Micron Technology Defied the Odds
Nobody believed in Micron Technology. A startup starting ten years late, dealing in commodity chips. Coming out of a city more known for potatoes than silicon: Boise. Facing down the world's most competitive semiconductor industry.
Somehow, they beat those odds. In this video, we look at the memory underdog from Idaho: Micron Technology. ## Beginnings I have never mentioned Boise before on this channel - it is not exactly near Asia - so let me introduce it to you. Boise the city began in 1835 as a trading post called Fort Boise set up by the Hudson Bay Company. The name's origin is uncertain. It grew into prominence during the American West's various gold rushes of the 1840s and 1860s supplying materials to miners. In 1864, it became the capital of the Idaho territory. Idaho then became an American state in 1890.
Idaho is particularly known for its potatoes. People were farming potatoes there as early as 1837. But the industry really sprouted in the 1950s, after the newly-established Idaho Potato Commission began marking their potatoes as being "Grown in Idaho". The harsh lands of Idaho gave rise to some of America's most extraordinary success stories like the billionaire Jack R. Simplot, founder of the massive JR Simplot Company.
Simplot started hustling for business at the age of 14 and slowly built a processed food business empire. That includes supplying half of McDonald's French fries - or chips if you want to call it that. These Boise lands will also give rise to another extraordinary success story: Micron.
## The Parkinsons Identical twins Ward and Joe Parkinson grew up in a small town in Idaho called Blackfoot. Their father was a hard-working man who ran a 1-man insurance agency. The Parkinson family emphasized education, so they studied hard while working in potato fields owned by the JR Simplot company. Both boys got good grades, won school debates and science fairs so they were accepted to Columbia. Joe Parkinson - the more eloquent of the two - went into law, and eventually returned to Boise to open his own law firm. Ward, on the other hand, took a different path. He dropped out of the Columbia pre-med
program and did an EE degree at Utah State while also working as a farmer. After that, he joined the famed Fairchild Semiconductor where he designed ECL logic and memory circuits. Fairchild was then going down the drain, but Ward distinguished himself. So much so that they offered to pay for his Masters in computer architecture at Stanford.
After he graduated, he was poached by his old boss at Fairchild to work at a company called Reticon, producing image sensors. Two years after that, he jumped again to Motorola where he designed memories. It was the late 1970s and it was becoming clear to everyone in the industry that Metal Oxide Semiconductors or MOS was the future. Ward wanted to get more experience in that, so he jumped to the leading MOS memories company - Mostek.
## Mostek Mostek was founded in 1969 by a few Texas Instruments folks led by LJ Sevin. Sevin and his co-founders were frustrated with their managers' conservative insistence on bipolar transistors over the high-potential MOS as a challenger to magnetic core memories. The new Dallas company introduced a plethora of memory breakthroughs like ion implantation and address multiplexing. The latter let Mostek make more complex Dynamic RAM modules without adding more pins and thus taking up valuable space on the board. Mostek quickly snagged the lead in world Dynamic RAM market share. Particularly,
a brilliant 4K product called the MK4027, first introduced in 1976. And the strength of their design was what convinced Ward Parkinson to join. There, he joined a design team he greatly respected. This included his manager, Dr. Paul Schroeder, who helped design the aforementioned MK4027. Dennis Wilson and Doug Pitman in particular were similarly great memory designers.
## Inmos & Micron But unexpectedly in 1976, Mostek suffered a huge shock. One of its own cofounders Richard Petritz left to start a competing semiconductor company backed with British money called Inmos. Petritz hired over the aforementioned Paul Schroeder by giving him the deputy managing director title. Schroeder in turn convinced his colleagues Parkinson, Wilson, Pitman, and a fourth guy Dave Wooten to leave Mostek and join them at Inmos. They were not the first to leave, but for Mostek it was the last straw. Infuriated,
LJ Sevin filed a lawsuit against Inmos, hoping to get a permanent injunction against poaching. Doug Pitman and Ward Parkinson were personally included on this suit. It bombastically accused them of everything from stealing trade secrets to violating antitrust law by hoarding design talent. Inmos later handily won the suit, but did little to help the designers targeted by it. Probably should have, considering what happened next. In October 1978, the three designers ditched Inmos to start their own design consulting company, Micron Technology.
