In late October 2025, 1X Technologies opened pre-orders for its NEO home robot and ran through its entire first-year production slot, 10,000 units, in five days. A few months later, Figure took its BotQ factory from one robot a day to one robot an hour, a 24-fold jump in under four months. Tesla, meanwhile, announced it is ending Model S and Model X production at Fremont and converting that line to build its Optimus humanoid.

Unlocking labor at scale, across warehouses, assembly lines, and homes, requires robots that can generalize across tasks rather than execute a single routine. That is what the industry is now building toward, and ARK puts the addressable market for this shift at ~$26 trillion, split almost evenly between industrial and household labor.

The Companies Building The Future

ARK follows both the incumbents that dominate today’s specialized robots and the pioneers building generalizable platforms.

• Tesla (Optimus) - Tesla is running its robot program on the same vision and compute stack that powers Full Self-Driving (FSD). The Gen 3 hand alone carries 22 degrees of freedom, twice that of its predecessor. The company plans to reveal Optimus V3 in summer 2026, the first version engineered for volume manufacturing, with production on the converted Fremont line to follow.

• Figure AI - Figure builds its robots in-house at BotQ, its high-volume manufacturing facility, and powers them with Helix, its proprietary vision-language-action (VLA) model. Every robot in the field continuously sends real-world data back to improve Helix, enabling fleet-wide over-the-air updates that make the entire group smarter with each deployment. Recently, the company live-streamed its Figure 03 robots completing what began as a planned 8-hour logistics shift, which ran autonomously for over 200 hours straight, sorting nearly 250,000 packages with zero hardware failures.

• Apptronik (Apollo) - Apptronik’s Apollo is already working in cordoned-off zones at Mercedes-Benz, GXO Logistics, and Jabil, performing real industrial tasks such as sorting, material transport, kitting, and lineside delivery. The company benefits from a major strategic partnership with Google DeepMind, which integrates advanced AI models, including Gemini Robotics, to accelerate Apollo’s embodied intelligence and real-world reasoning. With deep roots in NASA robotics expertise, including work on the Valkyrie platform, Apptronik places strong emphasis on safety, reliability, and industrial readiness.

• Physical Intelligence (π) - Physical Intelligence is building the “brain” layer for the robotics industry rather than its own hardware. The company develops general-purpose vision-language-action (VLA) models, with its latest, π0.7, demonstrating impressive open-world generalization, enabling robots to perform tasks they were never explicitly trained on.

• Unitree - China’s Unitree has emerged as one of the world’s highest-volume humanoid manufacturers, shipping over 5,500 units in 2025. It has achieved profitability since 2024 as per the prospectus filed for its Shanghai market initial public offering (IPO) in March 2026. The company is targeting up to 20,000 humanoid deliveries in 2026, highlighting China’s significant lead in cost efficiency and manufacturing scale. The units shipped so far skew towards research, development, and entertainment purposes. The breakthrough driving the shift.

What changed is how robots are programmed. Older industrial arms run fixed routines. Hand a part at an unfamiliar angle or an unknown object, and it stalls. The new platforms run on vision-language-action models: they take in camera input alongside a plain-language instruction and produce motion directly. They learn from human demonstrations and fleet data rather than pre-engineered motions. That ability to generalize is what finally opens up messy, unstructured environments for robots.

No ready corpus of this data exists, which makes it a near-term bottleneck, though start-ups around the world are tackling it. The real difficulty is the path to collecting it, and there are two. Field deployment is circular: a robot has to run at low failure rates before it starts logging useful data. Simulation only helps once there is enough real data to close the gap between simulated and real motion, which is why the parallel race is to build world models accurate enough that skills learned in simulation hold up on real hardware. Either way, the compute requirement only matters if one of these paths keeps the data flowing.

Source: ARK Investment Management LLC, 2026, based on data from Google Deepmind 2023, Humanoid 2025, and Figure AI 2025 as of December 18, 2025. In addition to those sources, certain information presented may be the result of ARK’s internal analyses, which draw on various additional sources of information. 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. Forecasts are inherently limited and cannot be relied upon.

