Tesla Does Not Have a Demand Problem

Follow Sam on Twitter @skorusARK

Like most weeks, Tesla was in the headlines this week, once again for Elon Musk’s tweets and the company’s delivery push at the end of the quarter. Though the two may seem unrelated, Tesla’s does drive sales leads primarily by giving the media reasons to cover it and by word of mouth.

Every few months bears argue that the demand for Tesla’s cars has been saturated.  Based on previous research, we concluded that Tesla’s decision to lower the sticker price of the Model 3 doubled its addressable market. This week we analyzed how auto manufacturers use advertising to convert potential customers into buyers. The chart below shows the ad dollars spent on traditional media per vehicle sold. Of the top selling car brands Toyota and Honda were the most efficient, with ad spend per vehicle sold of $273 and $261, respectively, both well above Tesla’s $0. Even if Tesla were to spend twice as much as Toyota to generate demand for 500,000 cars, the ~$250 million would be only 30% and 14% of the $823.6 million and $1.74 billion that analysts are projecting for its free cash flow this year and next year, respectively.

Tesla still has plenty of levers other than traditional advertising to pull. Besides social media and word of mouth, for example, it recently refreshed its referral program. That said, in the long-term ARK wouldn’t be surprised to see Tesla advertise to support its brand and to highlight the safety-related benefits of Autopilot, a major competitive advantage.

Novel Tools Can Lessen the Cost of Cancer Diagnostics

Follow Simon on Twitter @sbarnettARK

Because Next-Generation DNA sequencing (“NGS”) instruments are not always accurate, engineers have introduced hardware updates that not only reduce error rates but also cut the costs. Recently, several studies have illustrated that sequencing errors also can be corrected with bioinformatics tools, the benefits of which are likely to benefit cancer diagnostics first.

Typically, cancer mutations show up in traces, forcing clinicians to “find a needle in a haystack”. In such cases, sequencing error rates are similar to the variant allele frequency (“VAF”), making it difficult to separate biological signals from noise. As a result, researchers have to increase the intensity of sequencing, and the cost, of these diagnostics.

Source: https://www.nature.com/articles/s41467-019-09027-x

The figure above illustrates that, with less coverage and lower costs, bioinformatics tools and software updates have given clinicians greater confidence (“F1-Score”) in the results of next generation sequencing. As shown below, the results are more pronounced with liquid biopsies. To surpass the 95% clinical sensitivity threshold, more sequencing intensity, or coverage, is necessary when the VAF in the bloodstream is extremely small (<0.1%). By correcting errors in silico, with computing power and algorithms, tests that once were cost-prohibitive begin to make sense.

Source: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5680209/pdf/41467_2017_Article_1470.pdf

Google’s Stadia Game Streaming Service Has Great Tech But Few Games

Follow James on Twitter @jwangARK

Following in the footsteps of music, video, and enterprise computing, gaming is moving to the cloud. Running computer games in a cloud data center eliminates the need for expensive gaming consoles and PCs, opening up gaming to the mass market. The most popular console in history—Playstation 2—has sold roughly 160 million units to date, reaching not even 5% of internet users worldwide.

Google’s new Stadia game streaming service is its attempt to go mainstream with high quality video games. Powered by Google’s own data centers and a custom AMD GPU, Stadia will offer twice the graphics performance of existing consoles. Leveraging Google’s cloud computer expertise and its partnerships with internet service providers, Google aims to minimize latency and provide fast 4K gaming with a connection speed of only 30 Mbps.

Cloud-based game streaming is not a new idea: On Live, Gaikai, and Nvidia all have attempted it. Sony’s Playstation Now is the most mature offering, streaming more than 700 games for a $20 monthly subscription.

While well positioned to solve the technical and infrastructure problems, Google needs to sell streaming games as a subscription, a business model problem. Moreover, it has no catalogue. Stadia is a new console launch - albeit in the cloud - that will require years of attracting developers, promoting games, and acquiring customers, a task that both Sony and Microsoft started more than a decade ago. Stadia’s ultimate success will depend as much on Google’s business savvy as its cloud prowess.

Jack Dorsey Announces Square Crypto 

Follow Yassine on Twitter @yassineARK

On Twitter this week, Jack Dorsey announced that Square will hire “3-4 crypto engineers and 1 designer to work full-time on open source contributions to the bitcoin/crypto ecosystem.”  Square Crypto will be an effort to give back to the open source community.

Asked about Square’s commercial interests in the crypto community, Jack responded, “As long as we’re in business, [Square Crypto] will not be led or incentivized by our business incentives.”In addition, Square Crypto will be the “first open source initiative independent of Square’s business objectives. These folks will focus entirely on what’s best for the crypto community and individual economic empowerment, not on Square’s commercial interests. All resulting work will be open and free.” While supporting the open-source community is aligned with Square’s long-term success, Jack’s decision to distinguish between supporting open source and generating profits has resonated in the crypto community.

To date, few crypto companies have hired full time open source contributors. Square seems willing to forego short term profits to gain what it believes to be significant returns long term, as Bitcoin’s success will be Square’s success.

Fintech Companies are Leveraging Alternative Data

Follow Max on Twitter @mfriedrichARK

Peer-to-peer lending companies such as LendingClub (LC) differentiate themselves from traditional lenders with alternative data and artificial intelligence in assessing the credit risk of potential borrowers. Alternative data can include consumers’ payment history (e.g. utility, rent, phone, alimony), cash-flow data from bank accounts such as salary and cash withdrawals, credit card transactions, medical and insurance claims, personal data such as education and degrees, data from social networks, other online data such as loan applications and geographic location.

A recent study by the Federal Reserve Bank of Chicago suggests that, by leveraging alternative data sources, LendingClub can serve borrowers with low credit scores at lower interest rates than those charged by traditional banks. LendingClub itself believes it has saved borrowers $2.4 billion in loan costs since its inception.

Over the years, LendingClub has optimized its credit risk assessment model with more alternative data, lowering its reliance on traditional credit metrics such as the FICO score. It has decreased the correlation between its proprietary grades and FICO scores from more than 75% in 2007 to 35% in 2015, as shown below. With more data, presumably it is lower today.  

In 2007, LendingClub’s lowest quality grade was comprised almost entirely of individuals with low FICO scores and its highest quality grade of those with high FICO scores. Over time, however, many individuals with low FICO scores moved up in LendingClub’s grade system, accessing loans with lower interest payments. In 2015, LendingClub was able to offer many borrowers with FICO scores below 680 credit terms similar to those with FICO scores over 750.  Data made the difference.