Every kind of AI buyer now runs more of its AI on NVIDIA. That includes the largest cloud companies, the leading AI labs, governments, and the makers of self-driving cars and robots. Q1 sales reached $82 billion, up 85% in a year. That is the third straight quarter of accelerating (faster and faster) growth.

• Q1 revenue reached $82 billion, up 85% from a year earlier. Data-center sales alone rose 92% to $75 billion.
• Amazon’s cloud (AWS) will add more than 1 million NVIDIA Blackwell and Rubin chips, starting this year. These are the current and next AI processors.
• Vera is NVIDIA’s first processor built for AI agents (software that does tasks on its own). It opens a brand-new $200 billion market the company has never sold into. Every major hyperscaler (giant cloud company) and system maker has signed up to deploy it.
• Renting a prior-generation H100 chip now costs 20% more than at the start of the year. The older A100 rents for nearly 15% more. Older chips rent for more, not less.

This quarter, NVIDIA began to split its sales by buyer type. The split shows the five or six giant cloud companies now account for only about half of its data-center revenue: $38 billion of $75 billion. NVIDIA no longer leans on a handful of clouds. The other half comes from AI start-ups, enterprises, factories, and governments across nearly 40 countries. That half grows faster: up 31% from the prior quarter, against the clouds’ 12%. And the chips NVIDIA shipped years ago keep renting for more, not less. H100 rental prices are up 20% this year alone. That is pricing power that holds long after each chip ships.

Demand for AI chips has outrun what NVIDIA alone can supply. So every cloud company, server maker, and AI lab now orders from AMD as the credible second source. Q1 brought in Meta, which committed to deploy up to six gigawatts of AMD AI chips.

• Q1 revenue grew 38% year-over-year to $10.3 billion; data-center revenue hit a record $5.8 billion, up 57%
• Meta committed to deploy up to six gigawatts of AMD Instinct AI chips across several chip generations, including a custom accelerator co-designed on AMD’s MI450 architecture
• AMD began sampling its MI450 AI chips to lead customers. It stayed on track to ship Helios, its pre-built AI server rack, in the second half of 2026
• AMD raised its 2030 forecast for the server-processor market (the standard CPUs every server runs on) past $120 billion. That is more than double the $60 billion it projected last November. It also guided server CPU revenue up more than 70% year-over-year in Q2

Underneath the AI-chip headlines, AMD’s server-processor business is the quiet compounding story. It just posted its fourth straight record quarter, up more than 50% year-over-year. Cloud and enterprise sales each grew more than 50%. EPYC-powered cloud options rose nearly 50%, to more than 1,600. This means every AI buildout now budgets for AMD twice: once for the AI accelerator (the chip that does the AI work), and again for the standard server CPUs that direct it. And as AI agents multiply, they create more of the ordinary computing work that CPUs handle. AMD says the number of its CPUs per AI chip is climbing from as low as one-in-eight toward one-to-one. So every gigawatt of new AI capacity becomes fresh EPYC demand that outlives any single chip deal.

Every designer of the most advanced chips on earth (NVIDIA, AMD, Apple, Broadcom, Qualcomm) sends its designs to TSMC. No other factory can match TSMC’s manufacturing process. In Q1, demand for its three-year-old N3 process ran far ahead of supply. So TSMC broke its own rule and began building three new factories to make it.

• Q1 revenue reached $35.9 billion, up 6.4% from the prior quarter. Gross margin (the profit left after manufacturing costs) rose to 66.2%
• High-performance computing (the data-center chips that run AI) grew 20% from the prior quarter, to 61% of sales
• TSMC raised its full-year 2026 revenue forecast to growth above 30% in dollar terms. AI demand keeps outrunning supply into 2027
• TSMC is breaking its own rule against expanding a mature process. It is building three new 3-nanometer factories, in Taiwan, Arizona, and Japan, to meet AI demand. The first is due in early 2027

Underneath the headline, TSMC is spending to chase demand it cannot fill. Capital spending moves to the high end of its $52–56 billion range this year. Spending over the past three years totaled $101 billion. The next three years will run significantly higher. This means TSMC sells factory time that customers reserve years in advance, to buyers with nowhere else to go. Even a three-year-old process stays sold out through 2027. A new factory takes two to three years to build, with no shortcuts. So every dollar of that record spending locks in capacity only TSMC can supply. That is pricing power that compounds, because there is no second source at the leading edge.