The question now is how fast. Humanoid work is significantly harder than driving. ARK estimates the complexity ratio at around 200,000 times that of a robotaxi, accounting for degrees of freedom, balance, perception, and task variety. Tesla’s FSD shows what sustained compute investment can do for a hard perception-and-control problem. By mapping the compute behind FSD’s performance gains and extending the curve, ARK projects that Optimus could reach human-level task performance by the end of the decade, assuming compute infrastructure and robot hardware keep improving.

Note: *Cumulative AI Compute Units are defined as the total number of NVIDIA H100-equivalent compute units, benchmarked to H100 performance at launch, required to solve a given task. “MW”: Megawatts. Source: ARK Investment Management LLC, 2026, based on data Tesla 2025a, Tesla 2025b, and Tesla 2024, as of December 18, 2025. In addition to those sources, certain information presented may be the result of ARK’s internal analyses, which draw on various additional sources of information. 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. Forecasts are inherently limited and cannot be relied upon

What We Believe The Skeptics Get Wrong

Our view differs from the popular concern that robots this capable will displace large numbers of workers. From about 1950 to 2000, automation and productivity climbed while the US labor market kept expanding. Labor force participation only began to diverge after 2000, and that is mostly attributable to aging demographics and globalization rather than to machines replacing workers. As productivity rose, each hour of human work became more valuable, and people produced more while working fewer hours. ARK’s broader bet is that general-purpose robots will spin up entirely new job categories, the way earlier technologies did, even if we can’t picture them yet. The displacement risk is real and worth taking seriously, especially through short transitions, but ARK’s research points to augmentation as the dominant effect over time.

Source: ARK Investment Management LLC, 2026, based on data from U.S. Bureau of Labor Statistics 2025a, 2025b, and Organization for Economic Co-operation and Development 2023, as of December 18, 2025. 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. Forecasts are inherently limited and cannot be relied upon.

What To Watch Next

A few markers to watch over the next eighteen months:

• Tesla’s V3 ramp - The reveal is the easy part. The open question is whether Tesla can move from a deliberately slow first run to real volume on a brand-new line.

• Commercial conversion - A number of companies have pilots running inside partner facilities. Watch whether 2026 pilots turn into scaled, paid orders in 2027.

• The first home data - NEO deliveries this year will be the first real test of a consumer humanoid in actual houses, including how often it still needs a human in the loop.

• China’s output - Nearly 90% of humanoid robots sold globally in 2025 were Chinese. Unitree is already selling robots at price points well below Western competitors, the same cost dynamic that reshaped elective vehicles (EVs).

The Bottom Line

For sixty years, getting more out of a robot meant narrowing its environment. The bet now is the reverse: one platform that learns many jobs, in any setting, and improves with every unit deployed. If that holds, robotics stops being a line item in a handful of factories and starts looking like infrastructure across the economy.

Subscribe To ARK Research

Important Information

ARK Investment Management, LLC (“ARK”) may hold a financial interest in the companies discussed in this article through various strategies and investment vehicles it manages. Readers are urged to use caution when considering the forecasts and other forward-looking information provided in this article, as it is inherently subjective and reflects ARK’s inherent bias toward positive expected results. There is no guarantee that actual results will align with the forecasts, and they might not be predictive. The information provided in this article is for informational purposes only and does not constitute investment advice. All statements made regarding companies or securities are strictly beliefs and points of view held by ARK and are not endorsements by ARK of any company or security or recommendations by ARK to buy, sell or hold any security. Historical results are not indications of future results.

Certain of the statements contained in this material may be statements of future expectations and other forward-looking statements that are based on ARK’s current views and assumptions and involve known and unknown risks and uncertainties that could cause actual results, performance or events to differ materially from those expressed or implied in such statements. ARK assumes no obligation to update any forward-looking information contained in this material. Certain information was obtained from sources that ARK believes to be reliable; however, ARK does not guarantee the accuracy or completeness of any information obtained from any third party.

2025 marked an inflection point in AI-powered productivity, as AI agents evolved from something you could chat with to something that can perform work on your behalf. This advancement in capabilities was driven by an explosive expansion in the reasoning capabilities of large language models (LLMs), starting with nascent reasoning models like OpenAI’s o1 model in the tail end of 2024, as well as the maturation of agent harnesses. These harnesses, such as Claude Code and OpenAI Codex, include the tools, data integrations, and scaffolding that models need to escape the chatbot interface and begin performing real work.

