ARK • Disrupt

Last week, Shopify reported another strong quarter, with Gross Merchandise Volume (GMV) increasing 35% year-over-year and surpassing $100 billion, while revenue increased 34%.¹ Although investors focused on softer-than-expected near-term guidance, specifically revenue growth at a high twenties percentage year-over-year, the more important point is that Shopify is positioning itself at the center of agentic commerce.

AI is reshaping both product discovery and entrepreneurship, and Shopify has been benefiting from both. Product discovery is shifting from traditional search engines toward AI native interfaces like ChatGPT, Gemini, and Claude that are enabling agentic commerce: AI purchasing agents transacting on behalf of consumers. Shopify’s platform is enjoying early traction, AI driven traffic to its stores surging 8x year-over-year and orders originating from AI-powered search up ~13x.²

As commerce shifts to AI interfaces, Shopify’s role is becoming more important in the process. AI agents require accurate product catalogs, real time visibility into inventory selection and levels, payment processing, and fulfillment infrastructure to facilitate shipping and delivery. Powering millions of merchants with billions of products, Shopify’s infrastructure layer is connecting merchants to AI agents.

To standardize its ecosystem, Shopify partnered with Google to co-develop the Universal Commerce Protocol (UCP) enabling AI agents to scale across commerce workflows, from discovery through checkout across platforms and payment providers. Amazon, Meta, Microsoft, Salesforce, and Stripe have joined the initiative—which matters, according to Shopify, because catalog-powered searches convert at ~2x the rate of generic AI searches that rely on more outdated, web-scraped data.

During its earnings call, Shopify’s management noted that growth in orders from new customers in AI channels is nearly double that from traditional channels. US e-commerce penetration is less than 20% of retail sales, suggesting that agentic commerce could turbocharge the shift from offline to online purchasing.

AI is lowering the barriers to entrepreneurship, another trend that has benefited Shopify. Sidekick AI, Shopify’s AI assistant for merchants, is gaining traction now that the number of weekly active shops using Sidekick has increased ~4x year over year. During the last quarter alone, merchants have created more than 12,000 custom apps. Indeed, Sidekick is building nearly half of all Shopify Flow automations today.³

Meanwhile, Shopify Pulse is evolving Sidekick from an assistant to an autonomous operator that is identifying opportunities using merchant and market data and recommending and/or executing actions. In our view, AI-native Sidekick will help entrepreneurs to co-found companies over time.

Together, these initiatives suggest that Shopify is evolving from an e-commerce storefront provider to an—if not the—operating system and infrastructure layer for agentic commerce, powering the next generation of entrepreneurs.

A class of drugs originally developed to treat Type 2 diabetes now is impacting household spending and the financial results of industries far from healthcare, including fast food chains and apparel retailers. GLP-1 receptor agonists, medicines that mimic a gut hormone that regulates blood sugar and reduces appetite, have grown from a niche diabetes treatment into one of the fastest-growing drug classes in history. We believe that trajectory is likely to continue.

In the last few months, pharma giants Novo Nordisk and Eli Lilly have launched oral formulations of their GLP-1 drugs, removing the friction associated with injections that limited early adoption.⁴ Both companies have reported that ~80% of patients starting the pills are new to the drug class, suggesting that oral pills are expanding the user base instead of cannibalizing injectable users.⁵ While both Lilly and Novo Nordisk developed them for a narrow clinical purpose—managing blood sugar in patients with diabetes—the underlying mechanism of the drugs impacts multiple physiological systems, and their ripple effects are far-reaching.

GLP-1 stands for glucagon-like peptide-1, a gut hormone that signals the brain to reduce appetite and the pancreas to regulate blood sugar. While the U.S. Food and Drug Administration (FDA) initially approved the drugs to lower blood sugar in patients with Type 2 diabetes, their therapeutic reach has widened. In June 2021, for example, the FDA approved semaglutide, marketed as Wegovy by Novo Nordisk, for both chronic weight management in adults with obesity and overweight adults with at least one weight-related condition like high blood pressure or high cholesterol.⁶

Then, in March 2024, the FDA approved Wegovy for an additional indication—reducing the risk of major adverse cardiovascular events, including cardiovascular death, heart attack, and stroke in adults with established cardiovascular disease who also are overweight or obese.⁷ The FDA based its approval on the SELECT trial, a large cardiovascular outcomes study of ~17,600 such adults without diabetes. In that trial, Wegovy reduced the combined risk of major adverse cardiovascular events by 20% compared to the placebo group.⁸

In our view, Wegovy’s approval highlights a broader pattern showing that longevity-relevant therapies can reach the market through disease-specific regulatory pathways that, over time, generate broader clinical value as drug mechanisms impact shared biological pathways like metabolism and vascular risk. Their broader biological reach is central to GLP-1s’ success. The GLP-1 hormone itself is part of a coordinated regulatory system that the body uses to manage interconnected processes—appetite, satiety, glucose handling, and reward signaling, among others.⁹ Since GLP-1 receptor agonists impact this system, they can impact diverse processes simultaneously. Beyond regulating blood sugar and body weight, GLP-1 receptor agonists have impacted cardiovascular risk, kidney disease, sleep apnea, and the brain's reward circuitry—the neural system involved in cravings for food, alcohol, and nicotine.¹⁰ A single drug can impact multiple disease processes if the biology is interconnected.

