Tesla Projections Via Wright’s Law:
AI Future Projections:
The five key inputs that can change the model:
Tesla Projections Via Wright’s Law:
ARK Invest mentioned within an exclusive report that they expect sales to increases roughly 8-fold, from 4.8M units in 2026, to 40M units in 2030 – globally. This is roughly a 53% compound annual growth rate per-year. The reasons for the major increase within vehicles sold is driven by battery cost declines primarily. The analysts at ARK mention the use of Wright’s Law as a viable leading indication into the future of cost declines for Tesla.
Wright’s Law is a hypotheses that there is a “20% reduction in cost, when production doubles”. In so far as 1000 units have been produced, the cost per-unit decrease by 20% when production reaches 2000 units. Another 20% reduction is apparent at 4000 units, and at 8000 units, and so on. This idea states that cost decreases as a power law of cumulative production. Interestingly, in relation to innovative investments, MIT researchers have pointed towards Wright’s Law having validity, and coming closest to the truth – in comparison to other laws (for example Moore’s Law).
Importantly, it is vital to point out that whilst innovation by definition is unpredictable and unexpected, even if we can predict to some extent the level of innovation, and the areas of innovation in the future, this will have major benefits from a societal point of view. For example, in the context of GDP projections, public policies and also from an investment standpoint. If Wright’s Law has validity, this indicates towards the potential for major investment opportunities that will produce non-linear growth. Excitingly, MIT state how through Wright’s Law, there is a strong tendency across different types of technologies, not just solely within the context of manufacturing a tangible good. This has led to conclusions from commentators that there is an ability for super-exponential improvements for technologies over longer time spans because through new inventions, as cost decreases, output still rises, leading to a flywheel of innovation.
We can conspicuously see Wright’s Law in action, through the Human Genome Project. As reported by NHGRI Genome Sequencing Program, the cost of sequencing DNA bases has fallen majorly (more than 700,000 fold) since the first sequencing project. This outstanding improvement from the Genome Sequencing Program has revealed the exponentially large cost reductions that are occurring, whilst simultaneously increases within output.
When looking more closely at Tesla, ARK predict that through Wright’s Law, there will be a rough 28% cost decline in lithium ion batteries for every cumulative doubling within production. For Tesla, this means that by 2023 all EVs will reach up-front cost of prices in which are equivalent or lower than that of an ICE vehicle. This will cause a huge shift in demand in regards to EVs say the ARK analyst.
The reason as to Wright’s Law holds validity within light of innovation is based on the concept of the innovation flywheel. This is achieved when costs decline significantly, whilst output increases, leading to an ability to reinvest and peruse a flywheel of innovation.
Over time, this generates non-trivial exponential growth curves for innovation, in which can become very lucrative for investors.
AI Future Projections:
William Summerlin is an analyst at ARK who solely focuses and specialises on AI technologies. Will mentions how, he is most excited for the opportunity for foundation models. Will states that the larger AI models are, is often a good leading indicator as to the quality of performance. This is a trend that will continue, says Will. This has been seen across many other language models, and a range of other AI technologies.
What is an AI foundation model?
Foundation models work by training a single huge system on large amounts of general data, then adapting the system to a new problem. This interlinks with the future of AI in which refers to AI being flexible, reusable models in which can be applied to just about any domain or industry task. IBM mention how, “the next wave of AI looks to replace the task-specific models that have dominated the AI landscape to date”. The future is models that are trained on a broad set of unlabelled data that can be used for different tasks, with minimal fine-tuning. These are called foundation models.
Foundation models however are not trivial, and results within the real world have already been seen. For example, the first glimmers of foundation models have been shown within GPT-3, BERT or DALL-E 2. One can input a short prompt, and the system then generates an entire essay, or complex picture based on set parameters – even if it was not trained on how to execute that image or task explicitly.
Looking at the foundation model of GPT-3, which is a powerful language model, this has shown outstanding results in which give Tesla investors hope in regards to the future of FSD. The model is used to create articles, poetry, stories, news reports and dialogue using just a small amount of input text that can be used to produce large amounts of quality copy. GPT-3 has over 175B parameters and can generate any text including guitar tabs, or computer code.
“Others have found that GPT-3 can generate any kind of text, including guitar tabs or computer code. For example, by tweaking GPT-3 so that it produced HTML rather than natural language, web developer Sharif Shameem showed that he could make it create web-page layouts”.
When relating these foundation models back to Tesla, this gives ARK validity in terms of FSD being available within the future, and perhaps even becoming applicable to other domains with little friction.
Tesla has a unique data asset because the company has seen billions of miles in terms of data points within the physical world, and these data points are then relayed back into the AI system in which allows for greater learning over time in terms of AV capabilities. Tesla has an opportunity to make these models open-sourced, meaning that Tesla is likely going to allow other manufactures to build on-top of these models.
ARK mention how, based on their research this opportunity could be worth around $25T by 2030. There will be models used to solve issues that can help edit text, or generate images, however there is also a use case in terms of aiding robots moving within the physical world. Whilst this has not been explicitly stated by Tesla, given Elon Musk’s proclivity and excitement for AI within the real world, this seems like a fairly plausible statement to push forward.
One of the biggest markets within AI is automated driving says ARK. All vehicles within the future will be electric, & fully automated. Autonomous driving will majorly lower the cost to get around. For example, an automated taxi could drive a mile for as low as 25 cents. This is currently less than half of the cost of driving a new-personal car. Through this dramatic cost decline via automated capabilities, this can result in a range of previously excluded individuals gaining access to vehicle mobility at a much lower cost than present today.
“Cost will drive demand” say ARK.
On a revenue basis, ARK states that the TAM is around $11T for autonomy. ARK mentions clearly how, even if costs decline do not end-up reaching 25 cents, there is a possibility for still increased demand at cost levels of $1 per-mile.
The platform providers, and those that own the technology AI stack, these companies could enjoy an enterprise value of $11-12T by 2026. Tesla is uniquely set-up to participate within say’s ARK.
ARK mentions how their expected value for Tesla in 2026 is $4600 p/s, however this has potential to rise upward of $5,800 p/s or more by 2026.
The five key inputs that can change the model:
- Capital efficiencies (how much able to increase production y/y on a percentage basis).
- How many cars will be on the ride-hail network.
- Estimated launch for robo-taxis.
Capital efficiency is vital to understand: how much money needs to be spent for an incremental unit of capacity to be built within the automotive industry. When Tesla first started the Model 3, their capital efficiency was $84,000 per incremental unit of capacity. However, this has majorly improved – in 2021 this is down to $7,700 per incremental unit of capacity. Within ARKs upside case, this figure is likely to move downward to £2,000 per incremental unit of capacity.
By far, the single biggest driver for Tesla’s valuation is robo-taxi launches. ARK believes that around 60% of the value within the model can be attributed to the taxi launch.
ARK expect Tesla to launch FSD between 2024, & 2026 as a more conservative estimation.