The largest AI labs and cloud companies are signing longer contracts to rent CoreWeave’s AI computing capacity. Q1 brought record bookings: more than $40 billion of new commitments in a single quarter. That lifts signed future revenue toward $100 billion.

• Q1 revenue reached $2.1 billion, up 112% from a year earlier and 32% from the prior quarter
• CoreWeave signed more than $40 billion of new customer commitments in Q1, its largest bookings quarter ever. That lifted contracted future revenue (signed deals not yet delivered) to $99.4 billion, nearly 4x a year earlier
• Meta signed a $21 billion agreement, announced in early April. Anthropic came on as a new customer to build and run its Claude AI models
• CoreWeave passed 1 gigawatt of active, running power. It now holds more than 3.5 gigawatts under contract, up more than 400 megawatts in the quarter
• Rental prices rose across every chip generation, from the older A100 and H100 to the newest Blackwell. CoreWeave remains sold out of its 2026 capacity

Underneath the record bookings, CoreWeave’s customer base is widening fast. The AI-native labs and foundation-model makers that once dominated its order book are now less than 30% of the backlog. Ten separate customers have each committed at least $1 billion. This means CoreWeave’s revenue no longer leans on a handful of AI labs. New enterprise buyers now form whole new billion-dollar verticals, each on its own contract and schedule. They range from trading firms like Jane Street, which added $6 billion of capacity in Q1, to robotics and drug-research companies. And the same customers keep buying more of the platform: more than 90% now use at least two CoreWeave products, and 75% use three or more. Rental prices rose across every chip generation this quarter. That pricing power holds long after each contract is signed.

AI labs train their models on Oracle. Large enterprises replace SAP and Workday with Oracle Fusion. Oracle’s existing database customers now run Oracle, plus its new AI database and AI data platform, inside whichever cloud they already use. Q1 drove contracted future revenue to $638 billion, up 363% in a single year.

• Contracted future revenue reached $638 billion, up 363% year over year
• Cloud infrastructure revenue grew 93% year over year. The quarter added $67 billion of new AI infrastructure contracts. Most of that is prepaid or runs on hardware the customer buys
• Oracle delivered more than 1.2 gigawatts of data-center capacity to customers over the year. The next single quarter alone approaches 1 gigawatt, nearly the prior four quarters combined
• Multicloud (Oracle’s database running inside AWS, Azure, and Google Cloud) revenue grew 404% year over year. Bookings grew 325%
• 1,000 AI agents now ship inside Fusion (Oracle’s enterprise software). Q1 wins included Vodafone and Claro, which is automating service for its 30 million subscribers. Oracle also won a U.S. government-wide award for Fusion HCM

Underneath the headline demand, customers increasingly buy the AI chips themselves and prepay Oracle to run them. Such contracts now total $75 billion, at the same profit margin as the rest. This means Oracle earns full margin on infrastructure it never had to finance. The customer supplies the capital. Oracle supplies the part that is genuinely hard: it designs, secures, and operates the cluster every day. When 35,000 chips came up for renewal across 59 separate customers, Oracle resold the freed capacity inside the same quarter. That held chip utilization (the share of chips in paid use) at 97.5%. The scarce asset is not the chip; it is the ability to run it. That keeps the pricing power with Oracle long after each contract is signed.

The largest institutions on earth run more of their critical operations on Palantir’s software. They span defense, manufacturing, healthcare, banking, energy, and government. In Q1, the company’s U.S. revenue growth passed 100% for the first time since it went public.

• Q1 revenue grew 85% year over year to $1.6 billion. That is the fastest growth rate in Palantir’s history as a public company.
• U.S. commercial revenue grew 133% year over year to $595 million. One account moved to a government contract; without that move, growth would have been 143%.
• Existing customers spent 50% more than a year earlier. That is net dollar retention of 150%, up 11 percentage points from Q4.
• The U.S. Department of Agriculture awarded a contract of up to $300 million. It covers securing farmland and shielding farm programs from fraud and foreign influence.
• ShipOS (Palantir’s software for the Navy’s shipbuilding base) cut one supplier’s bill-of-materials approval from 200 hours to 15 seconds.

Underneath the headline, Palantir still employs about 70 salespeople. Only 7 of them really sell. Yet the company’s biggest problem is demand it cannot meet. This means customers pull the growth in; a sales force does not push it out. An institution runs a real problem through the software. Then it signs a large contract on its own timeline. And once a customer builds on the Ontology (Palantir’s layer that lets AI act on a company’s own data), it stays and spends more. Existing customers expand what they run on Palantir year after year. At Thomas Cavanagh Construction, 97% of employees now use the software every day. So every other tool that company runs has to justify why it is still there.