The new Micron team called LJ Sevin, and this might sound strange but Sevin quickly contracted them to design the next generation of memory chips for Mostek. But since the price was a bit low, Micron asked for the option to manufacture any design they made. Micron was then just three dudes without even an office, so Sevin gave it to them. They worked out of Dallas for the first year, but soon Pitman, who was from Idaho Falls, wanted to go home to the Northwest and raise his four kids. After looking at Montana and Utah, Boise was chosen because it had an airport and Joe Parkinson had many business contacts there. Whilst on a scouting trip, Pitman scoured Boise for new space and found a half-finished dentist's basement for half the normal rental price.
A donation to the Catholic Church next door gave them free parking. ## Japanese Memory Despite receiving the option from LJ Sevin, Micron never intended to go into manufacturing - though they sometimes fantasized about it. Then the market changed. In the late 1970s, Japanese competition in semiconductor memories intensified. Intel shipped the first 1K Dynamic RAM in 1970. In the next generation, 4K, they and the Americans took 90% market share while the Japanese grappled with new innovations like the aforementioned address multiplexing. Note that memory incremented four times each generation.
But in the generation after that 16K, the Japanese began gathering substantial market share - thanks in part to memory module pin standardizations that Mostek pioneered. Then in 1979, Hitachi, NEC and Fujitsu shipped the 64K DRAM at the same time as the Americans. And thanks to their lower prices - stemming from superior processes and government subsidies - they took the market share lead. Mostek and the other American memory makers began to experience declining, volatile profits.
It allowed the acquisitive United Technologies - a conglomerate that made helicopters, elevators and other equipment - to buy them out in August 1979. They promptly canceled the design contract with Micron Technology. ## Hell No Micron now had no customers and no source of revenue. But they still had the 64K DRAM design, which they completed, and a few others. So in what is perhaps a fit of mania, the Micron team decided to take matters into their own hands and go make these chips right in Boise, Idaho.
It was legitimately a bit crazy. Leslie Gill was Micron's first CTO. When Joe Parkinson told her that they were going to make chips, she thought "Potato or corn?" No surprise that traditional venture capitalists refused to invest. Some asked if they had to close the factory during the harvest. Ward Parkinson famously recalled that people didn't just tell them "No", they said "Hell no". ## Yanke and Noble As fund-raising went on, Joe Parkinson became increasingly important to the company.
He eventually joined Micron Technology as its President and fourth co-founder, though he still practiced at his law firm for a while. It is through Joe that the Micron team met Ron Yanke. Yanke took over his father's machine shop and turned it into a success. He parlayed that into successful mining, timber, real estate, and aviation investments.
Yanke had a business partner - a potato farmer named Allen Noble. Noble lived one of those classic American success stories. He ran away from home at the age of 14, bought a few acres for farming potatoes, and turned that into an empire. His flagship project was pulling water up 500 feet from the Snake river into the Dry Lake area, transforming what before had been a dry desert into a potato bonanza. A few months earlier, Noble and Yanke asked Ward through Joe Parkinson to troubleshoot an electronic controller for a broken irrigation robot they were working on.
Ward warned them that this wasn't his specialty but nevertheless worked for several weeks on the side to fix it. Noble grew to like the guy and later said that he would have backed them on anything they were going to do. So when the Micron team came calling, he knew he would be in. Over lunch, Noble and Yanke agreed to kick in some seed money to get things rolling. ## The Odds Neither Noble or Yanke knew semiconductors. Ward could not take their money in good faith without telling them what they were up against.
