1. Coinbase And BlackRock Partner To Provide Institutional Clients Direct Access To Bitcoin
Last week, Coinbase announced a partnership with Blackrock to provide institutional clients with direct access to crypto, starting with bitcoin. At the core of the partnership will be an interconnection between Coinbase Prime and BlackRock’s Aladdin. Coinbase Prime is an institutional prime brokerage platform that offers its 13,000 clients custody and advanced trading solutions. Aladdin is BlackRock’s portfolio management software that supports nearly $22 trillion USD in assets managed by 55,000 investment professionals. Through this partnership, Aladdin’s user base will have access to Coinbase Prime’s prime broker, trading, reporting, and custody solutions. According to Max Branzburg, Vice President of Product at Coinbase, this deal could usher trillions of dollars into the crypto asset class in the coming years.
Blackrock’s decision to partner with Coinbase is a strong signal that institutions consider crypto––starting with bitcoin––a new asset class. We agree that bitcoin has earned an allocation into well diversified portfolios.
Based on daily returns across asset classes during the past ten years, our analysis suggests that an allocation to bitcoin in well diversified portfolios should range from 2.55%, when minimizing volatility, to 6.55%, when maximizing returns per unit of risk, as shown below.
In this analysis, we ran a Monte Carlo1 simulation of 1,000,000 portfolios composed of various asset classes. The efficient frontier captures the highest returns possible for a given level of volatility. The stars indicate allocations associated with maximizing the Sharpe Ratio and minimizing volatility.
Based on ARK’s simulated portfolio allocations, institutional allocations between 2.5% and 6.5% could impact bitcoin’s price by $200,000 and $500,000, respectively, as shown below.
 A Monte Carlo simulation mitigates this interference by focusing on repeating random samples to output a result. While typically more effective than relying on a single variable to forecast or estimate an outcome, our simulation assumes perfectly efficient markets and does not account for factors that are not built into the price movement, including macro trends and market sentiment. As a part of the simulation, we selected the following commonly used asset class benchmarks, analyzing their price behavior from 2011 to 2021: Real Estate – The Morgan Stanley Capital International (MSCI), US Real Estate Investment Trust Index (RMZ), Commodities – The Crude Oil Futures (CL1 COMB), Currencies – MSCI Global Currency Index, Bonds – Bloomberg Barclays US Aggregate Bond Index, Equities – S&P 500, Gold – GLD.
2. Consensus Capex Forecasts Suggest That Tesla Could Continue To Take EV Market Share
Commenting on competition in the electric vehicle (EV) space, media pundits tend to frame new EV capital spending (capex) announcements as threats to Tesla. Some point to the $35 billion and $50 billion, respectively, that GM and Ford have committed to EV production, concluding that the impact on Tesla will be negative.
ARK believes this logic is flawed for two reasons. First, new EVs are taking share from internal combustion engine (ICE) vehicle suppliers. Second, according to consensus forecasts, Tesla’s EV capital spending will surpass that of both Ford and GM by 2025, as shown below.
According to the consensus forecast, Tesla’s capex will exceed the budgets of Ford and GM in 2023 and 2025, respectively. If they do not match Tesla’s capex in 2025 and beyond, we believe Ford and GM are unlikely to keep pace with its growth. During its recent earnings call, Elon Musk added to the competitive dynamic, announcing that Tesla intends to expand its network of gigafactories from six to twelve sites during the next few years.
3. Gene-Editing Research Shows Progress In Treating Hereditary Cardiomyopathy
Intellia Therapeutics (NTLA) and Regeneron Pharmaceuticals (REGN) made history at last year’s Peripheral Nerve Society Conference when they presented the world’s first data on CRISPR Cas9 in in-vivo gene-editing therapy treating patients with Hereditary Transthyretin Amyloidosis (hATTR) polyneuropathy. hATTR is a disease in which amyloid proteins build up and, if untreated, can cause the failure of multiple organs, typically the heart (cardiomyopathy) and nerves (polyneuropathy). Notably, the researchers demonstrated that editing the TTR gene might be effective in delivering a one-time treatment for polyneuropathy that reduces serum TTR, eliminating the need for chronic therapy.
This week, Alnylam Pharmaceuticals (ALNY) reported positive Phase-3 data on its APOLLO-B study of Patisiran, an RNA interference drug for patients with hATTR cardiomyopathy. The company met its primary and first secondary endpoints––based on metrics associated with a six-minute walk and a questionnaire on quality-of-life––demonstrating the clinical benefit of TTR knockdown for cardiomyopathy.
Once approved, the important question is whether the 50,000 patients suffering from hATTR Amyloidosis will be willing to take a one-time gene edit or stick with a chronic gene therapy?
4. AI Programming Tools Are Likely To Boost The Productivity Of Software Engineers
In its recent earnings call, Microsoft announced that more than 400,000 developers subscribed to GitHub’s AI pair programming tool in its first month of general availability. Called “Copilot,” the tool is built on OpenAI’s Codex model, which uses the same large-language model framework as GPT-3.
Microsoft announced Copilot a few months after Alphabet released Gato, DeepMind’s new artificial intelligence program that has ranked 54% in coding competitions. Gato pre-trains a transformer-based language model on open-source code and fine-tunes it with samples from competitive programming datasets.
Recently, Google explained that a code completion tool combining a semantic engine with machine learning wrote nearly 3% of its new code to help engineers code more efficiently. Collectively, by using the semantic machine learning tool, more than 10,000 engineers reduced both coding iteration time and the number of context switches by 6% and 7%, respectively.
ARK forecasts that AI training costs could decline by 60% per year, boosting the performance of AI (artificial intelligence) coding models exponentially during the next five to ten years. Based on our research, AI tools will increase the productivity of software engineers more than two-fold by 2030, automating a range of tasks from coding and documenting to note-taking.