#289: DNAnexus Has Supercharged Access to the UK Biobank Database & More

Last week, the UK Biobank and DNAnexus released their Research Analysis Platform (RAP), a cloud-based tool that substantially increases access to the UK Biobank database. Aided by the National Health Service (NHS), the UK Biobank has collected genetic and longitudinal clinical data from more than 500,000 people over the last 15 years. During the next four years, it expects the already massive database to quadruple from 11 petabytes to more than 40 petabytes— equivalent to a century’s worth of HD video footage.

Although the UK Biobank has been accessible for several years, most life science researchers have had neither the cloud resources nor the expertise necessary to access it. Anecdotally, just downloading the data has taken academic research labs months at a time, highlighting the importance of the DNAnexus RAP interface. Thanks to the RAP, researchers do not need to download data on-site and can run their models in the cloud.

Based on the need to store, organize, and share large volumes of data, and according to their website, DNAnexus has become the leading genomics cloud provider. Powered and supported by Amazon Web Services (AWS), the RAP interface should provide compute capacity that is secure and scalable as the UK Biobank grows exponentially.

DNAnexus also has been a pioneer in the multi-omics analysis that integrates data across layers of molecular biology, including genomics (DNA), transcriptomics (RNA), and proteomics (protein). Last December, the UK Biobank initiated the largest blood plasma proteomics project to date. Sponsored by a consortium of large pharmaceutical companies, the Pharma Proteomics Project is tracking 1,500 circulatory proteins in 53,000 UK participants. Combining the RAP and DNAnexus’s multi-omics capabilities, researchers should be able to generate novel insights about health and disease.

Last week, Coinbase announced a direct deposit feature that will allow US customers to convert their paychecks into crypto automatically. Users will choose what percent of their paychecks to deposit with Coinbase and then the currencies into which to convert: the USD or one of the more than 100 cryptocurrencies on its platform. Coinbase will not charge direct conversion fees but will earn a spread when buying and selling the cryptocurrencies. This service will become part of a broader Coinbase roll-out as it aims to become the consumer’s “primary financial account”.

Digital wallets typically focus on the user payroll opportunity to increase recurring fund flows, user retention, and user engagement with other revenue-generating services. Recently, for example, Square has emphasized that boosting Cash App recurring inflows has been a top priority, and such deposits have been the primary reason its gross profit increased. Similarly, the Chinese government has worked with companies to integrate salary deposits into its Central Bank Digital Currency (CBDC) pilot, hoping to embed the digital yuan in users’ daily lives.

ARK believes that, with payroll direct deposit, Coinbase will increase the adoption and monetization of its trading, yield-generating, peer-to-peer, and commerce products. Moreover, by facilitating inflows into crypto, Coinbase is likely to increase user activity across the broader crypto ecosystem as well.

Last week, President Nayib Bukele announced that El Salvador is mining bitcoin with energy from its volcanos. Just three months ago, he had instructed state-owned geothermal electric company La Geoto plan for bitcoin mining and, in a tweet on Friday, not only detailed that the operation is in testing mode but also provided a breakdown of the mining revenue generated so far.

Last month, El Salvador’s government rolled out its bitcoin wallet, Chivo. According to the President, Chivo has been downloaded by 2.73 million people, suggesting a penetration rate of 75% among eligible citizens. For perspective, prior to the rollout of Chivo, only 1.9 million and 1.3 million El Salvadoran adults used bank accounts and debit cards, respectively. If correct, scaling to 2.73 million digital bearer accounts in less than 1 month, Chivo has provided unprecedented financial inclusion, a promising sign for bitcoin adoption in other emerging markets.

President Bukele stated that Chivo is settling $1 million in remittances every day. If true, according to Alex Gladstein, the state-run bitcoin app accounts for roughly 6% of El Salvador’s $6 billion in annual remittances, or more than 1% of its GDP.

While these numbers have yet to be verified, El Salvador’s experiment to date appears to be an instant success. ARK believes that over time other countries will follow El Salvador’s lead in a race to become global leaders of this new monetary world order.

As artificial intelligence (AI) has evolved during the last decade, algorithms and architectures seem to have homogenized. Recently, Stanford’s Center for Research on Foundation Models (CRFM) published a paper arguing that the homogenization trend could extend to models, with large foundation models ultimately underpinning most AI applications.

Foundation models are trained with self-supervised learning on diverse sets of data and fine-tuned to perform various tasks. Like humans, foundation models learn from a wide range of data sources and can perform new tasks after exposure to a handful of examples.

Surprisingly, foundation models often outperform specialized models designed to excel at one task. In the case of LAMBDA text completion tasks, GPT-3 scored 18% higher than the previous state-of-the-art model after being trained with just a few examples. Foundation models also are finding success in complex vision tasks such as perception, so much so that some researchers are betting that they will achieve breakthroughs in higher-order skills like commonsense reasoning.

If foundation models were to underpin future AI applications, the economic implications would be significant. Today, the cost to train a large-scale foundation model is prohibitive. Although AI training costs are falling at a rate of roughly 68% a year, the exponential growth rates of training data and model complexity are likely to sustain the cost barriers.

Thanks to economies of scale and certain data advantages, we believe a few AI leaders could capture most of the revenue opportunity, much as has been the case in the cloud infrastructure market. If so, given our expectations for the AI opportunity, the market capitalization of just a few companies could scale to $10 trillion+ each during the next 10 years. We look forward to surfacing who we believe will be the winners in the AI age.

This week, the California DMV granted Waymo and Cruise Automation permits to operate commercial autonomous taxi services, allowing both to charge customers for driverless rides. Previously, each could test driverless services without charging customers, the service was limited to participants in testing programs, mainly employees and friends. This approval signals that California regulators are supporting the rollout of autonomous taxi technology, and that Waymo and Cruise are competing to launch the first robotaxi service for the general public in that state.

ARK estimates that autonomous taxis could generate roughly $10 trillion in revenue within the next ten years, with platform operators capturing roughly $4 trillion in net revenues. If earnings were to reach half of net revenues, autonomous platform providers could scale from nil today to approximately $40 trillion in enterprise value.

Importantly, we believe consumers would be major beneficiaries as autonomous technology frees up time spent driving and lowers the cost of personal travel dramatically, resulting in more Netflix binges and fewer accidents!