So he arranged for them to go to Texas to meet with LJ Sevin. The August 1979 meeting was like one of those "scared straight" therapies they do for 12 year olds but for semiconductor investors. The charismatic Sevin, himself a former designer, laid out Micron's challenges. Micron was a design consultancy with little experience in manufacturing.
They were entering the market in 1979 - some ten years behind established players like Mostek, Intel, AMD or IBM. Sevin estimated it would cost about $100 million in startup money to get into the memory market. And that market is extremely, sickeningly volatile, estimated to lose $300 million in 1981. The Japanese were then going nuts and pouring what was said to be billions of state-backed dollars into the industry. They were already seeding
next-gen 256K and everyone expected them to steamroll the Americans. Noble then asked Sevin, "Well if the market is so bad, why was Mostek still in it? What hope did they have?" To this blunt question, Sevin gave a sad smile and replied with a generic answer about the importance of being nimble before adding at the end, "But sometimes I wonder ..." On the way back to the airport, Ward asked Noble and Yanke what they thought of what they just learned - fully expecting the worst. Noble reportedly said, "Let's do it". Yanke and Noble helped bring in a third investor, Tom Nicholson, who owned one of the biggest sheep ranches in the country. Nicholson provided the initial land and water rights for the factory. For all that, they were still short. The Micron team estimated that they needed $6 million in
new loans and investments. Right then they had a million, and another million in bank loans. Noble and Yanke were rich. But not rich enough to single-handedly fund a semiconductor fab. However, Noble knew someone who was and that was how Micron got a meeting with JR Simplot, the potato billionaire.
"What'll take it to get her rolling", he asked. "$4 million", they said. Simplot later said that he saw that these farm boys - referring to Yanke, Noble, and the Parkinsons - had gone all in on this venture. And for him, that was enough. "Alright, we'll go for it", he said. Simplot,
his son Scott, and his companies eventually bought 40% of Micron, bankrolling chip R&D and factory construction much to the consternation of his finance team. ## Building the Factory Micron's corporate strategy was simple - a reliable product at a low cost. They wanted to be the low-cost leader. Their first 64K chips were built in a subcontracted National Semiconductor fab in Utah while the Micron fab was being built. That was completed in 1981 at the cost of $20 million, or about $76 million today. Plenty of money, but just a fifth of what a DRAM factory cost even in those days. How?
First, land, labor, water, and energy were far cheaper in Idaho than they were in Silicon Valley and even Japan. Hydropower in Idaho is quite common, and rates were a third of those in California. But that is not enough, so we should talk about equipment. One reason why a TSMC fab is so expensive is because they have a lot of expensive tools inside. Micron generally does not use the most advanced tools. For instance their lithography tools were older models from Perkin-Elmer, ASML, or Canon.
In fact, whenever new technology came out, Micron would not adopt it to move on to the next-generation density memory like others did. Rather, they preferred to use it to make a far cheaper version of the current generation out in the market. Next, redundancy. Another reason why a big TSMC fab is expensive is because it has a lot of tools for redundancy. So if one breaks down, the line does not stop because you always can route the work-in-progress to another, working machine.
When Micron was building their first factory, they gave up on that option. As CFO Gill said: > "Long term, you may need to back up all your equipment, but to make the first chips, you don't. You have one machine and hope it works" ## Design Cleverness Last and perhaps most importantly, design and manufacturing were lock-step in sync about being the low-cost leader. That means pulling out all the stops on the design side to achieve the lowest cost memory bit.
They were especially good with layouts. Doug Pittman was a layout guy. If you recall in the semiconductor design process, layout is where you wire together the various memory cells after they have been chosen for the design. Companies doing logic chips usually do the layouts with an electronic design automation software, which does it relatively inefficiently but cheaply. And that's fine for logic, where there is relatively available space on the silicon. But memory is different. A smart layout can let Micron cram more memory cells
into the same piece of real estate. High density can mean a small die, which in turn means more dies on a single wafer. Which in turn means a cheaper per-die cost. Members of the product and process engineering team also pitched in. For instance, by reducing the number of masks. Masks represent a layer in the final design, so the more you have, the more complex the design is, and thus the more expensive it is to make. Low mask counts - sometimes 30-40% lower - were an underrated factor for Micron’s cheaper costs.
