The extent to which Joe Biden will be able to follow through on his plan to make the “largest-ever investment in clean energy research and innovation” is currently unknown, but just his presence in the White House will likely lead to, at least, a modest boost in the number of clean energy startup companies entering the market. The more of those companies that succeed, the better off we’ll all be. For policymakers, investors, program managers, entrepreneurs, and advocates who want to help those companies, here’s some advice: encourage startups to seek out new customers that aren’t served by the market leaders. By market leaders, I mean those companies who lead and sometimes dominate the sectors they operate in. Examples include GE Lighting (LED lighting), Carrier (air conditioning), AO Smith (water heating), Sunrun (PV panel installation), and Tesla (electric vehicles).
Startup companies often compete with market leaders out of good intentions. Entrepreneurs, investors, and managers correctly identify the market leaders’ customer base as a rich source of revenue. Policymakers and advocates are eager to quickly gain the environmental benefits that accrue by replacing the market leaders’ product with a more environmentally sound alternative. If you’ve read business or technology development plans you probably ran across a sentence that looks something like this: “If we capture only a tiny portion of (fill in the blank company)’s customers, the revenues (or benefits) would be huge.” Those alluring revenues or benefits are like the Sirens’ songs to Odysseus’ ears.
While the idea of winning those customers is enticing, competing with market leaders is often ruinous for startups because the market leaders’ customers expect high levels of performance, quality, and reliability. Startup companies need time to work out their new products and processes, often through trial and error. They simply have to get through that shakeout period first before they can be ready to meet high customer expectations. Instead, startup companies are more likely to succeed if they begin with offers crafted for undemanding customers who are willing to accept the young companies’ limitations.
One thing that makes customers undemanding is that they don’t need the high performance associated with the market leaders’ products, and they either can’t afford or don’t want to pay for it. If a new company enables these customers to gain some of the features offered by the market leaders’ products, while paying less, then they’re satisfied. They don’t mind that they’re not getting the full performance of the market leaders’ products, because they weren’t going to pay for that performance to begin with. They’re happy to have good-enough performance at a lower price. While the concept seems simple in theory, it takes a great deal of imagination for companies to design offers that appeal to undemanding customers. It’s obvious what the market leaders’ customers want. It’s much harder for startup companies to understand what their core competencies are, and how to apply those competencies to meet the needs of customers who aren’t being served by the market.
A superb example of a company that succeeded by tailoring its initial offerings to undemanding customers has been in the news lately: Arm Holdings. Arm was originally founded to design energy efficient computer chips, but quickly figured out that its fortunes lay elsewhere. Instead of competing head on with the computer processor chip market leader, the Intel Corporation, Arm went into the business of designing chips for the fledgling cell phone and mobile devices industries. These companies didn’t need the high performance of Intel’s products. Instead, they highly valued the energy efficiency of Arm’s chips because it enabled them to provide their customers with longer battery life. It was only after Arm came to dominate these markets did it set out to compete with Intel. To demonstrate the wisdom of Arm’s strategy, history offers us an excellent example of a contemporaneous energy efficient chip company that went after Intel right out of the box and failed: the Transmeta Corporation.
The rise and fall of the Transmeta Corporation
Founded in 1995 by refugees from Sun Microsystems, and other California high-tech manufacturers, Transmeta developed and sold energy efficient microchips for laptop computers. The company’s first product, dubbed the Crusoe chip, cleverly relied on two tricks to reduce energy consumption. First, instead of using hardware—specifically, transistors embedded in silicon—to order the execution of digital instructions, the way conventional computer microprocessors do, the Crusoe chip used software. As a result, Transmeta was able to get by with one-fourth the transistors of its competitors, avoiding the associated cost and energy losses. Second, the company’s microprocessors monitored computing operations as they were being performed, and sped up and down as required. For example, a Crusoe microprocessor might operate at low speed when doing word processing, but shift into high gear when photograph-editing software was started up. In tests, laptops with Transmeta’s chips did consume less energy than the same devices with Intel chips, but also exhibited slower operation and poorer performance.
