Tesla Is Poised To Launch A Multi-Trillion Dollar Robotaxi Opportunity In 2025, & More

Want more research from the ARK Team? Have feedback on our publications? Click here to help inform our content creation.

1. Tesla Is Poised To Launch A Multi-Trillion Dollar Robotaxi Opportunity In 2025

By: Tasha Keeney, CFA, Sam Korus, & Daniel Maguire, ACA

Last week, Tesla delighted investors with its third-quarter earnings. The company reported better-than-expected profits, thanks to record-low vehicle costs, increased revenue from its high-margin Full Self-Driving (FSD) software, and surprisingly strong growth in revenue from energy storage and services. Tesla also reaffirmed its plans to launch a low-cost vehicle in the first half of next year that will support its 20-30% growth target for unit volume growth of its vehicles. Most important to ARK, Elon Musk confirmed that next year Tesla will launch a ride-hail service—in our view, unlocking a multi-trillion-dollar robotaxi opportunity—in Texas and California. We also learned that Tesla employees have been testing the service for quite some time in California.[1]

During its earnings call, Tesla shared ride-hailing insights that many expected to hear at the Cybercab event on October 10. Clearly, Musk made a strategic decision to defer some details important to investors until the earnings release. During the investor call, management noted that some states will require a safety driver until Tesla reaches specific time and mileage thresholds.[2] Given its large fleet, Tesla should be able to meet those thresholds handily. Even if regulators were to delay the removal of safety drivers, ARK has outlined the strategic and tactical advantages of launching a human-driven ride-hail service first.[3]

In our view, Tesla should enjoy a price umbrella at its robotaxi launch, thanks to the high level of current ride-hail prices, as shown below, while leveraging a lower cost per mile than the average vehicle on the road. Why? The operating costs associated with electric vehicles are roughly one-third those of their gas-powered counterparts.[4] Without safety drivers, Tesla has suggested that, at scale, its robotaxi rides will cost consumers only $0.30-0.40 cents per mile,[5] slightly higher than ARK’s estimate of ~$0.25 per mile but well below current ride-hail costs of ~$2 per mile and personal car ownership costs of ~$0.70 per mile.[6] Lower price points could unlock ~$11 trillion in revenue potential, ~80 times larger than the addressable market that Uber and Lyft target today, as shown below.

2. Will Genomic Screening Of Newborns Shift Healthcare From Reactive To Proactive?

By: Nemo Marjanovic, PhD

What if every baby were to be assessed at birth by whole genome sequencing (WGS) to identify health issues, including hundreds of potential genetic disorders? New studies like the

NHS Generation[7] Study in the UK and GUARDIAN[8] in the US are providing a glimpse into how WGS will shift healthcare from reactive care of the sick to a proactive, precision-based solution. How close are we to realizing that future, and what would it mean?

Today, most hospitals in the US screen newborn babies for ~30 conditions through an assessment that often misses treatable genetic disorders that surface only after children have suffered irreversible damage. The Generation and GUARDIAN studies are screening newborns for more than 200 genetic conditions, providing early diagnoses—and precious time—for early intervention, targeted treatments, and better outcomes.

WGS can flag severe combined immunodeficiency at birth, for example, leading immediately to treatments like the stem-cell transplants or gene therapies most effective in preventing immune complications. Based on data from GUARDIAN’s first 4,000 participants, the study found actionable diagnoses—conditions that standard newborn screening would have missed—in 3.7% of newborns, 92% for which early treatment already exists.

What about the economics? In neonatal intensive care unit (NICU) settings, ultra-rapid genome sequencing (URGS) has reduced healthcare costs by $14,265 per child, altering patient management in 37%[9] of the cases. In a system plagued by escalating costs and inefficiency, this model could be game-changing.

As the cost of the sequencing of whole human genomes has plummeted—from more than $100,000 ten years ago to less than $500 today and possibly ~$100-200 in the not-to-distant future—and as AI transforms data analysis, not administering WGS at birth should become unethical.

