Please enjoy ARK Disrupt Issue 141. This blog series is based on ARK Brainstorming, a weekly discussion between our CEO, Director of Research, thematic analysts, ARK’s theme developers, thought leaders, and investors. It is designed to present you with the most recent innovation takeaways and to keep you engaged in an ongoing discussion on investing in disruptive innovation. To read the previous issue, click here.
1. A New Wave of Electric Vehicles Isn’t Competition for Tesla: It’s Competition for the Internal Combustion Engine
Follow @skorusARK on Twitter
NIO’s upcoming IPO and BMWBMW.DE, AudiVLKAY, and Mercedes’sDDAIF EV announcements have caused a resurgence of the narrative that TeslaTSLA will struggle once competition arrives. ARK’s research suggests that the EV adoption cycle is much too early for analysts to focus on other EVs as competition to Tesla. Instead they should expand the EV market, at the expense of the internal combustion engine (ICE).
How will Mercedes, Audi, and BMW convince consumers that their EVs are better than Tesla’s without suggesting also that EVs are better than gas-powered cars? If they do succeed in converting consumers to EVs, then traditional auto manufacturers will accelerate the EV supply chain development and management that Tesla and other early movers started.
NIO’s roadshow offered a good example of what a fully formed supply chain will enable. While it took Tesla six years to hit an annual production rate of 100,000 EVs, NIO hopes to reach that level in half the time, as shown below. That time compression marks the difference between a technology pioneer and a fast follower.
2. Here is ARK’s Take on Benedict Evans’ Thoughtful Post on Tesla
Follow @tashaARK on Twitter
This week, Benedict Evans wrote an interesting thought piece on Tesla as a disruptor. One of his conclusions is that Tesla’s vertical integration is superior to the horizontal structure of traditional auto manufacturers. Munro Associates discussed this difference in its teardown of the Model 3. While the Model 3 has a centralized circuit board that consolidates 3 or 4 modules, traditional cars are manufactured with several electronics systems from different suppliers, each with separate software and chipsets. Tesla’s approach not only saves costs and processing power, but also enables over-the-air (OTA) software updates to enhance performance. OTA performance updates require that all of a car’s subsystems communicate with one another. Unlike Tesla, traditional automakers have little control over changes to their disparate software ecosystems and chipsets. As a result, the shift toward OTA and, ultimately, autonomous electric vehicles probably will require a transformation more in the massive supply chains that support traditional auto manufacturers than in the auto manufacturers themselves.
Evans also notes that battery cells are becoming commodities, and ARK agrees. Battery cell price declines should lower the cost of Tesla’s highly engineered and integrated battery pack system. While Evans postulates that other auto manufacturers will erode Tesla’s competitive advantage in back pack systems in the next 5 years, ARK argues that it’s easier said than done. General MotorsGM outsources battery pack production today for a reason.
Like Evans, ARK believes that mobility-as-a-service (MaaS) will evolve into geographic monopolies, thanks to the region-specific data collection that will feed and drive autonomous systems. Machine and deep learning systems improve with more data. We believe Tesla enjoys a significant advantage as the only automaker collecting data from its customer fleet. Launching an autonomous, highly utilized fleet of vehicles, each collecting 100,000 miles of data per year, should turbocharge this advantage.
All that said, Evans does argue that the absence of LiDAR in its autonomous sensor suite is a mistake and will prevent Tesla from developing a reliable perception system. For background, the perception system of a car deciphers the environment around the car, while the path planning system determines how to get from A to B. The better the perception system, the easier path planning becomes. Conversely, the worse the perception system, the harder path planning becomes.
While time will tell, ARK is more optimistic than Evans about Tesla’s prospects without LiDAR. In our view, Tesla has allocated resources more toward path planning, resulting initially in a lean, less detailed perception system. Arguably, path planning is a more difficult challenge than perception, a reason perhaps why Elon Musk calls LiDAR a crutch and why he believes a system without LiDAR can solve perception.
3. The Difficulty in Producing Leading Edge Nodes for the Bitcoin Mining Stack is Increasing
Follow @yassineARK on Twitter
Semiconductor device fabrication is the process that creates the integrated circuits present in Bitcoin mining rigs. At the highest level, the process involves creating electronic circuits on a wafer made of semiconductors. These wafers comprise the bulk of the mining box.
As Bitcoin mining has transitioned from CPUs and GPUs to Application Specific Integrated Circuits (ASICs), technology nodes largely determine the efficiency of mining machines. Shrinking sizes characterize successive generations of a node. The latest generation ASICs contain 7 nm nodes.
Mining machine manufacturers have no control over node advancement and, beyond ASIC design, little control over the efficiency of their mining machines. Instead, they depend exclusively on foundries both to develop the latest generation nodes and to supply them with enough wafers to manufacture new models at scale.
At the foundry level, leading the charge in technology is highly capital intensive, with only a few foundries dedicated to researching and developing the cutting-edge node. To date, of the twenty-eight foundries providing fabrication, only three provide leading edge fabrication, as shown below.