As a result of these improvements, the length of tasks AI agents could automate inflected in 2025. At the start of the year, leading models could only autonomously perform tasks that would take a human 12 minutes to complete on their own. By the end of the year, that number had grown over five-fold to 66 minutes, and in the first three months of 2026 the boundary has been pushed yet again to nearly 90 minutes.

Source: METR, as of May 12, 2026, ( https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/ ). For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

Capability is scaling exponentially, costs for commoditized workloads are collapsing, revenue is compounding at unprecedented rates, and we believe the downstream economic opportunity is measured in trillions.

Leaders Are Driving Innovation Across The Stack

Anthropic + OpenAI

Anthropic is the frontier AI lab behind the Claude family of models. The success of their dedicated coding product, Claude Code, and their general knowledge work platform, Claude Cowork, has led to explosive usage and revenue growth. Their annualized revenue grew at an 850% compound annual growth rate from $0.1 billion[1] at the end of 2023 to $9 billion by the end of 2025, before more than tripling in the first three months of 2026 to $30 billion.[2] The success of Claude Code and Claude Cowork demonstrates that embedding a highly capable model into a harness that can create and edit files, learn and remember skills and processes, and integrate with users’ data sources drives enterprise adoption far more rapidly than progressing model intelligence alone. OpenAI has been quick to refocus its efforts on enterprise tasks in response, with competing products like Codex, and will likely release a competitor to Claude Cowork soon to maintain its strong frontier position. While there is significant room for models to continue to improve in their innate capabilities, we think broad productivity unlocks will continue to be driven by the tight integration of these models into enterprise workflows.

Palantir

While tools like Claude Cowork serve as an agent harness for individuals, Palantir’s platform acts as a combined agent harness and operating system for enterprises, allowing agents to act on an enterprise’s actual operational data. Their products have been battle-tested via their success with the U.S. Government but have since seen significant traction on the commercial front, with year-over-year commercial revenue growth within the U.S. reaching 137%[3] in the fourth quarter of 2025. Palantir threatens to disrupt many legacy one-size-fits-none software products currently used by enterprises, by enabling those enterprises to efficiently create custom software products and automations to manage and streamline their core business processes.

Cloudflare

Cloudflare provides infrastructure and security products for over 20%[4] of the internet and has increasingly evolved into an execution layer for AI agents. Their edge inference capabilities, observability and security tools, storage products, and ability to block or monetize agentic traffic make them a versatile player as AI agents drive increased web traffic and demand for localized inference. The interactions between AI agents and the internet are still in early stages, and we believe that companies that provide security, guardrails, and monetization strategies for those interactions are well positioned moving forward.

Declining Costs And Improving Performance Are Fueling Adoption

Large language models are becoming more performant, and the ecosystems built around them have encouraged exponential growth in usage. OpenRouter, a unified application programming interface (API) through which developers can access LLMs, saw weekly tokens inferenced on their platform increase 25-fold between December 2024 and the end of 2025. Since then, in the first few months of 2026, usage on their platform has tripled again[5]. Drivers of this usage are twofold. As AI tools have become more performant, the price for frontier capabilities has remained relatively flat, offering enterprises and individuals significantly higher return on their investments in frontier AI tools compared to a year ago. Second, while costs for best-in-class models have remained steady, the price for what would have been cutting edge a year ago has collapsed. In many cases, the price for capabilities that would have been considered frontier last year has since fallen over 90%.

Source: ARK Investment Management LLC, 2026, based on data from Artificial Analysis as of 12/17/2025. For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

This reduction in price for now-commoditized capabilities is making it easier for AI applications to proliferate across enterprises and consumers in ways that may have been prohibitively expensive in prior years. This has allowed us to witness the Jevons paradox in action: as the cost for relative performance has fallen, both usage and spend on OpenAI and Anthropic’s products has grown accordingly.

*Note: ARR: Annualized Revenue Run Rate. OpenAI generally reports net ARR, after revenue share with those reselling its models. Anthropic generally reports gross ARR, before a cut is taken by hyperscaler partners. Source: ARK Investment Management LLC, 2026, based on data from OpenAI, Anthropic, as of May 2026. In addition to those sources, certain information presented may be the result of ARK’s internal analyses, which draw on various additional sources of information. 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.

Will Intelligence Become Commoditized?

