solar roof, ark brainstorming, investing in disruptive innovation

6 Takeaways from ARK’s Brainstorming: Solar City, Google Self-Driving Cars, Intel Enters AI, …

This is a blog series 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 keep you engaged in an ongoing discussion on investing in disruptive innovation.

1. The Allure of a Solar Roof
This past week, on SolarCity second quarter earning’s call, Elon Musk made a guest appearance and a surprise announcement that SolarCitySCTY is working on a “solar roof”. While historically ARK’s analysis had concluded that the residential solar panel business would remain uneconomic, a solar roof could be the “no brainer” that Musk referenced when TeslaTSLA announced its acquisition of SolarCity a few months ago.

We think the sales efficiencies that Musk described come into clear focus when the product on offer is a solar roof. Targeting customers already spending a significant sum to build or rebuild their roofs, SolarCity will be offering incremental solar hardware, not installation services. Not only adding to resale value as they build or rebuild their homes with solar roofs, homeowners will be able to capitalize on mortgage financing, a much cheaper source of funding than SolarCity itself could offer.

Building-integrated photovoltaics (BIPV) is not a new concept. In 2011, Dow Chemical offered solar shingles in the U.S. under a product line called Powerhouse. Two months ago, Dow announced that it was ending the business, highlighting that a solar roof might make sense economically on paper, but execution is critical. ARK will follow Tesla’s solar partnership with its new subsidiary, SolarCity, closely as it releases more details.

2. Google’s Self-Driving Car Project, Without Chris Urmson
Chris Urmson, lead engineer and former head of the GoogleGOOG Self-Driving Car Project, left this week, a major setback for the project. Rumor has it that Urmson clashed with John Krafcik, the Hyundai executive who took over as CEO of the project. Many other autonomous car teams will welcome Urmson with open arms, so it will be interesting to see where he ends up.

Urmson is one of many key people who have left Google’s X projects during the past two years. Anthony Levandowski, who ran the startup Google acquired to jumpstart its car project, also left earlier this year. Under new CEO Andrew Conrad, Verily’s life sciences venture has lost a number of its top scientists. Tony Fadell, CEO of Nest, left in June. Most recently, Bill Maris, who started Google Ventures, departed this week. What is going on? Is Google restructuring as it holds its X division’s executives more accountable for financial progress, and/or are startups luring these entrepreneurs back to their “early stage” roots?

We think the Self-Driving Car Project is arguably one of the most successful divisions of X, having built the most advanced autonomous car in the world. ARK believes that if the Google car were to continue improving at its current rate, it could produce a fully autonomous car for commercial use in the next two to three years. Autonomous taxi rides could cost half of the price we pay to drive own cars today, suggesting that the global market for autonomous taxis will be worth trillions.

3. Intel Enters AI Arms Race with Nervana Acquisition
On August 5th, ARK tweeted: “Prediction—Big tech companies are going to gobble up AI startups for the next decade.” Just four days later, investing $400 million IntelINTC acquired Nervana, a semiconductor startup that aims to build chips optimized for deep learning.

Based on ARK’s analysis, as shown below, when Nervana migrates its manufacturing to the 16 nanaometer (nm) process, Nervana will be able to build chips with speeds more than four times faster than Nvidia’s GPUs today. Currently Nervana is designing its processor at 28nm, delivering speed as measured by tera operations per second not much faster than Nvdia’s.


We’ve seen previous attempts to supplant the GPU, notably Sony/IBM/Toshiba’s Cell processor and Intel’s Larrabee chip, that have been unsuccessful. Nervana, with its focus on deep learning, may fare better in our opinion.

Unlike other parallel processors, Nervana’s need to succeed at only one workload—deep learning. We think deep learning requires little more than multiplication and addition operations. In theory, a clean sheet design maximizing the number of hardware units for these two operations should outperform existing GPUs–which also have to support graphics, video, and a plethora of other workloads–significantly.