In its early years, Transmeta achieved impressive and intoxicating success. Not only did the company raise a hundred million dollars before it went public, it also launched the last great IPO of the dot-com boom, in which it raised hundreds of millions more. Such lavish funding wasn’t able to make up for Transmeta’s foundational mistakes, as the company began to suffer from a host of marketing, performance, quality control, and supply chain problems.
Chief amongst Transmeta’s problems was that the company set out to compete with Intel, then and now, the world’s largest computer microprocessor chip manufacturer. At the time Transmeta entered the market, Intel was already cranking out annual revenues exceeding $30 billion and had a reputation for being a fierce competitor. Customers who were already buying from Intel expected Transmeta’s chips to be not only more efficient, but at least similar to Intel’s chips in nearly every other way.
First, Intel took Transmeta’s threat seriously and quickly brought out a more efficient chip of its own. Whether or not Intel’s new chip was as efficient as Transmeta’s was never clear, but it didn’t matter. Transmeta did succeed in selling chips to some major Japanese computer manufacturers, such as Sony, Toshiba, and Fujitsu, but because Transmeta couldn’t demonstrate that its chips performed as well as Intel’s, it never caught on much with other leading computer manufacturers. The company managed to stumble on for a few years with lower than planned revenues, until it attempted to switch manufacturing partners.
Unlike Intel, Transmeta didn’t actually manufacture its chips. It designed them and then farmed out their manufacture to others. Transmeta’s original chip manufacturer was IBM, which maintained high-quality standards, but charged high prices and missed delivery dates. In a misguided attempt to cut costs, Transmeta attempted to switch to the Taiwan Semiconductor Manufacturing Company (TSMC) to produce its next generation chip. The complexity of that chip proved to be beyond TSMC’s capabilities at the time. It took so long for TSMC to learn to produce Transmeta’s microprocessors that most of the company’s customers gave up and walked away. The company’s revenues fell from $35 million in 2001 to $16 million in 2003. By 2009, Transmeta was out of business.
Some analysts have suggested that had Transmeta been able to match Intel on performance it would have prevailed, but that conclusion is an exercise in circular logic. It was silly to expect a small startup like Transmeta to compete with a giant like Intel on the basis of quality, price, and performance. Why did the company even try to do so? Because it had spent nearly $100 million and 5 years developing its initial product, it had to achieve enormous revenues just to stay afloat. The only plausible story Transmeta’s managers could possibly tell investors, their employees, and ultimately themselves, was that they were going to capture a significant portion of Intel’s $30 billion of annual sales.
The rise of Arm Holdings
Originally dubbed Advanced RISC Machine, Arm Holdings was founded in 1990 in Cambridge, England, to be everything that Intel was not. The RISC acronym stands for “reduced instruction set computer.” For those not familiar with computerese, a computer’s instruction set is a list of commands designed to operate a processor. It may include instructions like add two numbers together, store information to random access memory, or output information to a particular device. Intel, even then, made chips that used a complex instruction set. Because Arm’s instruction set was small compared to Intel’s, chips based on Arm’s instruction set required fewer transistors, which then enabled them to dissipate less heat and be more energy efficient.
Instead of using this ability to compete with Intel in the laptop and desktop computer markets, Arm focused on markets that Intel wasn’t interested in: cell phones, personal digital assistants, digital cameras, global positioning systems, and hard drives. Now, we know that these devices would eventually form the basis for huge markets, but at the time, in the early 1990s, many of these industries were in their infancy. They needed to offer their customers low costs and long battery life, and the simpler, cheaper, and highly energy efficient chips designed using Arm’s reduced instruction set made that possible. Indeed, they so valued low costs and long battery life that they were willing to put up with slower operating speed to get them.
