원문 보기
In October 2014, a chip company called AMD was worth about $2 billion. Intel, its main rival, was worth around $180 billion, 90 times bigger. Analysts were openly using the word bankruptcy. The company had just cut about a quarter of its staff. And then, a soft-spoken engineer took over as chief executive and told people it was her dream job.
Today, AMD is worth close to $900 billion. It passed Intel for the first time in history in 2022. And the engineer, Lisa Su, has become one of the most quietly ruthless operators in modern business. This is a story about a comeback, but it is not a story about luck. And it is not a story about a genius having one big idea.
It is about somebody picking one extremely hard problem and refusing to blink for 10 years, while the biggest company in the industry assumed its lead was permanent. So, let's go through how she actually did it. And then, let's be honest about the parts of this story that people usually skip, including one that has changed completely in the last 12 months.
First, understand how bad it really was. Lisa Su became chief executive of AMD on the 8th of October, 2014. On her first day, the stock traded a little over $3 a share. The company was losing on every front at once. Its share of the market for x86 processors, the chips inside most laptops and servers, had collapsed to below 20%.
Intel controlled roughly 90% of the server market. AMD's flagship processor architecture, Bulldozer, had launched in 2011 and been a commercial disappointment. Revenue was shrinking. Debt was around $2 billion. And here is the detail most versions of this story leave out, because it spoils the shape of the comeback.
It got worse before it got better. Over the following year, the stock fell further, touching roughly a $1.60 in 2015. She took the job at the bottom and then watched it go lower. Now, who takes that job? Sue has a PhD from MIT in electrical engineering and her doctoral work was on transistor physics.
That is the actual physics of how a switch inside a chip works. She then spent about 13 years at IBM building semiconductors before stints at Texas Instruments and Freescale. She joined AMD in 2012 and was promoted to chief executive 2 years later. The first woman to lead the company. She would later become the first woman to receive the IEEE's Robert Noyce Medal in 2021.
So, the person who took over was not a turnaround specialist or a finance executive brought in to manage a decline. She was a chip engineer who had been building the things since her 20s. That matters for everything that follows. Her plan was not complicated. It was just unusually disciplined. Rule one, stop doing things you cannot win.
AMD had spread itself across PCs, graphics, embedded chips, and low-power mobile and it was losing in most of them. Sue narrowed it down to markets where AMD could realistically be first or second and put the weight behind one. High-performance computing. Rule two, be honest about the timeline. She told customers that a serious product was several years away and published a multi-year roadmap saying exactly what would arrive and when.
That sounds like basic management. In this industry, it was not. Roadmaps in semiconductors slip constantly. The unusual thing about Sue's was that AMD kept hitting it quarter after quarter and after a few years, customers started planning their own products around it. Trust in a roadmap is not a soft thing.
It is the reason a cloud company will commit to your chip 2 years before it exists. Rule three. And this is the one the story hangs on. Put nearly everything into one new architecture. So, here is the mechanism. For decades, a processor was etched onto a single piece of silicon. One slab, every component on it.
The problem is that the bigger that slab gets, the more likely it is that a tiny defect ruins the entire chip. Yields fall, costs climb, and it gets harder every generation. AMD's engineers went a different way. Instead of one big piece of silicon, they built the processor out of several small ones and fused them together.
The industry calls them chiplets. Think about why that is clever. Small pieces of silicon have far fewer defects than large ones. So, more of what you make is usable. You can mix and match them to build a cheap desktop chip or an enormous server chip from the same parts. And you can make the different pieces on different manufacturing processes,
using the newest and most expensive one only where it actually helps. The engineers named the new architecture Zen because it was designed around a balance between power efficiency and performance. And the decisive part was where those chiplets got made. AMD did not own advanced factories. So, it bought manufacturing from TSMC in Taiwan, which meant AMD could simply use whoever had the best process available.
Intel built its own chips in its own factories. That had always been its greatest strength. It was about to become its greatest weakness. Zen reached the market in March 2017 as the Ryzen line of processors. And it worked. Ryzen benchmarked within a few percent of Intel's best desktop chips at noticeably lower prices.
It was AMD's first genuinely competitive high-end processor in about a decade. Then came the part that actually pays. In servers, AMD launched EPYC and started taking data center share off Intel, which is the most profitable business in the industry. Console chips for PlayStation and Xbox brought in steady revenue that funded the rest.
In 2018, AMD was the best performing stock in the entire S&P 500. And now the other half of the story, because leaving it out would make this a worse video and a less accurate one. Sue had help. Intel tripped over its own feet. Intel had run for years on what it called tick-tock. Shrink the manufacturing process, then refine the design, then shrink again on a predictable two-year rhythm.
