Amplitude, Inc. Class A Common Stock 2026 Q2 Earnings Call

Thanks, John, and good afternoon, everyone. Welcome to Amplitude's second quarter 2026 earnings call. Today, I'll cover three things. First, our Q2 results. Second, how we transformed Amplitude into an AI company, and why every company I talk to now wants to learn how they can do the same. Third, a look at our product and a spotlight on our customers. Let me start with the numbers. Q2 revenue was $101 million, up 21% year-over-year. Total annual recurring revenue was $410 million, up 22% year-over-year and up $36 million from last quarter. That was made up of two parts: inorganic ARR from Statsig of $17 million and organic ARR growth of $19 million. Andrew will walk through the details. Non-GAAP operating loss was $1.5 million. Customers with more than 100K in ARR grew to 824, an increase of 30% year-over-year.

Both AI natives and large enterprises are driving this growth. Let me step back and tell you about our transformation and then how we're helping customers along their AI journeys. We help companies build better products. Every company wants to transform to deliver software products in an AI native way. We've made that transformation at Amplitude over the last two years, and now our customers are looking to learn from us. Becoming an AI company starts with the organization. Two years ago, we first transformed our engineering team by bringing in AI engineers who built with it for years. We moved into adjacent functions like product management, design, and the more technical parts of go-to-market. We also brought in AI expertise through acquisition. Founders and other members of the team from these companies have taken leadership roles across Amplitude.

I have focused on bringing in leaders who are former founders and who have a technical background. Gab, our Chief Product Officer, started multiple companies, including Loom Systems, which sold to ServiceNow in 2020. In addition, Nate, our Chief Commercial Officer, has a degree in math and physics and started his career as an engineer programming in C++ and Java and building databases. Most recently, we added Angela Ferrante as SVP of Marketing. Angela founded Laudable, which went through Y Combinator Summer 2021, sold it in 2025 and is a technical marketing leader who builds apps with AI in her spare time. In addition to all of this, we're continually re-educating everyone at Amplitude through initiatives like AI Week, unlimited token spend, and a living token leaderboard. This has all resulted in three times the number of pull requests in six months.

We've reduced our pull request cycle from five hours to 44 minutes. Bug reports are down 55%. Five percent of our pull requests are submitted from designers and product managers with no engineering involvement. We've leveraged AI to shorten our closing process by a day. We built customer health dashboards that enable our sellers and leaders to track customer usage, bringing our own Amplitude data alongside Salesforce data and data from other sources. When I talk with our customers, they are all focused on how they can transform their business to be AI native like we have done at Amplitude. The AI landscape is changing rapidly, and they want to learn how to adapt. Our customers are on a spectrum of AI adoption. Our job is to meet them where they are and then educate them on how to take the next step.

We work with leading AI companies to learn what the bleeding edge in product development looks like. We use that knowledge to educate the rest of the market, including the largest enterprises deploying at scale. More than 40 AI native companies now pay us over $100,000 a year. Those customers include Harvey, Midjourney, Character.ai, and one of the leading foundational AI model companies. On the enterprise side, enterprises are now more than 68% of our ARR. This quarter included agreements with Paramount, Jaguar Land Rover, and Domino's Pizza. We've improved our pricing and packaging. We reduced down to a single meter to make it simpler for enterprises to add additional products. We increased the amount of data on our free plan, so we're the best for those just getting started. Amplitude has the best pricing, whether you're a startup or a large enterprise.

One of the biggest changes with building an AI native company we're seeing at Amplitude and with our peers in private markets is in the cost structure. A lot of inference spend is required in order to deliver AI native products, which increases the amount spent on cost of goods sold. On the other hand, you do not need to add as much operating expense to continue to grow a business at scale. We are embracing this change in cost structure as part of our transition to an AI native company. For now, we expect gross margins to stay in the low 70s. We will offset that with a commensurate reduction in operating expenses. That allows us to continue to show the same leverage in operating income as we have planned. I am continuing to drive Amplitude to a 20%-plus operating margin business over the long term.

We offer three products to meet customers wherever they are on their AI journey. Amplitude gives you the deepest understanding of how people use your product. Our agents increasingly do that discovery for you. Statsig gives you feature flagging and experimentation built on the world's most advanced stats engine with an engineering-first view. It's also integrated natively with data warehouses. Wave is the future of product development, self-improving products where we automatically recommend what to build next based on signals from users. While we're early here, I'm actually excited to show you a demo today. Together, these three products close the product development loop. Understand what's happening, measure what ships, and ship what matters. That loop is how AI-native business is built. Let me go deeper on Amplitude. Global chat is becoming the primary way our customers interact with their product data.