A common pushback we hear on the AI productivity thesis is that intelligence itself is on a path to full commoditization. Price per token has collapsed over 90% in a year for some models, and Chinese labs are catching up on paper. The concern is that the economics of leading AI labs and AI-native software companies will erode in a race to the bottom.

We think this view conflates two different markets. Commoditization is already underway at the trailing edge of capability, where 6- to 12-month-old models are dropping into single-digit cents per million tokens. But even as those costs fall relative to frontier performance, enterprises are increasingly willing to pay for cutting-edge reasoning as performance gains outstrip the cost of increased inference budgets. More importantly, the durable value in this ecosystem sits in the harness, not just in the model. Claude Cowork, Palantir’s AIP, and tools like Cursor are defensible because they embed into workflows and data that are difficult to rip and replace. Even in a scenario where models themselves face commoditization, a software market growing from roughly $1.4 trillion today toward a potential $3–$13 trillion by 2030 offers significant returns for the application and harness layers.

2026 And Beyond

The next leg of this story hinges on two linked developments. First, agents are continuing to scale the length of work they can complete autonomously. Today’s ~90-minute threshold is already reshaping how knowledge workers approach their days, but crossing into multi-hour and eventually full-workday autonomy would be a fundamental shift. The METR benchmark has not yet shown signs of plateau, and upcoming model releases, such as Anthropic’s Mythos model and OpenAI’s GPT 5.5, suggest the trajectory remains intact.

Second, enterprise adoption should broaden as that capability ceiling rises. Early commercial traction has been concentrated in verticals like software engineering and customer service. We expect the next wave to expand more broadly across knowledge work, including finance, healthcare operations, and legal workflows, where agent harnesses with strong data integration and enterprise-grade guardrails are required. Platform revenue durability at the AI labs, continued adoption of Claude Cowork and competing offerings, and vertical ARR ramps should be the leading indicators of how quickly this adoption curve steepens.

In 2025, we witnessed AI agent capability begin to catch up to the early ambitions of the labs that created them. In 2026 and beyond, we expect the reverse, as our ambitions expand and catch up to unexpected capability unlocks. Every knowledge worker is on track to have a digital coworker working in parallel with them and getting measurably smarter every quarter. As agents become increasingly integrated into our daily lives, ARK believes we are entering a new period of economic opportunity and growth.

[1] SaaStr, “How Anthropic Rocketed to $4B ARR”

[2] Anthropic, “Anthropic expands partnership with Google and Broadcom for multiple gigawatts of next-generation compute,” April 2026.

[3] Palantir, “Q4 2025 Earnings Release,” February 2026.

[4] Cloudflare, “Cloudflare, Inc. Q4 2025 Earnings Call,” February 2026

[5] OpenRouter, “LLM Rankings,” May 2026

Important Information

ARK Investment Management, LLC (“ARK”) may hold a financial interest in the companies discussed through various strategies and investment vehicles it manages. Readers are urged to use caution when considering the forecasts and other forward-looking information provided in this article, as it is inherently subjective and reflects ARK’s inherent bias toward positive expected results. There is no guarantee that actual results will align with the forecasts, and they might not be predictive. The information provided in this article is for informational purposes only and does not constitute investment advice. All statements made regarding companies or securities are strictly beliefs and points of view held by ARK and are not endorsements by ARK of any company or security or recommendations by ARK to buy, sell or hold any security. Historical results are not indications of future results.

Certain of the statements contained in this material may be statements of future expectations and other forward-looking statements that are based on ARK’s current views and assumptions and involve known and unknown risks and uncertainties that could cause actual results, performance or events to differ materially from those expressed or implied in such statements. ARK assumes no obligation to update any forward-looking information contained in this material. Certain information was obtained from sources that ARK believes to be reliable; however, ARK does not guarantee the accuracy or completeness of any information obtained from any third party.

Fully driverless vehicles are already ferrying paying passengers in the US, China, and the United Arab Emirates (UAE)—and between commercial operations and active testing, robotaxis are now present in more than 50 cities across the globe.[i] The question is when do they scale? Waymo, which leads in commercial robotaxi miles, operates just 3,000 vehicles.[ii][iii] Even with that small fleet, Waymo is already pressuring Uber’s and Lyft’s market share in San Francisco within its operating zone. As autonomous technology moves from pilots to production volumes, the economics of personal transportation should fundamentally change.