The degree of difficulty in leading edge fabrication escalates from generation to generation, as successive nodes require bursts in expertise and capital expenditures, as shown below.
GlobalFoundries’ recent decision to halt its 7 nm production is a case in point. In late August, GlobalFoundries announced it would cease development of the leading-edge node, in addition to research and development for 5 nm and 3 nm nodes. This decision was purely economic, provoked by GlobalFoundries’ unwillingness to fund the $15 billion initiative. Instead, it has shifted its focus to “specialized process technologies for clients in emerging high-growth markets. These technologies will include RF, embedded memory, and low power features.”
Taiwan Semiconductor Manufacturing CompanyTSM, SamsungSSNLF, and IntelINTC are the last foundries focused on cutting-edge technology, as shown in the first figure. Only TSM and Samsung are relevant to Bitcoin mining. ARK’s current research is focusing now on how this leading-edge fab duopoly will impact the Bitcoin mining stack.
4. Server Sales Reach an All Time High as Moore’s Law Slows
Follow @jwangARK on Twitter
For decades, Moore’s Law provided the closest thing to a “free lunch” for computing, doubling performance roughly every two years at the same cost. Now that the free lunch is ending, computing is about to get more expensive, as shown in the latest global server sales figures from IDC.
During the second quarter, server revenue grew 44% to reach $18.4 billion, an all-time high, all the more impressive after flat server sales for nearly 20 years. Strong server sales reflect robust business computing demand, in contrast to tepid consumer demand, with PC and smartphone sales trending down. A key difference between enterprise and consumer is data—enterprises own tons of it and consumers own little. At a very high level, enterprises could be buying servers at an accelerated rate because they can convert data into improved business outcomes. Consumers, on the other hand, have little incentive to buy marginally better smartphones. If AI-powered analytics is driving this spend, accelerated server growth could become a secular trend, greatly benefiting microprocessors, memory, and other component suppliers.
5. Plaid is a Plumber for Fintech Platforms
Follow @bhavanaARK on Twitter
Plaid, a San Francisco based startup, is becoming a building block for the fintech ecosystem. The company facilitates communication and connectivity among banks and fintech companies such as VenmoPYPL, Acorn, Coinbase, Betterment, and Robinhood.
Until recently, fintech companies had to process small dollar transactions through banks to verify customer identity, a cumbersome and time-consuming process. During its early days, for example, Venmo’s customers had to wait a day to open their accounts. Plaid eliminated this friction with some software plumbing. With access to background information about user history from fintech companies, Plaid can get instant authorization from the banks. Its software and services should make Plaid a valuable partner not only in the financial sector but also potentially in other verticals, including the sharing economy, travel and e-commerce.
6. Did Sangamo’s Trial Results Hurt Prospects for Gene Editing?
Follow @msamyARK on Twitter
This week, Sangamo TherapeuticsSGMO released disappointing results from the first US-based trial of in vivo human genome-editing safety and efficacy. Data from this Phase 1/2 CHAMPIONS trial were inconclusive, casting a pall over the entire genome-editing space. While disappointing, we believe that the results illustrate the limitations of one form of gene-editing, zinc finger nucleases (ZFNs), not CRISPR or other forms of gene-editing.
Sangamo’s lead gene-editing therapy, SB-913, aims to treat MPS II, also known as Hunter’s Syndrome. Seemingly caused by a deficiency in iduronate-2-sulfatase (IDS) enzymes, MPS II is a lysosomal storage disorder characterized by the build-up of glycosaminoglycans (GAG) molecules around internal organs, including the brain. While the FDA has approved another drug for the treatment of MPS II, Shire’s Elaprase is an enzyme replacement therapy that requires repeat dosing which is removed from the body after 24 hours: molecules never cross the blood brain barrier and do not prevent the buildup of GAG. According to various studies, a sustained expression of IDS should reduce the build-up of GAG in the brain, so Sangamo had hoped that SB-913 in combination with Elaprase would prevent organ damage by correcting the IDS gene.
That drug combination did not work. At the lowest dose of both SB-913 and Elaprase, two patients saw no change in IDS levels and suffered anincreasein GAG levels. At the middle dose level, two patients did see a significant 51% reduction in GAG protein levels, but study investigators were unable to detect levels of IDS, causing questions as to whether patients were responding to the Elaprase treatment or the SB-913 gene correction.
Unlike CRISPR, Sangamo uses adeno-associated viruses (AAV) to deliver therapeutic ZFNs into patients. While AAV vectors are space-constrained, ZFNs are large protein molecules that demand a lot of space. Investigators in this study will have to increase SB-913 dosing levels with more AAV vectors, which will increase the risk of immunogenicity. CRISPR’s counterparts are much smaller in size, and under similar circumstances, probably would not face such tradeoff between safety and efficacy.
ARK's statements are not an endorsement of any company or a recommendation to buy, sell or hold any security. For a list of all purchases and sales made by ARK for client accounts during the past year that could be considered by the SEC as recommendations, click here. It should not be assumed that recommendations made in the future will be profitable or will equal the performance of the securities in this list. For full disclosures, click here.