#236: Baidu’s Apollo Go Is A Smart Step Toward Launching Fully Autonomous Taxis, & More

This week, Baidu extended its free robotaxi pilot, Apollo Go, from Changsha to Changzhou and now is testing autonomous cars in more than 20 cities in China and California. Its geographic expansion suggests that Apollo Go is winning support from local authorities, particularly because the Chinese government has designated it as the nation’s autonomous driving platform.

Apollo Go still seems to use safety drivers and limits rides to 55 drop off and pick up spots in Changzhou. Although its technology appears to be years behind its US competition, Waymo and Tesla, its limited launch will help Baidu to commercialize a minimum viable product critical to launching a fully autonomous taxi service.

While incremental updates to Tesla’s Autopilot also have created a minimum viable product in the autonomous space, Waymo tried to perfect its technology before launching Waymo One to the general public. Because the collection of real-world driving data, not simulated data, will be critical to the success of autonomous driving platforms, more companies should take a leaf from Baidu’s book as they plan their long-term strategies.

The Wuling Hongguang MINI EV is a neighborhood electric vehicle priced at ~$4,000, with a limited range of 75 miles and a maximum speed of 60 miles per hour. Within a month of the MINI’s launch, the SAIC-GM joint venture received a surprisingly high number of orders, more than 50,000.

Given the pace of Tesla’s innovation as well as its vertical integration, ARK believes that traditional automakers will not be able to compete economically on standard metrics like range and performance and will have to segment the market in other ways. Competing at the low end with neighborhood electric vehicles (EVs) or electric scooters/bikes and exceedingly low prices could be effective. Unfortunately, low end vehicles will impair their share of the EV market’s profits.

One solution to this dilemma would be to outfit low end neighborhood EVs with sensors, collecting real world driving data and training autonomous vehicles for autonomous taxi networks. Another would be to offer a software layer for the management of ridesharing and/or autonomous taxi fleets. Both strategies could generate high margins and salvage the traditional automakers’ share of the economic pie.

According to three recent independent studies, CRISPR Cas9 – a novel gene editing technology – could cause unintended changes, or off-target effects, to the human genome. Among them are gene deletions or unusual rearrangements of chromosomes, with unknown consequences.

Currently, the Food and Drug Administration (FDA) is drafting regulations for gene editing therapies. With roughly 1,000 human gene-editing trials already under way, regulators probably will have to review 20 new drug applications per year to keep up.

This week, the FDA rejected BioMarin’s (BMRN) gene therapy for Hemophilia A, a severe chronic bleeding disorder, and requested more data, perhaps out of concern that off target effects could jeopardize patient safety. As a result, BioMarin’s potential cure could face a two-year delay, giving other pre-clinical Hemophilia A trials like Intellia Therapeutics’ (NTLA) an opportunity to catch up.

CRISPR Cas9’s potential is monumental. That said, regulators will have to standardize testing to identify off target effects and maximize its safety profile.

The American College of Obstetricians and Gynecologists (ACOG) recently published new guidelines on non-invasive prenatal testing (NIPT). Previously, ACOG recommended NIPT for women older than 35 as well as those with high-risk pregnancies. Now ACOG is recommending NIPT for all women regardless of age or other risk factors. In our view, these new guidelines could motivate payers to offer more NIPT comprehensive coverage. That said, a number of other barriers will continue to inhibit NIPT adoption, including access to genetic counseling, patient and physician education, socioeconomic factors, and high costs.

In 1970, the securitization of mortgages was an important real estate-related innovation that made homeownership more accessible by lowering the cost of capital. Lenders packaged pools of mortgages into mortgage backed securities, giving fixed income investors access to pooled mortgages and home buyers access to lower mortgage rates thanks to the increased liquidity associated with institutional markets. Mortgage securitization as a percentage of total mortgages rose steadily, from nil in 1970 to 61% in 1997, more than 90% during the housing bubble 15 years ago and, even in the aftermath of the housing bust, 81% today.¹

A few weeks ago, Intercontinental Exchange acquired Ellie Mae, a cloud based digital mortgage origination platform for mortgage lenders. In our view, Ellie Mae could standardize and digitize the paper based and error prone mortgage application process. Ellie Mae’s standardization and digitization should help mortgage lenders, small or large, securitize more of the ~$2.2 trillion in annual mortgages, as measured by the Mortgage Bankers Association. According to our research, in 2019 ~$404 billion of those originations were not securitized, giving Intercontinental Exchange the opportunity to facilitate more mortgage securitizations, benefitting both lenders and borrowers.

[1] Source: Mortgage Bankers Association ad SIFMA

Heading into the second post-COVID quarter, certain technology companies are gathering momentum as the world economy moves online. One company that has caught fire is Fastly, a provider of content delivery networks (CDNs), which enjoyed second quarter revenue growth of 62%, a significant acceleration from 38% in the first quarter.

Storing copies of data in access points across the world, CDN technology is more than 20 years old. Since 2000, for example, Akamai ‘s growth had decelerated into the single digits last year, partly because public cloud companies leveraged their scale to turn CDNs into commodity services.

Why is Fastly outgrowing its far larger peers? In a word—programmability. Fastly has turned “dumb CDNs” into programmable software. With a developer-first mindset, simple APIs can control Fastly’s CDNs to purge content, balance loads, and control access. Like Twilio, Fastly has turned an old technology into a programmable interface, winning developer mindshare and enabling new use cases.

Fastly’s latest product, Compute@Edge, aims to expand its product stack from CDNs to serverless computing at the edge. The technology is new and seems niche but, like many “niche” products, serverless computing at the edge could become the next big thing.