The US: Waymo Leads, Tesla Could Be First To Scale

Waymo leads in commercial driverless mileage, delivering over 500,000 paid rides per week across 11 US cities.[iv] Backed by a $16 billion funding round in February 2026 that values the company at $126 billion, Waymo is targeting 1 million rides per week by year-end and planning expansion into more than 20 additional cities, including its first international launches in London and Tokyo.[v] Waymo has paved the regulatory pathway for autonomy in many US cities and demonstrated that its vehicles are safer than human drivers.[x] Waymo is positioned as a technology supplier, partnering with automakers like Hyundai, Geely, and Toyota for vehicles and fleet operators including Uber, Lyft, Moove, and Avis in addition to its own platform.

Tesla is taking a fundamentally different approach: vision-only autonomy running on consumer hardware without lidar or radar. Its Austin robotaxi pilot launched in June 2025 with safety monitors, and earlier this year, some vehicles began operating fully unsupervised. While Tesla’s robotaxi fleet is roughly 500 vehicles today, its vertically integrated manufacturing could allow it to scale robotaxi operations faster than competitors. For perspective, Tesla produces more vehicles in a single day than Waymo has in its entire fleet.[xi] Tesla’s vertical strategy has cost advantages as well - ARK’s research suggests the Cybercab’s incremental cost-per-mile could be roughly 50% cheaper than Waymo’s 6th-generation robotaxi.[xii] Tesla plans to be in dozens of US cities by year-end.[xiii]

China: Expansive Rollouts Abroad, Fierce Price Competition At Home

Baidu’s Apollo Go has deployed over 1,000 vehicles across 22 cities, logging more than 17 million cumulative rides.[vii] Pony.ai operates nearly 1,200 vehicles across four major Chinese cities and the UAE, and WeRide runs a fleet of over 1,000 vehicles in more than 30 cities across 11 countries.[viii]

Chinese operators face a challenge their US counterparts do not: fierce price competition with local ride-hail. Chinese ride-hail prices are already a fraction of US levels — roughly $0.50 per mile compared to $2.80 in the US.[xiv] As a result, many Chinese players are increasingly looking abroad. Baidu has partnered with UAE-based Autogo to bring Apollo Go to Abu Dhabi, and is testing in Switzerland and the UK.[ix] WeRide has achieved fully driverless commercial operations in Abu Dhabi and is testing with Grab in Singapore. Pony.ai is expanding into Hong Kong and the UAE. The Middle East is increasingly a strategic growth region for Chinese robotaxi operators - governments there are actively courting autonomous vehicle operators, offering favorable regulatory environments and direct investment.

Most Of Today’s Auto Fleet Will Be Rendered Obsolete

Today, a human-driven ride-hail trip in the US costs roughly $2.80 per mile. ARK projects that robotaxis could reach $0.25 per mile at scale — cheaper than owning a car and roughly one-tenth the cost of current ride-hail pricing. At that price point, it’s possible that incremental buyers will instead forgo personal car ownership in urban areas. Running at high utilization rates, only 24 million robotaxis — under 10% of the current US registered vehicle fleet — could cover all US urban miles traveled.

Who Captures The Economics?

ARK’s research suggests that autonomous technology platforms will capture the vast majority of the ~$34 trillion in enterprise value by 2030, as shown below.[xv] Americans spend 96 billion hours per year behind the wheel, time that our analysis of transportation economics literature values at roughly $3 trillion per year in US urban markets alone.

Note: Numbers are rounded. ARK continuously adjusts its forecasts based on latest available information and technological and industry developments. Updates this year include giving the fleet owner category a higher multiple and adjusting the adoption curve for autonomous driving adoption based on updated forecasts from market leaders. “EBIT”: Earnings Before Interest and Tax. “Autonomous Technology Providers”: refers to companies that develop and provide autonomous driving software, such as Waymo. Source: ARK Investment Management LLC, 2026, based on data from AAA 2025, Capital IQ 2025, and New York City Taxi & Limousine Commission 2014.
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. Forecasts are inherently limited and cannot be relied upon.

The primary risk to ARK’s forecast is how quickly automakers and partners are able to scale robotaxi fleets and service operations. State level regulation is a gating factor on the speed of rollout today, although it’s possible that there could eventually be Federal guidelines.

