#267: Apple Is Expanding Its M1 Product Lineup, & More

This week, Apple hosted its first event of 2021 and debuted products incorporating M1, the first silicon it has designed and developed in-house for the Mac. Dubbed “Spring Loaded”, the event showcased new M1-powered iMac and iPad Pro models which will join the MacBook Air, MacBook Pro, and Mac Mini in supporting M1.

The ARM-based M1 delivers impressive power and efficiency. By combining key components like the CPU, GPU, Neural Engine, and Memory on a single System-on-a-Chip (SoC) and then vertically integrating hardware and software, Apple claims CPU and GPU performance 3.5x and 6x faster, respectively, and battery life twice as long, compared to previous-generation Macs. With smaller and easier-to-cool packages, Apple will release the thinnest iMac ever, and power an iPad with a desktop-class CPU that will outperform the highest end desktop competition in many tasks.

In our view, its new products signal that Apple is moving quickly on a two-year plan to transition the Mac line onto its own silicon. Having powered Apple computers with its x86 architecture since 2005, Intel could be the big loser.

In other news this week, Intel reported a 20% decline in datacenter sales for the first quarter, purportedly because hyperscale players like Amazon and Facebook are going through a period of inventory “digestion”. AMD and Nvidia, its competition, also could be playing a role. Last week, for example, Nvidia announced Grace, a new ARM-based datacenter CPU.

Short-read sequencing (SRS) has dominated the molecular diagnostics industry by lowering the barriers to accessing genetic information significantly. SRS platforms, such as Illumina’s (ILMN) NovaSeq 6000, are accurate, highly scalable, and enormous data generators.

Now that the clinical community is capitalizing on their speed and scale, ARK believes that the shortcomings of SRS-generated data are becoming more obvious. BRCA1 and BRCA2, for example, have been two of the most widely studied hereditary genes associated with cancer. After decades of SRS, newer technologies are surfacing novel and clinically relevant findings associated with BRCA genes. Last year, Mary-Claire King, who discovered BRCA’s link to breast cancer, applied long-read sequencing (LRS) and CRISPR/Cas9 targeting to BRACA1. In doing so, she discovered a hidden, pathogenic structural variant (SV) within the non-coding (intronic) region of BRCA1. King’s team remarked that SRS is “…of limited use for identifying complex insertions and deletions and other structural [variants].”

Despite a decade of bioinformatic innovation, SRS appears ill-equipped in identifying structural and smaller variants obscured in dark corners of the genome. King notes that “…complex mutations have thus far been rarely encountered because they are difficult to detect with existing approaches.”

State-of-the-art clinical labs have invested aggressively in attempts to improve upon SRS data, to varying degrees of technical success. In a recent publication, scientists at Color Genomics described how they enhanced the detection of SVs in routinely tested genes. They combined a novel target capture method with a suite of eight open-source and proprietary bioinformatics algorithms. While the approach increased SV sensitivity, the authors concluded that LRS could have resolved variations within the myriad regions not accessible to SRS.

In our view, the bias and lower resolution associated with SRS has forced researchers into complex clinical workflows that cannot match the comprehensiveness and standardization that LRS offers.  Indeed, one leading clinical lab has underwritten the R&D risk of creating an ultra-high-throughput LRS platform to increase accuracy, eliminate cost centers, and simplify downstream analysis. In our view, at some point in the not-too-distant future, clinicians will be focused not on the higher costs associated with LRS but instead on the opportunity costs associated with SRS.

This week Meituan raised $10 billion to fund its last-mile autonomous delivery minivan and drone program. After nearly a year of testing, Meituan is launching its next generation autonomous delivery minivans in Beijing’s Shunyi district.

The latest robo-delivery vans can move up to 150 kg and 135 gallons at a top speed of 12 mph. From roughly $50,000 today, Meituan expects their cost to drop below $20,000 at scale during the next three to five years. Because existing data indicates that drivers account for more than 90% of Meituan’s food delivery costs today, if one van were to displace two drivers, each earning $600 per month, costs could drop more than $14,000 a year, suggesting a payback period of one and a half years at scale.

Reportedly, Meituan has been testing a drone delivery food service in Shenzhen since January. Thus far, the report indicates it has delivered more than 1,000 orders. At scale, drones should enable point-to-point delivery in a two-mile radius within 400 seconds, or fewer than seven minutes.

In our view, Meituan is challenging competitors like JD Logistics and Alibaba’s Cainiao in the last-mile autonomous delivery race. While wider adoption will depend on regulatory approval by district, slower moving robo-delivery vans without passengers probably will have to clear lower safety hurdles than robotaxis that are transporting passengers.