Large enterprises run more work through Snowflake. They run analytics queries there, and now AI agents act on the same data. In Q1, revenue growth accelerated to 34%. That was the fastest sequential dollar gain in company history.

• Q1 product revenue grew 34% to $1.3 billion. Growth was 30% the prior quarter and 26% a year ago. The sequential dollar gain was the strongest Snowflake has ever posted.
• Cortex Code, or CoCo (Snowflake’s AI coding assistant), went live on February 5. It is already in use across more than 7,100 accounts. That is the fastest adoption of any new product in company history.
• Snowflake Intelligence lets business users query company data in plain English and act on it. Accounts using it more than doubled from the prior quarter.
• Snowflake signed a new five-year, $6 billion agreement with AWS. That is more than double the prior contract. Snowflake also passed $7 billion in lifetime sales through the AWS marketplace.
• OpenAI expanded its partnership on the platform to $200 million. Snowflake added 616 net-new customers in Q1, up 38% and a company record.

Underneath the headline, existing customers spend more. Net revenue retention (how much last year’s customers spend this year) climbed to 126%. At the same time, Snowflake signed its most net-new customers ever. It raised its full-year growth outlook from 27% to 31%. This means Snowflake’s growth now comes from two directions at once. New customers keep arriving, and existing accounts keep spending more. That spend spans analytics and a fast-growing layer of AI agent work. CoCo and Snowflake Intelligence run on the governed data customers have already loaded and configured. Every agent a customer builds inside Snowflake pulls still more work onto the core platform. So each new AI workload makes leaving harder and the next one cheaper to add.

Engineering teams, security teams, and AI coding agents all use Datadog more. Q1 pushed quarterly revenue past $1 billion for the first time, and growth accelerated to 32%.

• Q1 revenue rose 32% to $1.0 billion. That is the company’s first quarter ever above $1 billion. Annual recurring revenue (the yearly value of signed subscriptions) crossed $4 billion.
• Datadog landed two of the world’s largest AI research teams. One deal pays seven figures a year; the other pays eight figures. They hired Datadog to watch their model-training runs. A year ago, that market was not yet a market for Datadog.
• New-customer bookings set an all-time record and more than doubled from a year ago. The average size of a new deal also more than doubled.
• Among Datadog’s AI-native customers, 22 firms now pay more than $1 million a year and 5 pay more than $10 million a year.
• 6,500 customers now send AI data to Datadog. They are 20% of all customers but 80% of revenue. Calls to Datadog’s MCP server (the connector that hands AI coding agents live production data) quadrupled quarter over quarter.

Underneath the headline, the acceleration is broad. Revenue from Datadog’s non-AI customers grew in the mid-20s percent, up from 19% a year ago. At the same moment, a genuinely new market opened for the first time: monitoring AI model training. This means Datadog is not riding a single AI wave. Its oldest business is speeding up just as a brand-new one appears. Customers also keep folding scattered tools onto the platform. 20% now run eight or more Datadog products, up from 13% a year ago. Each added product deepens the switching cost. So usage compounds whether humans or AI agents generate it.

Designers, product managers, engineers, and marketers now create more of their work inside Figma. So do the frontier AI labs themselves. In Q1, growth accelerated to 46%. And for the first time, Figma put a price on the AI these users consume.

• Q1 revenue grew 46% to $333 million. That is the second straight quarter of faster growth, up from 40% in Q4 and 38% the quarter before
• Existing customers spent 39% more than a year earlier. That is net dollar retention of 139%, Figma’s highest rate in over two years
• The paid customer base grew to roughly 690,000, up 54% from about 450,000 a year earlier
• One of the world’s largest hyperscalers (a giant cloud-computing company) folded its scattered Figma use into a single agreement of over 35,000 paid seats. It is one of the largest deals in Figma’s history
• Figma began charging for AI usage on March 18. Pro teams that bought AI credit add-ons already spend over 3x more per year than teams that did not

Underneath the headline, that 46% growth arrived almost entirely before Figma charged a cent for AI. Credit billing switched on only on March 18, near the end of the quarter. Figma has bolted a second meter onto a business that was already accelerating on seats alone. Every prompt an engineer or product manager runs now rings a register that was silent a quarter ago. And the buyers keep widening past designers. At one large European industrial customer, engineers now outnumber designers on the platform. Customers using Figma’s MCP (the connector that lets AI agents read and write design files directly) added full seats about 70% faster than customers who don’t. So the tools built to automate design work pull more paid seats onto Figma, not fewer.

