ARK • Disrupt

On April 18, President Trump signed an Executive Order directing the U.S. Food and Drug Administration (FDA) to issue Commissioner's National Priority Vouchers (CNPVs) to psychedelic drugs with Breakthrough Therapy designation, a mechanism that can compress FDA review timelines from ten months to one or two months.¹ More importantly, the order is a clear signal that the White House, the Department of Health and Human Services, and the FDA are aligned in accelerating psychedelic drug approvals. Earlier this year, third-party media reports indicated that, despite a supportive FDA, Administration officials had removed Compass Pathways' psychedelic candidate for treatment-resistant depression (TRD) from the FDA's CNPV list just hours before it announced the recipients.² After that mixed signal, Trump’s Executive Order signaled the Federal government‘s support.

Pharmaceutical psychedelics are not the same as the recreational substances that some associate with them. Investigational psychedelics are manufactured synthetically to pharmaceutical-grade standards, ensuring consistent purity, chemical identity, and precise dosing. In FDA-regulated clinical programs, researchers administer psychedelics in clinical settings under medical supervision and trained support from therapists. Classical psychedelics—a term applicable to compounds that act on mood-regulating serotonin receptors in the brain—include psilocybin, which is the active compound in certain mushrooms, as well as LSD and 5-MeO-DMT. Research demonstrates that classical psychedelics promote neuroplasticity—the brain's capacity to form new neural connections—which could result in durable antidepressant effects.³

Historically, the path to psychedelics approval has faced significant obstacles. The Drug Enforcement Administration (DEA) has classified most psychedelic drug candidates, including psilocybin, LSD, and MDMA, as Schedule I substances under the Federal Controlled Substances Act, the category for compounds with no currently accepted medical use and high potential for abuse.⁴ Scheduling applies to the molecule itself, which means that clinical sponsors and researchers face additional regulatory requirements under the DEA that does not apply to non-scheduled investigational drugs.⁵ On top of that baseline friction, in August 2024, the FDA rejected Lykos Therapeutics' MDMA-assisted therapy—an investigational treatment pairing MDMA dosing sessions with structured psychotherapy—for post-traumatic stress disorder, citing trial-design concerns including selection bias and functional unblinding.⁶ Then, last year, the FDA reportedly under administration pressure removed Compass's psychedelic therapy from the initial CNPV list, further clouding the regulatory picture. The new Executive Order is the most direct Federal response to the accumulated overhang.

In parallel, a growing body of clinical evidence suggests that psychedelics could change the treatment paradigm for mental health disorders in meaningful ways. Conventional psychiatric care for depression often relies on a daily dose of selective serotonin reuptake inhibitors (SSRIs) and other antidepressants, for months or years. The investigation of classical psychedelics, on the other hand, involves a single or small number of supervised dosing sessions, as late-stage clinical data has demonstrated that antidepressant effects can persist up to 26 weeks, if not longer.⁷ The combination of minimal dosing and durable therapeutic effect is particularly relevant for treatment-resistant depression, defined as failure to respond to prior lines of standard therapy and affecting roughly one-third of patients with major depressive disorder.⁸

Compass Pathways is leading the psychedelic pack with COMP360, an orally administered synthetic formulation of psilocybin.⁹ Both of its Phase 3 trials in TRD have delivered positive results, and on April 24, the FDA granted Compass a rolling review of its approval application and awarded COMP360 a Commissioner's National Priority Voucher, potentially positioning COMP360 to be the first classical psychedelic to achieve FDA approval.¹⁰ AtaiBeckley is advancing BPL-003—an intra-nasal formulation derived from 5-MeO-DMT—in treatment-resistant depression, with positive Phase 2b results supporting a Phase 3 initiation in the second quarter of 2026.¹¹ With FDA Breakthrough Therapy designation, BPL-003 has the potential to achieve fast-track review under the new Executive Order.¹²

Importantly, commercial viability for clinic-delivered mental health therapies is no longer speculative and could be ready to scale. Johnson & Johnson's Spravato is a ketamine-derived nasal spray that the FDA approved for TRD in 2019.¹³ While it works through a different mechanism than classical psychedelics, Spravato not only created the playbook for clinic-administered, supervised-dosing that COMP360 and BPL-003 will follow, but also has proven its commercial potential with an annualized run rate of ~$1.9 billion.¹⁴ COMP360 and BPL-003 are differentiated from Spravato on efficacy and dosing burden: Spravato requires twice-weekly, in-clinic dosing during the first four weeks and continued maintenance thereafter, while both Compass and AtaiBeckley’s treatments might involve only one or two dosing sessions.¹⁵

Trump’s Executive Order fits within a broad pattern of recent Federal government signals. FDA Commissioner Marty Makary publicly supports expedited reviews of psychedelic therapies¹⁶; HHS Secretary Robert F. Kennedy Jr. maintains that psychedelics warrant serious regulatory evaluation¹⁷; and the FDA hired Michael Davis as Deputy Director of its Center for Drug Evaluation and Research in June 2025, positioning the former Chief Medical Officer of a nonprofit company developing psychedelic therapies—and former lead of the FDA's psychiatry drug review division—to help spearhead the movement.¹⁸

In our view, the combination of late-stage clinical data, the commercial template established by Spravato, and federal signaling has converged to remove the main barriers to approval of the first classical psychedelic for a mental health indication. FDA Commissioner Marty Makary has suggested that the first approval could come as early as this summer.¹⁹ Compass Pathways is positioned to achieve the first approval, we believe, with AtaiBeckley advancing in its wake with a differentiated profile. To be sure, execution will be critical but, after years of stops and starts, the clinical and regulatory trajectories now are pointing in the same direction.