The auto industry took a century to put 280 million vehicles on American roads. Robotaxis could make most of them unnecessary within a decade—making our cities safer and bringing affordable ride-hail to everyone.

Subscribe To ARK Research

[i] ARK estimate based on aggregated city counts across major operators: Waymo (11 US cities), Baidu Apollo Go (22 Chinese cities), WeRide (30+ cities across 11 countries), Pony.ai (4 Chinese cities plus UAE), Tesla (2 US cities), Zoox (Las Vegas). Sources: TechCrunch, March 2026; Rhode Island Chronicle, citing fleet and city data for Chinese operators; Reuters factbox on global robotaxi deployments, December 2025.

[ii] TechCrunch, “Waymo’s skyrocketing ridership in one chart,” March 2026.

[iii] NHTSA data via TechCrunch: Waymo reported 3,067 robotaxis equipped with its 5th generation self-driving system as of December 2025.

[iv] TechCrunch, “Waymo’s skyrocketing ridership in one chart,” March 2026; TechCrunch, “Waymo opens robotaxi service in Nashville,” April 2026.

[v] TechCrunch, “Waymo raises $16B to scale robotaxi fleet internationally,” February 2026; Bloomberg, “Waymo Co-CEO Outlines Path to 1 Million Weekly Trips in 2026,” February 2026.

[vi] Basenor, “Tesla Robotaxi Expansion: 60 Model Ys Spotted in Phoenix,” April 2026.

[vii] Rhode Island Chronicle, citing Baidu fleet and ride data.

[viii] WeRide and Pony.ai fleet data from same source as [vii].

[ix] Reuters factbox on global robotaxi deployments, December 2025: Baidu/Autogo Abu Dhabi partnership, Baidu/Swiss Post Switzerland partnership, WeRide/Grab Singapore testing.

[x] Waymo Safety Impact Data Hub. Based on peer-reviewed analysis across 170+ million rider-only miles; see also Kusano et al. (2024), Traffic Injury Prevention, 25(sup1), S66–S77.

[xi] Tesla produced 434,358 vehicles in Q4 2025 (exhibit991), or roughly 6,894 per working day. Waymo’s fleet totals approximately 3,067 vehicles (NHTSA data via TechCrunch).

[xii] ARK Investment Management LLC, Big Ideas 2026, Slides 95–100. Autonomous Vehicles section; Cybercab cost-per-mile comparison on Slide 97.

[xiii] Tesla Q4 2025 Earnings Call, January 2026.

[xiv] ARK Investment Management LLC, Big Ideas 2026, Slide 97. US ride-hail price per mile and China comparison.

[xv] ARK Investment Management LLC, Big Ideas 2026, Slides 99–100. Enterprise value breakdown across technology platforms, automakers, and fleet operators.

Important Information

ARK Investment Management LLC (“ARK”) may hold a financial interest in the companies discussed through various strategies and investment vehicles it manages. Readers are urged to use caution when considering the forecasts and other forward-looking information provided in this article, as it is inherently subjective and reflects ARK’s inherent bias toward positive expected results. There is no guarantee that actual results will align with the forecasts, and they might not be predictive. The information provided in this article is for informational purposes only and does not constitute investment advice. All statements made regarding companies or securities are strictly beliefs and points of view held by ARK and are not endorsements by ARK of any company or security or recommendations by ARK to buy, sell or hold any security. Historical results are not indications of future results.

Certain of the statements contained in this material may be statements of future expectations and other forward-looking statements that are based on ARK’s current views and assumptions and involve known and unknown risks and uncertainties that could cause actual results, performance or events to differ materially from those expressed or implied in such statements. ARK assumes no obligation to update any forward-looking information contained in this material. Certain information was obtained from sources that ARK believes to be reliable; however, ARK does not guarantee the accuracy or completeness of any information obtained from any third party.

Why This Big Idea Matters Now

The first human genome cost nearly $3 billion and took 13 years to produce. That was about 25 years ago. Today, you can sequence a whole human genome for about $100 at scale. ARK’s research suggests that drops to roughly $10 by 2030. That cost collapse is turning biology into one of the largest data-generation engines on the planet — and the scale and complexity of this data demand AI.

This is the multiomics–AI flywheel: plummeting costs generate more biological data, more data trains better AI models, and better models produce better tools, tests, and treatments — which generate even more and richer data. It is a virtuous cycle, and ARK believes it is just getting started.