Important to note, 72% of families that GUARDIAN approached consented to WGS. Moreover, 90% opted for additional neurodevelopmental screenings.

The convergence of low-cost sequencing, rapid AI-driven diagnostics, and burgeoning genetic databases makes “P4 medicine”—predictive, preventive, participatory, and personalized care—a compelling proposition, as shown below.

Pioneering studies are demonstrating that proactive genetic screening is the next step in healthcare’s evolution. In a world where the parents of each newborn child receive his or her genetic blueprint at birth, true healthcare—as opposed to sick care—will begin with the baby’s first breath, transforming the future of medicine for generations.

3. Anthropic Has Unlocked New Ways For AI Agents To Work On Our Computers

By: Jozef Soja

Last week, Anthropic unveiled[11] an updated version of its flagship large language model (LLM), Claude 3.5 Sonnet. The new version features significant improvements in coding and reasoning skills, placing it in the top position on the challenging agentic coding benchmark SWE-bench Verified: now, it scores 49%, ~16 percentage points higher than Sonnet’s previous version.[12]

While benchmark improvement is a useful way to measure progress, productivity gains are likely to accelerate when AI companies unlock new ways for their models to interact with users and the world around them. Anthropic has done so by teaching Claude to interact directly with user computers—when given permission—enabling its model to take direct control of a computer's mouse and keyboard to use web browsers, spreadsheets, forms, and developer environments. This leap forward has expanded the number of tasks Claude should be able to automate significantly. Claude already is using its new computer vision capabilities to access multiple data sources and populate vendor request forms,[13] create and launch personal websites[14] using VSCode, and plan the logistics[15] for a sight-seeing trip to San Francisco.

As LLMs improve their abilities to navigate the user interfaces of computers and software applications autonomously, their use cases should expand from chatbots and search engines to agents capable of performing real work for users. ARK anticipates important productivity gains to materialize for most knowledge workers as AI begins to automate increasingly complicated tasks, both reliably and affordably.

Want more research from the ARK Team? Have feedback on our publications? Click here to help inform our content creation.

[1] Motley Fool Transcribing. 2024. “Tesla (TSLA) Q3 2024 Earnings Call Transcript.” The Motley Fool.

[2] Ibid.

[3] See Keeney, T. 2020. “Tesla Should Launch a Human Driven Ride-Hail Service to Accelerate Its Autonomous Strategy.” ARK Investment Management LLC. See also Keeney, T. 2020. “Could a Tesla Ride-Hailing Network Run Over Uber and Lyft?” ARK Investment Management LLC.

[4] This ARK analysis is based on a range of underlying data from external sources, which may be provided upon request.

[5] Tesla. 2024. “We, Robot | Tesla Cybercab Unveil.” YouTube.

[6] The market served by Uber, for example. This ARK analysis is based on a range of sources, including Moye 2024 and Grauer 2023, which may be provided upon request. See Moye, B. 2024. “AAA Your Driving Costs: The Price of New Car Ownership Continues to Climb.” AAA Newsroom. See also Grauer, J. 2023. “How much does Uber cost per mile?” Daily Dot.

[7] NHS England. 2024. “First newborn babies tested for over 200 genetic conditions as world-leading study begins in NHS hospitals.”

[8] Ziegler, A. et al. 2024. “Expanded Newborn Screening Using Genome Sequencing for Early Actionable Conditions.” JAMA.

[9] Kingsmore, S.F. et al. 2024. “Rapid genomic sequencing for genetic disease diagnosis and therapy in intensive care units: a review.” NPJ Genomic Medicine.

[10] Ibid.

[11 Anthropic. 2024. “Introducing computer use, a new Claude 3.5 Sonnet, and Claude 3.5 Haiku.”

[12] SWE-Bench. 2024. “Can Language Models Resolve Real-World GitHub Issues?”

[13] Anthropic. 2024. “Claude | Computer use for automating operations.” YouTube.

[14] Anthropic. 2024. “Claude | Computer use for coding.” YouTube.

[15] Anthropic. 2024. “Claude | Computer use for orchestrating tasks.” YouTube.