You ask it a question in plain language. It does the analysis, no dashboard building required. It's become the de facto way many companies do product analytics. Global Agent finds the root cause behind 75% of customer questions and hands you the answer. There are 1.3 million Global Agent interactions every week. Root cause discovery rates are improving by one percentage point every month. As of today, over 40% of all insights come from AI agents as opposed to humans. We expect this to continue to grow. Today, for our demo, I want to show you Custom Agents, Statsig, and Wave. Let's start with Custom Agents. Custom Agents are teammates that automate recurring workflows on your product data and push that work to other tools and systems. This is our chat interface. An increasing number of users are interacting with Amplitude mostly through chat and agents.

I'll ask a question. Which group of users are most likely to purchase next week? Chat can now write its own code to perform this analysis. This unlocks the ability to run deeper analysis and create powerful new graphs and artifacts, including diagrams like you see here, out of time decile lift, an ROC curve, segment propensity. You can dig in by seeing the actual code used and step-by-step analysis. This type of deep analysis has never been available before in analytics tooling. We are no longer bound by the constraints of a UI. We can also create automatic and recurring agents that run in the background. I give it these instructions. I want this analysis run every Monday morning, cross-reference with marketing activity in Confluence. DM me the results in Slack. Amplitude then creates the agent that you see here.

This is the entire prompt, including connectors to Atlassian and Slack. It will run regularly every Monday and push the results to me. We are building the best analytics agent across all data sources. Statsig is the leading product for experimentation and feature management. Statsig runs experiments natively on your cloud data warehouse, whether that is Snowflake, BigQuery, Databricks, or Redshift. Let me show you what this looks like. Here is the results page for one of hundreds of experiments that an e-commerce customer is running. This experiment is testing a larger product image versus the default size. There's a lot of statistical machinery behind a good experiment, but the UI makes it simple for an engineer to run. Up top, they can monitor exposure, which is saying is the experiment is healthy or not.

We expect to see a 50/50 split. We're doing good, and as you can see over here, we're getting a healthy check. We move to the scorecard that has the results. This has a confidence interval of 95%. Statsig uses advanced techniques like CUPED and sequential testing that allows engineers to speed up time to decision. We have those turned on. In monitoring, we see specific events we're tracking for this experiment. We're seeing positive results. The checkout event is up by 27.4%, ±2.3%. Cart conversion is up, total purchase dollars is up, while carts per session is down. For the rollout of this feature, we have a progressive rollout starting with employees, moving to early access users, then early release, and a scheduled rollout for everyone else. Statsig has a variety of advanced experimentation capabilities for rollout, like feature gating, dynamic configs, and automatic rollbacks.

Together, these are the mechanisms that a team uses to ship a change gradually, tune it while live, and pull back automatically it goes wrong. Last, I want to show you Wave, the future of product development. Wave allows for self-improving products that automatically recommend what to build next based on signals from your users. Wave is magical. Wave looks across all the different data sources you have, analytics, experimentation, Session Replay, Guides and Surveys, feedback, and many others. It then synthesizes that data into a set of product recommendations, plans those recommendations, and then helps you create those changes in your product. I'm going to walk you through a real example Wave suggested and built for Amplitude's documentation site. On our documentation site, Wave found a spike in failed searches through looking at Session Replay and analytics data.

The core problem was that search on our docs page fired on every keystroke. Typing a single letter to start a search returned an empty, no result state before the person finished typing their search, leading to a bad experience for users. Wave explains the reach of this issue. Every user who uses search, it has an expected impact of decreasing total search failures by 80%. Wave has automatically created a visual example of the problem below, so it's easy to understand. It also has a full explanation of the evidence. For the plan, Wave sketches a wire frame of the recommended update, setting a three character minimum and a 200 millisecond debounce to trigger the search. Wave can also drive execution. It automatically created the pull request and Cursor wrote the code. Mark, our technical writer, was able to merge this pull request and ship this.

No engineers, no designers, and no product manager. Wave measures the results of the change. There is a massive decrease in total search failures. Simply amazing. Let's talk about some of our customers. We had a great quarter with both new lands and expansions. We added or expanded our relationship with customers including Paramount Global, Jaguar Land Rover, Teladoc Health, Chime, Disney Ad Platforms, F5 Networks, Coursera, Grammarly, Kraken, and Crunch Fitness, among others. I want to tell you three stories about how these customers are leveraging our platform. First is Coca-Cola FEMSA, which sells to hundreds of thousands of small shops across Latin America. Every shop is different, but for years, they had to run the same broad campaign to everyone because there was no way to tailor a message to that many retailers by hand. AI changed that. They began sending each retailer its own recommendation every week written by AI.