jerry Sanders - who departed Fairchild semiconductor less than a year after Intel’s founders did the same.

In its early years, AMD manufactured logic and flash devices, not microprocessors. When IBM decided to enter the Personal Computer (PC) market in 1982, they procured microprocessors from Intel, with AMD designated as their second source. AMD was to build these processors to comply with Intel’s x86 architecture specification, to ensure software (Microsoft DOS) only needed to be written once and be compatible with both designs.

Thus began a ferocious rivalry that continues to this day. AMD, the perennial David, aiming for Intel the Goliath. Both companies would tussle for the title of fastest processor on the market, but Intel kneecapped AMD at a crucial time when PC adoption was booming. AMD attempted to come from behind with a fully in-house design, dubbed the K5; however, this project was plagued with delays that left AMD with a hole in its product line up. To ensure continued roadmap viability, AMD acquired NexGen. And then, to enter the graphics card market for video games, AMD spent a mind-boggling (at the time) $5.4B to acquire ATI. It then proceeded to give away the keys to the mobile processor kingdom to Qualcomm.

Even amongst all of this, the Achilles heel for AMD was always manufacturing. To ensure the unit economics make sense for a fabrication facility, there needs to be

a) consistent demand to keep the plant equipment busy producing round-the-clock

and

b) sufficient cash generated to enable continued investment in R&D, to ensure the facility can compete on the leading edge of manufacturing technology

So, you can see how product delays, near-cost pricing to overcome its second-best market status, and a huge pile of debt from acquisitions left AMD without the capital to invest in its plants.

In 2008, AMD was forced to spin off its manufacturing division to a U.A.E-led consortium, called GlobalFoundries. GlobalFoundries promptly fell behind, announcing it would cease bleeding edge manufacturing due to escalating R&D costs. It then became somewhat of a patent troll. AMD was a far cry from original CEO Jerry Sanders’ claim that “Real men have fabs”.

The third player in our story - IBM - was once one of the largest manufacturers of chips as well. IBM’s TJ Watson Research Center used to be a premier institute, with graduating engineers making a beeline for working there. The IBM, Motorola and Apple co-designed processor - codenamed Somerset - used to power the old school Macs. However, just like AMD, it too was unable to maintain sufficient demand - nor withstand pressure from Wall Street - to justify continuing to invest in manufacturing. In 2014, a couple of years after Ginni Rometty took over as CEO, IBM pivoted towards software and services, and literally had to pay Global Foundries $1.5B to take semiconductor manufacturing off its hands. Platformonomics has a damning set of charts capturing IBM’s last decade.

IBM’s offices in Austin, Texas is now an outdoor shopping mall.

And thus, suddenly in 2018, the U.S. Department of Defense found itself without a trusted foundry to manufacture the most advanced processors on U.S. soil.

5 years ago, AMD’s stock price was hovering below $2. It was in danger of getting delisted from the NASDAQ. A parade of CEOs had tried and failed to revive its fortunes. It had been forced to sell its Austin campus and lease it back, as a cash generating measure.

Since 2015, AMD’s stock price has gone up 41x, far eclipsing the 20x return on Nvidia’s stock in the same time frame.

How did it achieve this miraculous growth? The answer is Dr. Lisa Su. An MIT PhD, last year Dr. Su was the highest paid CEO in the S&P 500. Let me make that clear - not just the highest paid female CEO, the highest paid CEO of any gender.

Lisa Su is my queen.

In speaking with many who have worked for her now and in the past, it becomes clear that the key to Dr. Su’s success is re-focusing AMD on its core competencies in CPUs and graphics cards, and reinvigorating the engineering-first culture. She jettisoned distractions, even paying GlobalFoundries $325M to get out of a sweetheart deal so it could manufacture its designs with TSMC.

(Yes, it looks like GlobalFoundries keeps making out like bandits, so it is confounding that they cannot keep up on advanced manufacturing.)

When Samsung’s Austin design team was let go of last year, several of my ex-colleagues decided to go work at AMD over higher paying salaries at other companies, simply because they were more aligned with the mission and leadership at AMD.

Dr. Su is a huge inspiration to women in technology like me, and even to women outside our industry.

Side note: here are some AMAZING technical women I have worked with in my career, both mid-level and senior.

Even Jensen expresses admiration for her intense focus on execution; takes one fantastic executor to know another. Today, Nvidia and AMD are frenemies, competing in the graphics market on equal footing, and collaborating in the datacenter market, with AMD’s chips feeding data to Nvidia’s DGX deep learning accelerators. Nvidia picked AMD over Intel due to the higher core and memory channel count, and faster PCI Express lanes than anything Intel has to offer.

How does Dr. Su play a role in relieving tensions with China? In 2014, when AMD was desperate for a cash infusion, it set up a joint venture with a company called Sugon - a leading Chinese supercomputer developer - to license AMD’s processor technology. The $293M that AMD earned from the THATIC JV was the key deal to its turnaround in the last 5 years.

This JV is quite the complicated structure.

Source: https://wccftech.com/no-amd-did-not-sell-keys-kingdom-how-thatic-jv-works/

By the time the U.S added THATIC to its entity list last year, AMD claimed it had concluded the partnership.

So I posit that Dr. Su, also of Taiwanese descent, CEO of a resurgent U.S. technology company, with extensive experience doing complicated deals in China will play a crucial role in relieving tensions in the semiconductor industry.

(Obligatory: I am not a fiduciary and these are just my opinions. I’d like to thank the many folks who took time to speak with me as I researched this! In the next post, I’ll present some ideas on how we can make forward progress as an industry.)

Some other interesting things I found in my research:

  1. Jensen’s kitchen has become quite the sensation as he’s conducted keynotes from it during the pandemic. Nvidia made a high-res version of his kitchen available as a Zoom background!

  2. I don’t think you need to conduct any fancy sentiment analysis to understand how fanboy consumers feel about Nvidia (508K subs) and AMD(617K subs) vs. Intel (140K subs).

  3. Dr. Su’s keynote at Hot Chips (where I am on the organizing committee) 2019 is an excellent summary of the state of our industry.