Moving from 14-nanometer manufacturing to 10-nanometer was supposed to be the next step. It took years longer than planned. Intel's own factories, the thing that had made it untouchable, became the bottleneck. The company was stuck selling refinements of the same process generation while the competition moved on.
It got so bad that Intel publicly apologized to customers for processor shortages. So, look at what was happening in parallel. Intel was frozen on its own manufacturing. AMD had no manufacturing to be frozen by and was buying the most advanced process in the world from TSMC. The outcome is almost unfair.
For a stretch of years, the smaller company was shipping chips built on better manufacturing than the largest chipmaker on Earth. AMD did not have to beat Intel at Intel's own game. Intel stopped being able to play it. Then the results compounded. In October 2020, Sue agreed to buy Xilinx, which makes programmable chips in an all-stock deal.
By the time it closed on the 14th of February, 2022, AMD's own share price had risen so much that the deal was worth about $49 billion, the largest semiconductor acquisition in history at that point. The day after it closed, on the 15th of February, 2022, AMD's market value passed Intel's for the first time ever, just barely.
AMD, at roughly $198 billion, Intel, a fraction below it. By 2024, AMD's annual revenue reached about $25.8 billion, a record, with the data center business nearly doubling. That December, Time named Lisa Su its CEO of the year. And in that same year, Intel's market value fell below $100 billion, its first time under that mark since before its dot-com peak.
In a single decade, the two companies had swapped positions completely. And then the AI boom arrived, and the chart stopped looking like a business, and started looking like a rocket. In June, 2026, AMD closed at an all-time high of just under $581 a share. As of this week, the company is worth close to $900 billion.
Now, let's be precise about the returns, because two very different numbers get used interchangeably here, and they are not the same thing. The share price has gone from a little over $3 to around 550. That is roughly 170 times your money. $10,000 invested on her first day would be worth well over 1 and 1/2 million today.
The company's total value has grown by considerably more than that, from around 2 billion to close to 900 billion, measured as an annual return to shareholders across her first decade, that works out at roughly 46% a year against about 13% for the S&P 500 over the same stretch. Those two multiples disagree, and the reason is simple.
AMD issued a large number of new shares to pay for Xilinx. So, the company got bigger by more than the individual shares did. If you see a single number quoted for this comeback, check which one it is because the gap between them is not small. And now the part that is genuinely different from how this story was being told even a year ago.
The first thing is that Intel is not a corpse. It is having its own comeback. A new chief executive, Lip Bu Tan, took over in 2025. The US government took a stake of around 10% in the company, and the stock has been extraordinary. Intel is up more than 200% across 2026, which is roughly double AMD's gain over the same period.
Intel has climbed back past 300 billion dollars in value for the first time since the dot-com era, and now sits somewhere in the region of 500 billion. Read that again because it reframes everything. In the year we are talking about, the giant that got beaten has been the better investment. The second thing is that AMD's next fight is much harder than the last one.
In artificial intelligence chips, it is not the leader. It is the challenger, and the company it is chasing is Nvidia, which is vastly larger and got there first. To be fair, the wins are real. AMD's Instinct accelerators have landed a reported agreement with Meta covering up to 6 gigawatts of capacity and a partnership with OpenAI.
Its data center processor revenue has set records for five straight quarters, and it is still taking server share off Intel. But, the gap is a different kind of gap. SU beat an incumbent that had stopped running. Nvidia is sprinting, and the third thing is that semiconductors are brutally cyclical. Export restrictions, demand swings, and political decisions move these companies enormously.
A chart that goes vertical can go the other way. So, the honest version is not that AMD conquered the industry. It is that the underdog caught the king, and the king got back up, which brings me to why I think this story is worth your time anyway. Intel did not lose to a flashier company. It did not lose to a bigger checkbook, because for that entire decade, AMD did not have one.
It lost to an engineer who picked one genuinely hard problem, told everyone exactly what she was going to do about it, and then did that same thing for 10 years without deviating, while the industry leader assumed its position was structural, rather than earned. There is a version of this where she got lucky with Intel's stumble.
But, Intel's stumble only mattered because somebody had spent years building the thing that was ready to walk through the door the moment it opened. So, here is what I would leave you with. Every dominant company believes its lead comes from something permanent. Intel thought its advantage was owning the factories.
That was true, right up until it was the reason it fell behind. So, what is the thing your industry treats as an unassailable advantage? And what happens if it quietly turns into a liability while everyone is still calling it a moat? Thanks for watching. I'll see you in the next one.
Comments 0