Tens of thousands of businesses are turning their Salesforce systems into agentic enterprises (companies run partly by AI agents). They span every industry, from banks and hospitals to the U.S. Air Force and the AI labs themselves. Their agents act on their own customer data across sales, service, and marketing. In Q1, Agentforce passed $1 billion in annual recurring revenue, barely a year after launch.

• Q1 revenue reached $11.1 billion, up 13% year over year. Agentforce annual recurring revenue passed $1 billion. Total AI and data ARR reached $3.4 billion.
• Salesforce processed 28.6 trillion tokens (the text units AI models read and write), up 152% from the prior quarter. It converted them into 3.8 billion agentic work units (individual tasks its agents completed), up 111%.
• Salesforce closed 98 deals worth more than $1 million in new annual contract value. One was a new $72 million agreement with the U.S. Air Force. Half of all Agentforce and Data 360 bookings came from existing customers expanding what they already run.
• Agentforce went live on Salesforce’s own help site 15 months ago. Since then, it has handled 4 million customer inquiries on its own, double the volume human agents handle. Agentforce sales worked 220,000 leads on its own and generated $42 million in new sales pipeline (potential deals not yet closed).
• Slack (the workplace messaging tool Salesforce owns) drove nearly half of the quarter’s million-dollar wins, up 80% year over year. Its agent activity grew nearly 350% quarter over quarter. OpenAI and Anthropic both run their businesses on it.

Underneath the headline, the same customers keep buying more. Half of Agentforce and Data 360 bookings came from existing accounts. Salesforce’s ten heaviest agent users increased their total spend 1.5x over the past year. This means Salesforce no longer just sells seats for people to log into. It sells agents that do the work. The more work those agents do, the more the customer pays. That usage shows up as tokens and completed tasks before it shows up as revenue. The agents act on the customer’s own sales, service, and marketing data. That data already sits inside Salesforce, and Data 360 now unifies it. So the cost and risk of moving that work anywhere else rises with every agent a customer switches on.

More small and mid-sized businesses run their whole go-to-market (marketing, sales, and service) on HubSpot. Now they turn its Breeze AI agents loose to do the work. Q1 was the quarter those agents became a real, paid growth lever. Credit consumption rose 67% in a single quarter.

• Total customer count reached nearly 300,000 worldwide, up 16% year over year. HubSpot added 10,800 net new customers in Q1
• Total credits consumed (the metered fuel customers buy to run HubSpot’s AI agents) grew 67% quarter over quarter. Customer Agent (the AI support rep) led with 53% of all credits used
• Nearly 14,000 customers have switched on Prospecting Agent (which finds and reaches out to sales leads), up 33% in the quarter. Jotform, an online form builder used by over 35 million people, now buys 625,000 credits a month to run it fully automated
• Customer Agent’s resolution rate (the share of support tickets it closes with no human involved) climbed to 70%, up from 20% a year ago. Some customers exceed 90%
• Deals over $120,000 in annual recurring revenue grew 64% year over year. Customers with 500 or more seats grew over 450% year over year. Bigger companies are consolidating onto HubSpot

Underneath the headline, HubSpot rewired how it charges for AI. In April, Customer Agent moved to billing per resolved support ticket. Prospecting Agent moved to billing per qualified lead. Each comes with a 28-day free trial. Customers pay only when the agent delivers a result. This means the AI agents earn on top of the subscriptions, priced by the work they do rather than by the seat. One customer pushed a single agent from 100,000 to 300,000 credits a month because it could measure the outcomes. That spend climbs as trust in the agent grows. And every agent, HubSpot’s own or a third party’s, runs on the same shared ‘growth context’ (the accumulated knowledge of how nearly 300,000 businesses find customers and close deals). So the more agents run on HubSpot, the more valuable that context becomes. And the harder the platform is to leave.

Merchants of every size run more commerce through Shopify, from new entrepreneurs to General Motors and L’Oréal. AI shopping assistants like ChatGPT and Gemini do the same. In Q1, merchant sales passed $100 billion for a second straight quarter. Revenue grew at the fastest rate in over four years.