In our Big Ideas 2026 research, we lay out why we think this flywheel could reshape healthcare — from how biological data is generated, to how diseases are detected, to how drugs are developed, and ultimately to whether some diseases can be cured entirely. Across each of these areas, the companies driving this convergence are reaching meaningful inflection points.

Select Examples Of Companies Building The Future

Multiomics Tools: Twist Bioscience (TWST). Twist has built a silicon-based DNA synthesis platform that can manufacture synthetic DNA with a level of throughput and precision that was not previously possible. Synthetic DNA is a key input across drug discovery, diagnostics development, and across research disciplines from medicine to defense — and Twist’s platform has made it faster, cheaper, and more accessible to design, test, and iterate on biological constructs. This matters because the multiomics–AI flywheel depends not only on the ability to read DNA at falling costs, but also on the ability to write and engineer DNA at scale. In ARK’s view, as AI-driven drug discovery and molecular diagnostics continue to expand, demand for synthetic biology tools like Twist’s should grow alongside them.

Molecular Diagnostics: Tempus AI (TEM). Tempus has built one of the world’s largest libraries of multimodal clinical and molecular data, and an AI-enabled operating system to make that data actionable at the point of care. Over 50% of U.S. oncologists are now connected to the Tempus platform — a level of clinical adoption that is generating a powerful data flywheel, which improves clinical decisions and drives further adoption. The company is broadening this flywheel across disease areas including cardiology, neurology, and infectious disease. In ARK’s view, the scale and breadth of Tempus’s data moat reinforces a durable competitive advantage and as more clinical and molecular data flows through the platform, that advantage compounds.

AI-Developed Drugs: Recursion Pharmaceuticals (RXRX). Recursion is a clinical-stage TechBio company that has built the Recursion OS, an AI-native drug discovery and development platform powered by one of the world’s largest proprietary biological and chemical datasets. In late 2025, Recursion reported early positive clinical data for a therapeutic targeting familial adenomatous polyposis (FAP), a rare condition with no approved pharmacotherapies — a meaningful proof point that AI-native platforms can translate computational biological insights into real clinical outcomes. The company has earned more than $500 million in cumulative milestone payments from partners suggesting that large pharma is increasingly willing to invest behind AI-driven discovery workflows.

Cures: CRISPR Therapeutics (CRSP). CRISPR, in partnership with Vertex Pharmaceuticals, brought the first approved CRISPR-based therapy to market: Casgevy, a single-dose and potentially curative treatment for sickle cell disease and transfusion-dependent beta thalassemia. Patient infusions increased nearly 13x in 2025 versus 2024, demonstrating an early signal that gene-editing therapies can scale commercially. The company is also advancing an in vivo gene-editing program that demonstrated deep and durable reductions in triglycerides and LDL cholesterol — signaling a potential expansion from rare disease into cardiovascular-related disease, the world’s leading killer. ARK sees this rare-to-common disease trajectory as a key theme: companies proving gene editing in rare disease are building the clinical and manufacturing foundation to eventually address much larger patient populations.

The Breakthrough Driving The Shift

Plummeting costs are generating an ocean of biological data. According to ARK’s research, an increase in the number of diagnostic test volumes is generating more data than companies like OpenAI, Gemini, Anthropic, and xAI have used to train large language models — and that volume is projected to grow 10x by 2030. At the same time, AI models have become capable enough to extract increasingly meaningful insights from this data, from helping to identify drug targets to predicting patient outcomes.

This convergence is what makes the current moment so pivotal. The cost declines have been underway for years, but the AI capability to make use of the resulting data at scale is recent. One of the most tangible impacts of this development is the potential to shift the economics of drug development. Today, it takes more than a decade, can cost billions and roughly four out of every five clinical candidates ultimately fail. AI-driven drug development could reduce time to market by roughly 40% and total costs by roughly fourfold, according to ARK’s research. And downstream, improved biological insights should enable the capabilities to develop cures for more diseases.

The companies profiled above are among those powering this flywheel — generating the data, building the models, and developing the tools, tests, and treatments that keep the cycle accelerating.