Those impacts are beginning to surface in consumer spending data. One Cornell University study found that, within six months, GLP-1 users cut spending on groceries by ~5% on average, on savory snacks by ~10%, and on fast food chains by ~8%. Most of the users who discontinued the medication, however, largely returned to pre-adoption spending patterns. Survey data from EY-Parthenon found that 44% of GLP-1 users reported drinking less alcohol after starting treatment, with 82% of those impacted reporting less alcoholic consumption even after stopping the drug.¹² In apparel, the effects have surfaced in company earnings. In its 2025 results, Destination XL, a US plus-size men's retailer, reported that annual sales declined roughly 7% year-over-year, with management citing GLP-1 usage among ~25% of its customer base as a contributing factor.¹³

Fast food is another category that GLP-1 adoption could reshape, given its association with chronic conditions that GLP-1s are designed to address. The industry has faced significant pressure recently. Last week, while Novo Nordisk and Eli Lilly reported standout GLP-1 revenues, shares of restaurant chain Shake Shack plummeted ~30% after its revenues missed consensus estimates, while Papa John's stock fell ~8% after the company reported a ~6% drop in North American comparable sales on a year-over-year basis.¹⁴

Fully untangling the impact of GLP-1 compared to food inflation and affordability is difficult. Clearly, though, the addressable population for these drugs is set to grow substantially. As Eli Lilly highlighted at the time of its oral GLP-1 approval last month, fewer than one in ten people who could benefit from a GLP-1 have prescriptions, suggesting that the impact of GLP-1s on consumer-facing sales is likely to intensify as adoption broadens.¹⁵

That trajectory raises a longer-term structural question. Industries built around chronic disease—and the consumer behaviors that drive it—will have to pivot as effective treatments scale. While impacting outcomes well beyond approved indications is uncommon, it is not unprecedented. Approved in 1960 as a contraceptive, for example, the birth control pill influenced women's participation in the labor force, their education, and their consumer behavior patterns over the following decades.¹⁶

In ARK's view, GLP-1s are not an anomaly, but an early example of a pattern likely to recur with increasing frequency. AI and multiomics are reinforcing each other in a flywheel that is deepening our understanding of the biological pathways underlying chronic disease. That flywheel is making drug discovery and development more scalable and precise. As more therapies emerge from the flywheel, more are likely to engage interconnected biological systems effectively, and, like GLP-1s, produce effects that exceed the indications for which they were developed.

Medicine is shifting from managing the symptoms of chronic disease to treating its underlying biology. For industries built around chronic disease—and for the consumer behaviors that have sustained it—the consequences should compound.

Last week, Chief Technology Officer (CTO) Praveen Neppalli Naga disclosed that Uber wants to outfit its drivers’ vehicles with sensor kits to collect real-world driving data for autonomous vehicle companies.¹⁷ Paired with Uber’s 25 autonomous vehicle (AV) partnerships and its new “AV cloud” data platform,¹⁸ the announcement marks a decisive return to autonomy after Uber sold its in-house self-driving unit to Aurora during the pandemic in late 2020.¹⁹

In our view, Uber’s recalibration makes sense. According to ARK’s research, autonomous electric ride-hail is poised to dominate urban transportation.²⁰

Why didn’t Uber pursue this approach sooner? Notably, it has not committed to a timeline or a specific sensor kit, but its plan to democratize the data is telling. Uber would fare best in an autonomy ecosystem with many small winners, while a single autonomy player or a duopoly that scales could threaten its core business and compress its take rate.

Real-world data is essential for every autonomy player. Tesla’s full self-driving (FSD) fleet recently surpassed 10 billion cumulative miles,²¹ dwarfing Waymo’s ~200 million fully autonomous miles.²² Yet, with the largest data-collection operation in autonomous driving, even Tesla has a test fleet dedicated to finding corner cases that will inform its robotaxis. Whether Uber can translate its driver network into a meaningful data advantage remains to be seen.

Last week, Anthropic and SpaceX forged a partnership to enhance Anthropic’s Graphics Processing Unit (GPU) capacity and create a new revenue stream for SpaceX ahead of its Initial Public Offering (IPO).²³ Anthropic will lease more than 300 megawatts of capacity from SpaceXAI‘s Colossus 1, a data center with ~220,000 GPUs. That capacity will enable Anthropic to increase rate limits across its products, a much-needed response to massive demand that has forced the company to curtail usage in recent months. This deal shows that SpaceX can monetize the ground-based infrastructure that it acquired with xAI while it plans to monetize future space-based compute capacity as an infrastructure-as-a-service provider. Based on GPU rental prices, back-of-the-envelope math suggests that the deal could generate $5+ billion per year in revenue for SpaceX,²⁴ with some observers arguing that the rental fee could be higher.²⁵ As part of the agreement, Anthropic also expressed interest in multiple gigawatts of SpaceX’s orbital compute capacity in the future.