• Q1 revenue grew 34% to $3.2 billion, on $101 billion of merchant sales. That is the second straight quarter above $100 billion, and Shopify’s fastest revenue growth in over four years
• New signings include the luxury house LVMH, Mulberry, and the iconic Lands’ End. Orvis, the outdoor brand founded in 1856, is moving its whole business to Shopify. It gets a full unified commerce setup: online store, physical stores, and payments in one system
• Orders from AI shopping searches grew nearly 13x year-over-year. AI-driven traffic to Shopify stores grew 8x. Buyers who arrive from those searches are new to the merchant at nearly twice the rate of ordinary search
• Shopify’s catalog (the clean product data AI assistants read to find and price items) now holds over 1 billion products. Amazon, Meta, Microsoft, Salesforce, and Stripe all joined UCP, the open commerce standard Shopify co-developed with Google
• Shopify Payments processed $67 billion of merchant sales in Q1, up 41% and now 67% of the total. Shop Pay (Shopify’s one-tap checkout) processed $35 billion, up 59%

Underneath the headline, the numbers that matter most are the ones that pull merchants deeper in. Weekly active shops using Sidekick (Shopify’s in-store AI that now builds apps and automations for merchants) grew 385% in a year. Almost 90% of Q1 revenue came from merchants on the platform for more than a year. This means Shopify’s growth is not new sign-ups replacing merchants who leave. Each year’s cohort of merchants stacks on top of the last and keeps growing. The revenue base compounds instead of resetting. Shopify also built an open commerce standard with Google, and Amazon, Meta, Microsoft, Salesforce, and Stripe have now joined it. The very companies that could have routed around Shopify are instead building on the rails Shopify controls. And every sale, on a storefront or inside a chat window, still earns Shopify a monthly fee plus a slice of the payment.

Pinterest’s 631 million users now ask the platform more commercial questions. Advertisers trust Pinterest’s AI to do more of their targeting. Q1 pushed both further. Revenue growth returned to 18% and passed $1 billion, as spending outside the tariff-hit retail giants accelerated.

• Q1 revenue grew 18% to $1.0 billion, back above $1 billion for a third straight quarter. Growth excluding its largest retailers accelerated from Q4.
• Monthly users hit a record 631 million, up 11%, and all are signed in. That is a 10th straight quarter of double-digit user growth. Gen Z is now over half the base and its fastest-growing cohort (age group).
• Pinterest handles more than 80 billion searches a month. About half show commercial intent (the user wants to buy something). By ChatGPT’s own data, only 2% of its prompts are commercial.
• Performance+ is Pinterest’s AI ad-automation suite. It needs half the inputs to launch a campaign. It now carries roughly 30% of lower-funnel ad revenue (ads meant to produce a purchase now). Advertisers who adopt it grew that spend nearly twice as fast as those who don’t.
• Mejuri, a fine-jewelry brand, ran a four-week head-to-head test. Its Performance+ campaign delivered a 46% lift in return on ad spend and 62% more conversions (completed customer actions). Mejuri then adopted it more broadly.

Underneath the 18% headline, the growth came less from Pinterest’s largest retailers. It came more from a widening base of mid-market, SMB (small-business), and international advertisers. Pinterest now sends advertisers five times the clicks it did three years ago. Its ad revenue does not yet reflect that traffic. This means Pinterest’s revenue is diversifying off the handful of big retailers that softened last year. It is moving onto dozens of smaller, faster-growing buyers, each on its own schedule. The gap between the shopping Pinterest drives and the credit it collects is stored-up revenue. Pinterest’s AI bidding now plugs straight into advertisers’ own measurement systems; one lifetime-value pilot cited a 15% to 20% gain. So Pinterest starts getting paid for demand it was already creating.

Mobile-game studios let AppLovin’s AI handle more of their ad targeting. Now e-commerce brands do the same. They sign themselves up through a self-service portal. Q1 cleared the way to open that portal to the world in June, after 14 years closed.

• Q1 revenue rose 59% to $1.8 billion. The adjusted EBITDA margin (profit before certain costs, as a share of revenue) hit 85%, a new high. Revenue also grew 11% over Q4. Ad businesses almost never post a first-quarter gain on the holiday quarter
• The consumer vertical serves e-commerce and other non-gaming advertisers, and the product is only 1.5 years old. It set a record spend month in April, above any peak month in Q4. March ran about 25% ahead of January
• In June, AppLovin opens Axon (its AI ad engine) to the public after 14 years as a closed platform. Advertisers worldwide can sign up and run campaigns without talking to a salesperson
• Each new advertiser is expected to spend well over $70,000 a year. Signing 100,000 in the first year would bring roughly $7 billion in ad spend
• AppLovin’s generative-AI creative tools are now live for all customers. One is an interactive-page generator that auto-builds the tappable middle of an ad. AI video generation is in testing and days from a broad rollout

