ARK • Disrupt

Last week, Tempus AI signed¹ a $200 million strategic partnership with AstraZeneca and Pathos AI to build a multimodal oncology foundation model. Tempus is contributing its valuable data repository—over 8 million de-identified patient records, including clinical notes, genomics, imaging, and transcriptomics. Fueled by this data, AstraZeneca’s foundation model should unlock new drug targets, accelerate therapeutic development, and drive personalized care.

This deal signals an important shift as pharma companies advance from adopting AI models to building them. In our view, every pharma/biotech giant will have to train and finetune their own models to address their respective pipelines and strategies—a transformation as fundamental as personalized medicine itself.

At roughly $25 per patient record—$13–$53 depending on data depth²—the economics of this deal highlight the value of richly-annotated, multimodal data. Notably, 8 million records will become just the starting point of a much larger corpus. Building a robust foundation model that captures disease complexity across populations is likely to require tens or even hundreds of millions of deeply phenotyped, multi-omic, longitudinal patient journeys.

The arms race for AI dominance in healthcare has begun. The pharma/biotech companies that control and capitalize on the deepest, richest data will win.

Recently, OpenAI released³ the full version of its highly anticipated o3 reasoning model alongside o4-mini, a smaller model engineered for cost-efficiency at roughly one-tenth⁴ the price point of o3. On paper, o3 and o4-mini sit at or near the top of many public benchmarks, matching or eclipsing the performance of most of the other top models. That said, OpenAI's models are facing formidable rivals in xAI, Google, and Anthropic on marquee tests like the PhD-level science reasoning benchmark GPQA Diamond,⁵ as shown below.

Source: ARK Investment Management LLC, 2025. This ARK analysis draws on a range of external data sources as of April 25, 2025, which may be provided upon request. For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security. Past performance is not indicative of future results.

While OpenAI is no longer leapfrogging competitors as in its early days, the raw leaderboard performance tells only part of the story. OpenAI’s integrated toolset has become key to its differentiation. o3 and o4-mini, for example, have access to seamless web search, on-the-fly code generation and execution, dynamic charting, and now—for the first time—sophisticated image analysis. By chaining those capabilities together, o3 and o4-mini are leapfrogging isolated “reasoning” models and becoming versatile copilots for real-world workflows.

In our view, this tool-centric ethos will become critical to success. Model providers increasingly will embed large language models (LLMs) into robust, end-to-end products—either through internal development or strategic acquisitions⁶—to create positive data feedback loops, sharpen differentiation, and unlock sustainable monetization opportunities. In doing so, they will navigate around the increasing commoditization of LLMs with sticky, high value experiences that transcend raw performance metrics.

Last week during its first-quarter earnings call,⁷ Tesla reaffirmed that it is on track to commercialize its robotaxi service in Austin this June. Although it will start with a fleet of only 10–20 “robotic taxis,“ mostly Model Ys, on day one, the company plans to scale to millions of vehicles by the end of 2026.

Source: Tesla 2025. (https://digitalassets.tesla.com/tesla-contents/image/upload/IR/IR/TSLA-Q1-2025-Update.pdf). For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

Management also confirmed that production of its low-cost vehicle will begin by the end of this quarter. Given heightened macroeconomic uncertainty, management emphasized that affordable models have become more critical to Tesla’s strategy.

Shortly after the earnings call, Tesla announced⁸ that it has launched a live ride-hailing service for a small group of employees in Austin and San Francisco. In other words, internal robotaxi testing is underway. In our view, the launch of robotaxis in June will create a “seeing is believing” moment, encouraging analysts to rethink their financial models as they begin to understand how rapidly Tesla’s fleet could scale during the next year and a half, transforming the perception of Tesla from a hardware manufacturer into the largest embodied AI company in the world.

Source: Tesla AI 2025 (https://x.com/Tesla_AI/status/1915080322862944336). For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

Last week, ARK Invest published⁹ an updated 2030 bear, base, and bull case price forecasts for bitcoin.

Our methodology aggregates six TAMs (total addressable markets) relative to bitcoin’s mathematically metered supply growth toward 20.5 million coins by 2030. According to our research, three of the TAMs—institutional investment, “digital gold,”¹⁰ and an emerging-market safe-haven use case—should deliver most of the expected value accrual. Secondary growth levers include nation-state treasuries (El Salvador, Bhutan, and a prospective US strategic reserve); corporate treasuries (74+ public companies already hold ~$55 billion);¹¹ and accelerating on-chain financial facilitators and services like the Lightning Network and Wrapped BTC.

The forecast assumes conservative TAM penetrations in the base case and aggressive TAM penetrations in the bull case for 2030. In the base case, for example, the allocation to bitcoin accounts for 2.5% of the $200 trillion in global investable assets ex-gold in 2030, while in the bull case it accounts for 6.5%, roughly twice the allocation to gold today. Notably, “digital gold” dominates the bear scenario, accounting for 58% of bitcoin’s value, while institutional allocations drive 43% of the value in the bull case.

Finally, the last section of the piece focuses on the metric “on-chain network liveliness,” introducing the concept of “active supply” that discounts coins lost or held dormant. If only 60% of the supply were liquid in 2030, which is our expectation, the forecasted price in each scenario would increase by ~40%. While any long-range crypto forecast is highly speculative, this more experimental exercise illustrates that scarcity, capital-market accessibility, and layer-2 innovation could converge to push bitcoin past the six-figure mark within the next five years.

The price forecasts referenced above rely on assumptions made regarding the total addressable markets (TAMs) and penetration rates discussed, which are further explained in the published article. Bitcoin may fail to reach these price forecasts if any of the TAMs or penetration rates are not met. Risks and limitations exist that may prevent our forecasts from being realized.

The forecasts and price estimates herein are subject to revision by ARK Invest (“ARK”) and provided solely as a guide to current expectations. Forecasts regarding broad markets and bitcoin are not, and are not intended to be, representative of any ARK-managed investment product or the characteristics of any ARK portfolio.

FORECASTS ARE HYPOTHETICAL AND HIGHLY SPECULATIVE, AND PRESENT MANY RISKS AND LIMITATIONS. While ARK believes that there is a sound basis for the forecasts presented, they are provided for illustrative purposes only and no representations are made as to their accuracy.

Bitcoin is subject to rapid price swings, including as a result of actions and statements by influencers and the media, changes in the supply of and demand for bitcoin, and other factors. There is no assurance that bitcoin will maintain or increase its value over the long term. Bitcoin is largely unregulated and bitcoin investments may be more susceptible to fraud and manipulation than more regulated investments. Bitcoin is subject to unique and substantial risks, including significant price volatility and lack of liquidity, and theft.