The company regularly cites mask reductions in its shareholder letters as an indication of cost cuts. Micron never had a shortage of cleverness. Early on, Ward figured out an alternate use of their defective DRAM chips - turning them into "optic RAM" image sensors for early computer cameras. It sold quite well in the market for several years as the MicronEye accessory.
Though being sold in the early 1980s, it was probably far ahead of its time. After a dodgy few months, the first MT4264 DRAM chips started coming off the line in late October 1981. They let the first inventors - Simplot, Yanke, Noble and others - look at them through a microscope. JR remembers it being a wonderful day. ## Entering the Market Micron's chips first entered the market in 1982. That year, they sold a million chips. Customers noted how small the chips were: 33,000 square mils - 1 mil means 0.001 inch. This was 40% smaller than Motorola's, 15% smaller than Hitachi's. It was a size that even TI struggled to achieve.
They were an ideal fit for the growing PC market. Apple bought them for their Macintosh and Lisa PCs. Jack Tramiel personally guaranteed millions in sales from his company, Commodore. Despite this and improving yields over 30%, the company was still losing money on each chip sold. It became clear that something drastic was needed. Fortunately, designers had deliberately
made the chip bigger than it needed to be in case of manufacturing process variations. But Micron's second fab had better equipment and produced larger 5-inch wafers. So the designers removed those tolerances, shrinking the die to a mere 22,000 square mils. This and the bigger wafers right-sized the chip economics.
Released in January 1983, the new smaller 64K MT4264 chip was an immediate hit. Sales surged from half a million in 1982 to 10.9 million, net losses narrowed from $6.8 million to just $2.6 million. The backlog reached $23 million so Micron immediately started hiring and expanding the factory. Thanks to continued good DRAM sales from the growing PC revolution, Micron's sales the next year 1984 surged to $84 million, producing $29 million in profits which they funneled into 256K development.
That year, the company also went public, selling 2.1 million shares on the market. The stock were offered at $14 and went to $40. JR Simplot and his family and companies still held almost 19% of Micron, and made $150 million on his early investment. JR was such a believer that he told his employees to buy the stock, even on margin. And that he would make whole any loss they might take on that borrowing. He liked to say, "We are going to build the memory of the world here in Boise".
## From Up to Down But LJ Sevin had not been wrong. Memory is a commodity and the market can turn at the drop of a pin. In late 1984, Micron's salespeople noticed that customers were starting to ask for lower prices and concluded that a market downturn was on its way. They decided that they should get ahead of this decline. So in September, they shocked the industry by cutting the price of their 64K chips from $3.40 a chip to as low as $1.95.
It was a play for market share that puzzled market observers and infuriated the rest of the industry. The stock declined 25% in just two days. Lane Mason of the research firm Dataquest said, "A price drop of that magnitude was unwarranted, unwise, and leaves a lot of money on the table". NEC Electronics' president Keiske Yawata said, "I don't see marketing expertise in what they are doing".
VC Pierre Lammond, who invested for Sequoia and Khosla Ventures over a long career, really laid in there with: "I think their luck is going to run out. The large Japanese companies and some of the US manufacturers will eventually wipe them off the face of the map". ## 1985 Thanks to this price cut, Micron managed to pile up orders and win favor with customers. Initially. Then came a disastrous 1985. The year began with high hopes. The Semiconductor Industry Association predicted a nice 22% growth in sales. Then suddenly, it all fell apart thanks to weakness in the American electronics industry.
Prices on every piece of electronics from CD players to gaming consoles to PCs crashed. Several VC-funded PC makers collapsed, hurting demand. American chip sales fell by 28%. National Semiconductor in particular took a heavy beating. Their sales fell 17% and profits swung from plus 59 million to negative 117 million. They shut down for 39 days of the year and laid off 2,200 people. The Japanese continued to produce, even as the Americans cut back.