Look under the headline. The metric the company watches most closely is growth from advertisers it already has. That is what accelerated. A model upgrade a couple of weeks before the call drove it and made April a record month. This means AppLovin’s growth comes mostly from existing advertisers spending more as its AI improves, not from a sales push. Each model release lifts returns. Advertisers respond by adding budget. That is a virtuous cycle. AppLovin also wins new advertisers the same way its customers buy ads: it runs performance ads for its own platform, and those ads break even inside 30 days. So opening to the public in June pours buyers into a machine whose cost structure barely moves. Two markets now feed one auction: gaming and a faster-growing consumer vertical. Every new advertiser adds data that sharpens the models for both. And once a customer clears 30 days of spend, it almost never churns (stops spending).

From Bayer and IKEA to Nestle and Cheerios, the world’s largest brands trust Trade Desk’s AI to handle more of their ad-buying. They lock themselves in with multi-year contracts. Q1 brought the biggest month for those signings in the company’s history, with 45 new deals in March alone.

• Q1 revenue reached $689 million, up 12% year over year. The adjusted EBITDA margin was 30%.
• March was the biggest month on record for Joint Business Plans (multi-year advertiser commitments), with 45 signed in March alone. The total JBP count grew 55% year over year. New deal spend excluding renewals grew 40%.
• A top global pharmaceutical advertiser had shifted budget to Amazon. It returned in Q1 and signed a 2026 JBP that will lift its spend on Trade Desk 114% year over year.
• Audience Unlimited (Trade Desk’s new automated audience-building product) delivered strong results for one travel brand against a control group: 30% lower ad prices, 38% lower data costs, 75% more efficient cost-per-action, and a 2.7x jump in conversions.
• Retailers in Trade Desk’s data marketplace now represent more than 80% of sales from top U.S. retailers. Amazon covers under 15% of U.S. retail spend. New sponsored-listing tie-ins came from Dollar General, plus a deal to power Lyft’s ad business.

Underneath the headline, Q1 revenue growth slowed to 12% even as multi-year contract signings hit a record. The softness sits in tariff-hit consumer-packaged-goods and auto brands, not in the platform itself. This means the slowdown is cyclical: a macro squeeze on two advertiser categories. The structural asset compounds underneath it. Brands keep signing longer commitments. They keep their own data on a platform that never asks them to surrender it. Trade Desk owns no ad inventory. So its AI can weigh 20 million ad opportunities every second and pick the 300 or 400 worth buying, without steering a brand toward media it profits from. That is why a top pharmaceutical advertiser abandoned Amazon’s cheaper rates and returned in Q1 at 114% higher spend.

Duolingo’s language learners run more AI-tutored conversations every day. Q1 shows it plainly: words spoken per user in the AI video-call tutor more than doubled over the past year.

• Daily active users grew 21% year-over-year in Q1. Full-year guidance calls for roughly 20% daily-user growth.
• Duolingo published 20,500 course units in Q1 alone. That is more than 10 times its per-quarter pace two years ago. It is about as much as it shipped in all of last year.
• Courses now run all the way to professional proficiency across all 9 of Duolingo’s most-learned languages. That is B2 on the CEFR scale (the level needed to work in a language).
• Words spoken per user in the AI video-call tutor more than doubled over the past year. Duolingo now tests that feature with new subscribers on the cheaper Super tier. Those subscribers confirm they will pay more for it.
• Q1 adjusted operating profit (EBITDA) reached $83 million, about 29% of revenue. Duolingo entered Q2 with over $1 billion in cash and no debt. It expects over $350 million in free cash flow this year.

Underneath the guidance, Duolingo is making a deliberate trade. Only about 12% of its monthly users pay today. At Spotify, the figure is close to 50%. So Duolingo trades near-term price for a bigger base of engaged learners. This means Duolingo chooses reach over squeezing today’s users. Free trials now run one month instead of seven days. The video-call tutor sits on the cheaper Super tier. Both moves put the AI tutor in front of far more people without slowing daily-user growth. Those learners already speak twice as many words to the tutor as a year ago, and they say they will pay more for it. Each new habit becomes pricing power the company can collect later, once the paying base is larger.

Consumers now run more of their weekly errands through DoorDash: food, groceries, retail, pharmacy. In Q1, monthly active users hit an all-time high. DoorDash drove most of the industry’s growth.