Last week, Elon Musk modulated his expectations for the rollout of Tesla’s Cybercab robotaxi, previously slated for coverage of 25–50% of major cities in the US by year-end to a more conservative “dozen or so states.”²⁰ That said, having launched its unsupervised robotaxi officially in Dallas and Houston in April, paid robotaxi miles nearly doubled sequentially.²¹

Musk also confirmed that Hardware 3 vehicles will not support unsupervised full Self-Driving (FSD)²²—a long-awaited admission, though in our view less of a blow to the robotaxi thesis than many assume. Produced between early 2019 and early 2023, while its ~4 million HW3 vehicles constitute~43% of the 9.2 million fleet,²³ Tesla probably will offer a discounted trade-in program and hardware upgrade path for their owners.²⁴ Likely hundreds of thousands of HW3 owners purchased FSD outright, limiting the scope of the obligation.²⁵ Compared to other robotaxi players, Tesla’s production capabilities should secure its strong competitive advantage on a go-forward basis.

Counterbalancing its relatively conservative comments on the robotaxi rollout, Musk assured investors that Tesla’s dedicated robotaxi vehicle without a steering wheel and pedals has entered production²⁶ that will not be subject to a federal cap on production limits. While Musk noted that the Cybercab production ramp could be slow, the Model 3 took three quarters to scale to ~3,000 vehicles²⁷—roughly the size of Waymo’s entire fleet today.²⁸ Moreover, 3,000 vehicles today would account for less than a day’s production of Tesla's current lineup.

ARK estimates that ~24 million robotaxis—less than 10% of today’s registered US vehicle fleet—could accommodate the majority of all US urban miles traveled, and that Tesla already has the production capacity to build to that scale.²⁹ ARK also projects that autonomous technology platforms stand to capture most of the robotaxi ecosystem economics and could generate $34 trillion in enterprise value in 2030.³⁰

During the past two weeks, OpenAI has released a cluster of updates that reassert its dominant position at the frontier of AI. The company introduced GPT-5.5, workspace agents in ChatGPT, ChatGPT Images 2.0, major Codex upgrades, and multiple new go-to-market initiatives. Each one impressive on its own, collectively the updates are pointing to a compelling broader go-to-market strategy: OpenAI is focused on packaging models into highly capable agents and getting those agents into the hands of more users.

Codex is the clearest expression of that strategy. What began as a coding assistant can connect into more enterprise systems, operate apps on a computer, generate images, learn from previous actions, and conduct ongoing work more autonomously. The improvements appear to be working, with OpenAI disclosing that active users of Codex grew from 3 million to 4 million in roughly two weeks,³¹ and ~20x since the beginning of 2026, as shown below.

Note: Jan 1 (0.20M) and Feb 5 (0.81M) are back-solved from OpenAI's own disclosures — Feb 5 from the Feb 12 +60% WoW statement, Jan 1 from the Feb 26 "+400% YTD" statement. Source: ARK Investment Management LLC, 2026, based on data from OpenAI 2026a, 2026b, 2026c, and Tibo 2026.³² For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

Powered by Codex, workspace agents enable companies to create shared agents that run in the cloud and can be delegated tasks via ChatGPT or Slack.

On the model front, the Images 2.0 model is a leap forward in image generation, with better precision, realism, multi-language support, and the ability to reason and remain coherent while creating complex graphics. Importantly, strong image model performance historically has correlated highly with user growth.³³ Meanwhile, GPT 5.5 sets a new watermark for Large Language Model (LLM) performance across publicly available models and is within the performance range of Anthropic’s unreleased Mythos model. Cutting tokens used by one-third relative to Anthropic’s Opus 4.7 to complete the benchmarking, GPT 5.5 scores 60% on Artificial Analysis’s Intelligence Index, topping all other frontier models, as shown below.

Source: OpenAI. 2026. For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

Reflecting both the market’s willingness to pay up for frontier intelligence and the ongoing compute shortage, OpenAI’s new model comes with a blended per-token price of $11.30 per million tokens—double that of its prior generation—according to data from Artificial Analysis.³⁴ In comparison, the blended price for Anthropic’s Opus model is $10 per million tokens.

In our view, its launch cadence, Codex traction, and ability to price GPT-5.5 at a premium suggest that OpenAI is converting frontier-model leadership into a broader agent platform with growing distribution and pricing power, a formidable position.