The Debate: What Critics Get Wrong

Can AI actually improve drug development outcomes? No AI-designed drug has reached approval yet — and skeptics are right to note that. But the proof points are building. AI-driven candidates are entering and advancing through clinical trials, and major biotech and large pharma companies are putting real capital behind these workflows. We are starting to see these translate to tangible clinical proof points, including Recursion’s clinical data highlighted above and Generate:Biomedicines’ advancement of an AI-engineered antibody therapeutic into Phase 3 trials for severe asthma. The full validation cycle is still underway, but the trajectory is accelerating.

Can one-time cures work — for the healthcare system and for drug developers? Gene-editing therapies can carry high upfront price tags. Casgevy is priced at $2.2 million. But these prices are generally positioned to reflect the value of replacing the lifetime cost of managing the disease, which is why payers are willing to reimburse them. By the end of 2025, roughly 90% of U.S. patients had reimbursed access to Casgevy. In some cases, cures could ultimately drive total savings back to the healthcare system. ARK’s hereditary angioedema case study in Big Ideas 2026 estimates that a one-time gene-editing treatment priced at approximately $3 million could save the U.S. healthcare system approximately $52 billion in direct medical costs across current patients’ remaining lifetimes. The economics also favor drug developers: cures frontload cash in a drug’s commercial lifecycle, allowing them to capture more revenue before facing competition and generic entry.

What Happens Next

Multiomics tools getting cheaper and more powerful: Companies like Illumina and PacBio continue to push sequencing costs lower while improving the depth and completeness of biological data their platforms can capture. If these trends persist, the result is more data at higher quality.

Molecular diagnostics expanding in scope and scale: The range of conditions detectable through molecular testing continues to grow — from multi-cancer early detection blood tests like Guardant Health’s Shield to minimal residual disease monitoring and AI-enabled companion diagnostics across disease areas. Broader adoption of these tools should improve patient outcomes through earlier detection and more precise treatment decisions.

AI-driven drug development building clinical proof points and heading toward first approvals: A growing number of AI-driven candidates are advancing through clinical trials, and the coming years could see the first AI-designed therapies reach regulatory approval. Beyond individual programs — including progress from Recursion and Generate:Biomedicines as highlighted above — the broader development to watch is whether AI-driven approaches begin to measurably improve the probability of success in clinical development, a key lever in reshaping drug development economics.

Gene editing shifting to in vivo delivery and expanding into larger disease populations: The shift from ex vivo to in vivo delivery — where treatments are administered directly into the body rather than requiring cells to be extracted, modified, and reinfused — could dramatically broaden the range of patients and diseases gene editing can reach. Intellia expects Phase 3 data for its in vivo gene-editing treatment for hereditary angioedema by mid-2026 — potentially positioning it as one of the first in vivo CRISPR therapies to reach the market next year. At the same time, gene editing is moving beyond rare disease. CRISPR Therapeutics is advancing an in vivo program targeting lipid disorders, with 2026 updates that could further build the clinical evidence base for gene editing in common conditions.

Closing Thought

The convergence of AI and biology is not a future possibility — it is underway. The cost of reading and writing DNA continues to fall, molecular diagnostics are generating growing volumes of biological data, and companies across the multiomics landscape are translating these advances into real clinical and commercial progress. ARK believes this flywheel is poised to fundamentally reshape healthcare — and it’s just getting started.

Subscribe To ARK Research

Important Information

ARK Investment Management, LLC (“ARK”) may hold a financial interest in the companies discussed through various strategies and investment vehicles it manages. Readers are urged to use caution when considering the forecasts and other forward-looking information provided in this article, as it is inherently subjective and reflects ARK’s inherent bias toward positive expected results. There is no guarantee that actual results will align with the forecasts, and they might not be predictive. The information provided in this article is for informational purposes only and does not constitute investment advice. All statements made regarding companies or securities are strictly beliefs and points of view held by ARK and are not endorsements by ARK of any company or security or recommendations by ARK to buy, sell or hold any security. Historical results are not indications of future results.

Certain of the statements contained in this material may be statements of future expectations and other forward-looking statements that are based on ARK’s current views and assumptions and involve known and unknown risks and uncertainties that could cause actual results, performance or events to differ materially from those expressed or implied in such statements. ARK assumes no obligation to update any forward-looking information contained in this material. Certain information was obtained from sources that ARK believes to be reliable; however, ARK does not guarantee the accuracy or completeness of any information obtained from any third party.