• Monthly active users (people who order at least once a month) reached an all-time high in Q1. DoorDash drove the majority of the entire delivery industry’s user growth
• Subscription (DashPass and its overseas equivalents) had a record quarter across all three brands (DoorDash, Deliveroo, Wolt). Member sign-ups and retention both accelerated year over year
• DoorDash is now the volume share leader in US grocery delivery. It wins about 1 in every 2 customers who order grocery delivery for the first time
• Deliveroo, acquired last year, is growing at its fastest rate in four years. It reaccelerated in each month DoorDash has owned it. The European unit remains on track to add roughly $200 million of EBITDA in 2026
• Roughly two-thirds of DoorDash’s software code is now written by AI. AI agents already cut the cost of onboarding new merchants and building their catalogs

Underneath the headline, the growth is not one lever. Several levers turn at once: more users, higher order frequency, record subscription, and share gains. All of it happens across three brands and two continents in the same quarter. This means DoorDash’s growth no longer rides on any single market or category. When winter storms clip US order value by about 1%, record membership and a reaccelerating Europe keep the platform compounding. The year’s biggest project folds DoorDash, Wolt, and Deliveroo onto one global tech stack. A feature built once ships to all three. An improvement learned in London or Los Angeles spreads across the whole network. No one rebuilds it three separate times.

Restaurants of every size run more of their operation on Toast. That now includes chains like Applebee’s and Firehouse Subs. In Q1, Toast’s combined cut of every dollar spent through the platform crossed 1% for the first time.

• Toast added 7,000 net locations in Q1. It ended the quarter with 171,000 live locations, up 22% from a year ago.
• Recurring gross profit (software and payment processing combined) grew 27%. GAAP operating margin crossed 20% for the first time, reaching 21%, or $110 million.
• Toast’s total take of every dollar spent through the platform, software plus payments, crossed 1% for the first time. It reached 103 basis points on $51 billion of payment volume, up 22% year over year.
• Toast IQ (Toast’s AI assistant for operators) now has 40,000 weekly active locations. Its first AI agent, a marketing agent inside the new Toast IQ Grow, lifted pilot restaurants’ sales an average of 8%.
• Q1 marquee wins: The Alinea Group (Alinea, Next, The Aviary), 500-unit pizza chain Hungry Howie’s, and Papa Murphy’s. Toast also launched a Drive-Thru product that opens 140,000 new locations.

Underneath the location count, Toast is turning 14 years of restaurant data into work it can charge for. The Toast IQ Grow marketing agent now does the outsourced marketing job itself. AI also runs more of Toast’s own operation: 40% of support interactions are resolved by AI, coding velocity is up over 60%, and SaaS gross margin crossed 80% for the first time, up nearly 3 points. This means Toast no longer only sells software. It is starting to sell the labor restaurants used to outsource: marketing now, and soon bookkeeping, payroll, and inventory. It uses data only Toast holds, which raises what each location pays. And Toast’s take of every dollar now exceeds 1 cent. So 7,000 new locations a quarter and a rising take rate compound the same revenue line, while AI pulls cost out from underneath.

Dutch Bros’ 15 million loyalty members pull into the drive-thru more often. They now order breakfast alongside their daily drink. Q1 same-shop sales jumped 8.3%, the seventh straight quarter of transaction growth. That result was strong enough to raise the full-year outlook.

• Q1 revenue grew 31% to $464 million. Adjusted EBITDA rose 26% to $79 million. Average sales per shop hit a record $2.2 million.
• System same-shop sales grew 8.3%, with transactions up 5.1%. That is a seventh straight quarter of more visits, not just higher prices.
• Dutch Rewards, the loyalty program, reached an all-time-high 74% of transactions. Order-ahead climbed to roughly 15% of the mix.
• Food is now in 485 shops after a ninth item was added in Q1. Attach rates (the share of drink orders that add food) run in the low teens. The company-operated rollout is set to finish by the end of Q3.
• 41 shops opened in Q1, including 7 Clutch Coffee Bar conversions now running about 3x their pre-conversion volume. The 2026 plan rose to at least 185 shops. Full-year revenue guidance rose to about $2.1 billion.