Another way Arm distinguished itself from Intel, is that the company never developed the capability to manufacture semiconductor chips. Instead Arm focused on selling designs and instruction sets to others. Sometimes, those other companies incorporated Arm’s intellectual property (IP) into their own designs, and then farmed out the manufacturing of those chips to other companies. Many of the world’s leading chip companies buy IP from Arm, including Apple, Samsung, Qualcomm, Texas Instruments, and, yes, even Intel.
By 2010, Arm dominated the mobile computing market and nearly all cell phones, even to this day, contain at least one chip made using Arm IP. Only after Arm achieved dominance in these markets did it start to invade markets in which it faced competition with Intel, including data center servers, laptop computers, smart televisions, and supercomputers. For example, Apple is currently migrating its Mac computers from Intel microprocessors to chips designed using Arm’s instruction set. The world’s fastest supercomputer (at the time of writing), and which is also one of the most energy efficient, is based on Arm designed microprocessors. Arm is also moving into chips for autonomous vehicle computing.
This year, Arm gained another opportunity to drive energy efficiency into more markets. In September, Nvidia, one of the world’s largest semiconductor chip companies, announced that it was buying Arm for $40 billion. This deal, if it comes to fruition, will be the largest semiconductor chip deal ever (at the time of writing). Nvidia’s main business is designing processors for gaming, mobile computing, and automotive applications. One reason analysts have given for Nvidia’s interest in Arm is that in order for the company to expand into the data center and artificial intelligence markets, it needs to improve its energy efficiency. That’s certainly something Arm can help with. Furthermore, combining the skills of both companies could help them compete more effectively in the Internet-of-Things market.
Arm shows us the way
The contrast couldn’t be more evident:
Transmeta went into immediate competition with the market leader. Arm didn’t.
Transmeta’s products were designed to appeal to the market leaders’ customers. Arm designed new products for customers who weren’t being served by the market leader.
Transmeta tried to convince the market leaders’ customers to sacrifice performance for energy savings. Arm went after customers who were willing to sacrifice performance because energy efficiency was critical for their products.
Transmeta took on a huge debt burden that couldn’t be serviced by selling to undemanding customers. Arm didn’t.
Transmeta attempted to sell actual computer chips. Arm sold only designs and IP.
Transmeta went out of business. Arm is one of the most successful companies in the history of the semiconductor chip industry.
One facet of Arm’s success that energy efficiency advocates might find unsatisfying is that it’s not clear how much energy Arm’s technological advantage in energy efficiency actually saved. Because Arm came to dominate the mobile computing market early, we can’t actually point to another less-efficient chip and calculate how much more energy would have been consumed had that chip been used instead. Arm’s recent forays into the laptop computer, data center server, and supercomputer markets, are now manifesting discernable energy savings, but those efforts didn’t get underway for decades after Arm’s founding. Then there’s the incalculable energy savings that occurred because Arm’s competitors, including Intel, were driven to improve their energy efficiency. To these advocates, all I can say is that successful energy efficiency companies are far more likely to have a positive impact on climate emissions than failed companies. Even if the energy benefits aren’t calculable, or take decades to materialize, we’re all still better off when energy efficiency companies succeed.
Policymakers, investors, managers, entrepreneurs, and advocates, take note of the lessons that the epic saga of Transmeta and Arm Holdings holds for us. For companies to have a positive impact on the climate, they must first become successful businesses. Encourage clean energy, and especially energy efficiency companies, to design their early products for undemanding customers. Their chances for success will likely be improved.
Image caption: In this test conducted about 20 years ago, Transmeta engineers took thermal images of both an Intel processor and a Transmeta processor while they were playing a DVD video on a small notebook computer. The Intel chip, on the left, operated at about 222° F and melted shortly after the image was taken. The fact that it melted was not surprising, because it was designed to operate with a heat sink. The Transmeta chip on the right operated at about 119°F and did not require a heat sink.