Insulated from financial damage by their sibling banks, electronics, or real estate arms, they sold at whatever it took to get the sale. After the price cut in September 1984, prices for 64K memory fell from $2.20 per chip to under $1.00. During the summer, spot prices for 64K were hitting an unthinkable 35 or 40 cents. It felt like there was no bottom. In February 1985, Micron laid off nearly 50% of its 1,250 employee base, 500 people. They also dialed down 64K production, shifting more towards the next generation 256K.
Eleven American memory-makers exited the market that year - most famously, Intel. United Technologies sold Mostek to Thomson of France. By the end of the year, the only American companies still making Dynamic RAM memory were Micron and Texas Instruments. ## The Petition Micron eventually decided that greater measures were necessary. In June, they filed a petition with the US International Trade Commission and the US Department of Commerce alleging a material injury due to dumping practices by the Japanese - essentially selling below cost.
They requested an investigation into this and possible countermeasures. They also filed a lawsuit against Japanese firms themselves for dumping. Micron was the first to take such steps and it demonstrated their differences with the rest of the American semiconductor industry - represented by the Semiconductor Industry Association or SIA. Joe Parkinson said: > It was very clear to me that they [SIA] had a different agenda ... their strategy was whatever the Japanese get into,
let's get out. The people who are dominant in SIA are not taking the Japanese on. Indeed, the SIA was deeply divided. Some like Motorola and Charlie Sporck of National Semiconductor wanted to fight. But there was also a sizable faction that believed
that the US should instead move on to more valuable, design-rich products like Intel did. And several - including those in Japan - thought that Micron was being self-serving. Jack Beedle of In-Stat, a semiconductor market research firm, said it best: > Everyone was cruising along at $3 until they started the price war. Then they turn around and sue Japan on prices ... any company that only produces one kind of part when that part is also manufactured by the Japanese is going to have serious problems Not the most charitable view, but it was reflective of the larger industry in early 1985 - that Micron had made its own bed and should sleep in it. The Japanese makers cited Micron's actions as exhibit number one in their defense.
## The Agreement I shall sidestep this era of semiconductor trade discord between the US and Japan, but as it became clear that something was indeed going on, the industry joined in. In August 1985, the International Trade Association made a preliminary ruling that there was indeed some material injury. With the help of Idaho's senators, the issue escalated to the highest levels of American government.
Eventually the Japanese came to the belief that some accommodation was necessary despite continued disagreement on the dumping issue. Akio Morita of Sony led informal talks with the Americans on a resolution to bring back peace. Thus came the 1986 Semiconductor Agreement, which put price floors on Japanese semiconductor products and opened up new markets in Japan. The guys in Idaho were happier,
though skeptical that the Japanese would actually follow through. All throughout this ordeal, Micron still suffered $50 million of losses in 1986 and 1987. When the Japanese lagged in implementing the 1986 agreement as the Idahoans feared, Reagan imposed tariffs in early 1987, reducing supply and production from the Japanese. When the market turned again in early 1988, it caused a DRAM shortage. Memory prices shot up, irritating American computer and electronics makers, who screamed "Chip shortage!" But Micron - as one of the two last American memory makers standing - did very well, turning nearly $100 million in profit. They had survived the Japanese wave.
## Conclusion In 1998, amidst another brutal memory downturn, Micron bought Texas Instruments' DRAM business. It solidified their position as America's final DRAM memory-maker - the last survivor of one of the chip industry's most brutal spaces. Today, they are a $150 billion company. Armed with $6 billion in CHIPS money, they are setting out to build leading edge memory fabs in Idaho and New York. The Micron Technology story is remarkable. You can't make it up. The way these guys started out,
persisted, and eventually won the market through smart design, cost control, and sheer determination is a demonstration that semiconductor success is possible almost anywhere.
2024-09-06 20:49