Only about 1.5 points of the 8.3% same-shop sales gain came from price. The rest was more visits. Capital spent per new shop fell to $1.3 million, from $1.7 million a year earlier. This means Dutch Bros grows on traffic, not menu price. The gains hold even when customers tighten their budgets. The shops are largely company-operated and cost less to build. So the chain funds its own expansion from cash flow and keeps the full economics of each lane. Conversions like Clutch reopened at about $1.4 million each and run at roughly three times their old volume. They turn a competitor’s real estate into Dutch Bros routine.

16 million U.S. households now reach for Vital Farms in the supermarket cold case. Q1 was a reset. The price gap to rival premium eggs widened, so first-time trial stalled and margins compressed. But loyal buyers held, and the outdoor-access egg category kept taking share.

• Q1 net revenue grew 15.4% to $187.2 million. Volume drove all of it: a $34.7 million volume gain against a $9.7 million price/mix decline.
• The category kept shifting Vital Farms’ way. Outdoor-access eggs (from hens given real outdoor access, the segment Vital Farms leads) rose from 8% of U.S. egg volume in 2023 to 15% in 2026. They are up 32% year-to-date, while ordinary eggs grew just 4%.
• Existing buyers stayed loyal: units per retained household ran 2% above the prior eight-quarter average. But first-time trial fell from over 55% of buyers to 50%. Price gaps to rival premium eggs widened past what the brand could hold.
• Distribution keeps widening. Vital Farms secured at least a 50% shelf-presence increase with a top-3 retailer, plus the category-captain role for eggs at another top-3 banner. It expects its best distribution gain since its 2020 IPO: adding 20 to 30 TDPs (total distribution points, a measure of how widely a product is stocked).
• The company cut 2026 guidance: net sales to $775–$800 million and adjusted EBITDA to $0–$10 million. It absorbs about $32 million to clear an oversupply of eggs. And it is exiting its butter business to concentrate on the premium-egg core.

Underneath the headline reset, the oversupply is a timing mismatch, not a demand failure. Vital Farms buys its farmers’ eggs no matter the sales environment. It had built capacity ahead of demand. That is how a supply-constrained brand stays ready to grow. Then a mild bird-flu season pushed commodity egg prices to multiyear lows. Competing premium and private-label eggs cut their prices in step. The gap to Vital Farms widened past anything the brand had seen. New shoppers stopped crossing that gap while loyal buyers stayed. So committed supply briefly ran ahead of stalled trial, and the excess went to the breaker channel, where eggs are cracked for liquid products and brand earns nothing, at $0.10 a dozen. This means the glut is Vital Farms’ own eggs meeting a temporary air pocket in new-customer growth. The pasture-raised category is not fading: outdoor-access eggs went from 8% of the category in 2023 to 15% today, even with commodity eggs at their cheapest in years. And the fix sits with the company. At one top-10 customer, trimming the premium from about 35% to about 25% (still above rivals) lifted volume 18% in two weeks. Slowing the new Indiana plant and paying farmers to pause production pull near-term supply back into line. Adding real capacity still takes years of hens, pasture, and farm build-out. So once trial reaccelerates, the binding constraint returns to supply.

Celsius now sells three energy-drink brands: Celsius, Alani Nu, and Rockstar. Together they hold ~20% of the U.S. energy-drink market. Q1 pushed portfolio share to 20.9% on record revenue of $783 million.

• Q1 net revenue reached a record $783 million
• Alani Nu, acquired April 2025, delivered $368 million in Q1 sales, up ~60% year over year. Tracked scanner data shows it up roughly 100%
• Portfolio dollar share reached 20.9% in the four weeks ending April 12. One of every five energy drinks sold in the U.S. is now a Celsius-portfolio product
• Celsius completed the Alani Nu integration. It captured the ~$50 million in cost synergies it had promised a year earlier
• Adjusted operating profit (adjusted EBITDA) reached $195 million, up ~$125 million year over year. Margin rose to 24.9% from 21.2%

Underneath the headline, growth is shifting between the brands. The original Celsius brand grew about 6% in Q1. A deliberate SKU cleanup (trimming the product list) held that figure down. So did shipping less to stores while they sold down extra inventory. Demand at the register did not weaken. Meanwhile, Alani Nu grew roughly 60%. The company now runs two separate billion-dollar brands aimed at different consumers. This means Celsius no longer leans on a single brand or a single kind of buyer. Its revenue now comes from two large brands at different price points and occasions. A slowdown in one can be offset by the other. As the Alani Nu integration closes and its ~$50 million in cost synergies land, more of each additional dollar of sales falls to profit. Adjusted operating profit rose about $125 million year over year. Scale turns into widening margins that outlast any single quarter’s flavor launch.