While the deal should support near-term profitability and lower the skepticism about its xAI investment ahead of its IPO, SpaceX could limit its capacity by leasing scarce compute capacity to a direct competitor. If its next-generation models²⁶ training in the Colossus 2 data center see strong demand, xAI will have ~300 megawatts less capacity available to serve those models than otherwise would have been the case.

Block is back, delivering strong first-quarter results as CashApp’s gross profit growth accelerated to 38% year-over-year, and Square’s gross profit growth accelerated to 9%.²⁷ Block also raised its full-year guidance, signaling broad-based strength across its ecosystems.

Those results suggest that Block’s internal restructuring is working. The company has reshaped its organization dramatically during the past few years, moving from siloed business units to functional roles and reducing staff by ~40%. Management flattened its organizational structure, reduced layers of management, and streamlined decision making. As a result, its product velocity appears to have accelerated meaningfully.

AI is central to the acceleration.

Goose is Block’s automation layer for internal operations. Employees work with Goose through BuilderBot, Block’s internal AI development interface enabling anyone inside the company to use Slack to build or modify product features. Since January 2026, the number of code changes per engineer has increased more than 2.5x,²⁸ highlighting that AI-assisted development is improving productivity.

Block also embeds AI directly into its external ecosystems through MoneyBot for CashApp users, and ManagerBot for Square sellers. MoneyBot is a financial assistant that helps users manage spending, discover offers, and take actions within CashApp, while ManagerBot works like an AI co-founder, helping sellers analyze performance, identify opportunities, and manage their businesses more efficiently. Ahead of its broader launch slated for June 2026, both have increased engagement meaningfully, MoneyBot across the Cash App ecosystem and ManagerBot across more than 1 million sellers.

Both products are evolving from reactive chat interfaces to proactive intelligence layers. Instead of providing basic answers to questions, they increasingly surface recommendations and actions before users ask. Block’s management noted that ~70% of conversations are initiated through proactive prompts. Moreover, sellers completing an insight session with ManagerBot are 33% more likely to return to Block, while Cash App users are 5x more likely to revisit MoneyBot after taking a recommended action.²⁹ Those dynamics suggest that Block’s AI agents are increasing engagement and strengthening retention across both of its ecosystems.

AI agents also are setting Block up for Neighborhoods, in our view Block’s most ambitious strategy ever. Neighborhoods connects Cash App’s 59 million monthly active users (MAUs) with millions of Square sellers in their local communities, turning Block into a localized commerce network. Sellers can reach nearby customers, gain followers, promote products, and offer loyalty rewards directly within Block’s ecosystem, while Cash App users can discover businesses, redeem rewards, and transact seamlessly inside the app.

According to management, Neighborhoods adoption has accelerated nonlinearly, with ~100,000 CashApp users already following local sellers, putting it on track for gross payment volume of $320 million at an annual rate.³⁰ Clearly, Block is leveraging AI to connect its two-sided seller and consumer ecosystems into a single commerce network built around local relationships, discovery, and engagement.

Last week, Genesis AI unveiled GENE-26.5, its first foundation model for robotic manipulation, alongside a series of striking demos that attracted widespread attention to the robots’ human-like dexterity.³¹

Source: Genesis 2026.³² For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

Backed by Khosla Ventures, the Paris- and California-based company is building a full-stack robotics platform aimed at general-purpose physical intelligence. Central to the announcement is news of a human-scale, five-fingered dexterous robotic hand, which it developed in partnership with China's Wuji Tech and engineered to match human proportions and degrees of freedom.

To train its model, Genesis built a proprietary 1:1 data glove like the safety gloves common in industrial settings,³³ enabling Genesis to capture rich motion, force, and tactile data directly from humans performing everyday specialized tasks. GENE-26.5 is trained not only on the glove data but also on internet-scale video and high-fidelity simulation.

The result is a foundation model capable of fluid, coordinated behavior across complex, long-horizon tasks. Viral demos showcased one-handed egg cracking, multi-step cooking, precise pipetting for lab environments, two-handed smoothie preparation, piano performance, and mid-air Rubik's Cube solving—all without teleoperation.

In our view, this full-stack approach is a notable milestone in the physical AI landscape that has the potential to accelerate reliable manipulation capabilities. Gensis AI CEO Zhou Xian noted that Genesis also is developing a full-body humanoid robot. As highlighted in ARK’s Big Ideas 2026,³⁴ robotics software is likely to advance rapidly, supported by continued improvements in GPU compute density and software efficiency on existing hardware. We look forward to seeing and sharing the next stages of this technology adventure.