Reality may prove more challenging. Now owned by Intel, ARK believes Nervana must satisfy the needs of its new corporate owner which needs to pivot from its dependence on CPUs. Nervana needs not only to build a world class chip, but to orchestrate its design, manufacturing, and go to market within the belly of a $160 billion market cap lumbering giant. Needless to say, Intel has not had the best record in the realm of mergers and acquisitions.

4. Bitcoin News
This week we met with a number of bitcoin and blockchain companies to discuss the future of “initial coin offerings (ICOs),” the equivalent of an IPO in the realm of cryptocurrencies. ICOs, as illustrated by both Ethereum and the DAO, enable a small team of developers to raise millions of dollars and capture a broad base of enthused users in a breathtakingly short period of time. But what, exactly, does the U.S. Securities & Exchange Commission (SEC) think of this new way to raise money? Do the “coins” qualify as securities?

While we can’t know what the SEC’s is thinking precisely, we can apply the Supreme Court’s “Howey Test” to these offerings to determine if they are “investment contracts” and where danger may lie. Coincenter, a highly respected non-profit research center focused on cryptocurrency policy, published a 58 page paper on this topic here.

Essentially, an investment contract (i.e., security) is an investment of money in a common enterprise with an expectation of profit in a third party’s efforts. With many nuances in this definition, we think the most thorny in the context of ICO’s is the “common enterprise”: what is the “common enterprise” when dealing with distributed systems of trust? If the entire system makes up the common enterprise, who is responsible if the system is compromised? If ICOs qualify as securities, the SEC will have to make that decision.

5. Can DNA Solve Tech Giants’ Data Issues?
While Deoxyribonuclease (DNA) is the blueprint of life, it also may hold the key to the explosive demand for data storage. ARK believes traditional magnetic storage technology solutions cannot keep pace with current demands. In a recent collaboration, Microsoft (MSFT) and the University of Washington set a record in storing 200 megabytes of data in just a microgram of liquid, smaller than the width of a sharpened pencil.

Thanks to advances in DNA synthesis and the concomitant rapid decline in cost curves, DNA’s role in research is extending beyond genomics. Storing data in DNA is not a new concept.

DNA storage is best suited for long term archives. DNA is composed of a series of four small nucleobases strung together in a double helix structure. The double helix structure is highly stable and heat-tolerant. While the smallest transistor to date is 10 um, one nucleobase is 0.5 um, which provides a large amount of storage in a small space. In addition, DNA exists in a three-dimensional space, whereas transistors have to be stacked and require expansive cooling solutions.

Initially, DNA-based storage will be limited to archival applications because random access memory (RAM) is prohibitively expensive. Based on ARK’s research, today it costs roughly $1,000 to sequence a whole human genome. “Reading” and “selecting” a file stored on DNA would be both costly and time-consuming. In addition, sequencers with long read capabilities are not precise enough to work well with binary code. Content consumers will not wait hours to play back anything these days!

6. Exoskeletons and Virtual Reality technology help restore feeling in paraplegics
A recent study conducted in Brazil—with the aim of utilizing advanced computing and robotic technology to improve the lives of paraplegics—has yielded surprising results. The trial was intended originally to test robotic aids signaled by brainpower. Unexpectedly, eight paraplegics moved their legs and regained partial sensation after 10 months of “brain training.”

See a video of Miguel Nicolelis, co-director of the Duke University Centre for Neuroengineering and part of the Walk Again Project, discussing the breakthrough here.

This training incorporated an exoskeleton, sessions using Virtual Reality (VR) technology, and a non-invasive system linking the brain with a computer. In the experiment, the use of VR technology helped patients learn how to control the exoskeleton. While the patients controlled their avatars while walking in VR, researchers studied each patient’s brain and generated a set of commands for each exoskeleton based on his or her unique neural activity. Gradually, as they weaned off full virtual reality, patients experienced changes in their states of paralysis.

ARK believes that VR will lead to some exciting health applications. While VR holds much promise for the 5.6 million people in the U.S. living with some degree of paralysis,1 ARK intends to explore its potential in many other therapeutic areas.

  1. The Christopher & Dana Reeve Foundation

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.