Answer Amjad Masad on the Cursor deal, fighting Apple, and why he’d rather not sell

Amjad Masad has been building Replit for ten years, but the last 18 months have been something else entirely. The AI ​​coding company has gone from $2.8 million in revenue by 2024 to what Masad describes as a multi-billion dollar annual run rate.

At StrictlyVC’s sold-out TechCrunch event in San Francisco on Thursday night, we covered a lot in a short amount of time, starting with the question everyone in the industry is asking right now: in a world where rival Cursor is reportedly in talks to be acquired by SpaceX for $60 billion, is Replit also bound to sell? We also delved into Replit’s net retention – a measure of how much existing customers are increasing their spending – which Masad says is as high as 300%, his willingness to take Apple to court for what he clearly calls its App Store battle with Replit, and the opportunity for the company to start investing in its customers.

On the issue of freedom, Masad was unclear. Unlike Cursor, which he said was operating at 23% gross margins, he said Replit has the economics to make that approach work — even if he stopped short of deciding whether to sell it entirely.

The following is edited for length and clarity:

TC: The SpaceX contract reported by Cursor was the talk of the industry last week. What do you do about it?

AM: It’s hard to be an independent, small AI company building on basic models, especially if you’re burning a ton of money. Cursor’s suggested portion of the report has negative 23% margins, and if you want to invest in training models, that makes it very difficult to stay independent.

For us at Replit, in part because we target a different set of customers, we’ve been able to run a business that makes sense. We have been in full margin for over a year. We are more expensive, but we offer much more. Our audience is often non-technical users who have not previously been able to create any software. We provide an end-to-end platform – from an all-in-one notification to a scalable deployed application. We handle security, databases, database migrations. And we’ve been doing this long enough to build many of those classics on the platform.

Is Replit for sale? I would imagine that you are talking to potential beneficiaries all the time; it is your responsibility.

Yes. We have amazing partners, and sometimes they bring up these topics. But we will try to stay independent. I would like us to remain a private company. We’ve been around for 10 years, before it was accepted that you can make apps out of ideas. We were talking about creating a billion software creators back in 2018 at YC, and people sometimes laughed at that dream. Now that dream has come true, and we’ve started this transition with our coding experience in September 2024. It feels like we can go a long way with it.

He works closely with Anthropic, Google, and OpenAI. If you had to rate yourself – who did the best?

Anthropic is still undefeated in the core agetic loop. They have the best tool call; the agent can stay together for a long time. GPT-5 catches fast. Google’s Flash family of models is amazing for price performance. If you want something quick and cheap, they’re actually beating open source right now. We use all three, and I honestly wouldn’t discount the new labs. Reflection AI comes with open source models that we hear great things about. And the Chinese models are impressive – Kimi has been the model of the Anthropic generation since January, so she only has three months to go.

If you’re scrambling to find a business deal, what’s in it for you?

Most of our sales are inbound or organic – very product led. We got customers like Zillow and Meta just by using people who accepted the product and raised their hands to buy the business plan. When it comes down to the bottom and there is legal baking, we usually win on the product. But even in cases where we might miss a feature, when it comes to the C-suite and IT team, Replit wins with security. Many vibe-coding tools will generate a website and connect it to an external database – good products, but they make security more difficult, because the database is open to the public and you need to configure line-level security, which is especially difficult for non-technical developers. Replication is a full stack, with a database built into the project and open to the public – that makes the application inherently secure.

We’ve also spent 10 years fighting crypto fraudsters and hackers, so our cybersecurity practice is as good as a dedicated cybersecurity startup. Every time you deploy an app to Replit, we create a completely new distributed project in Google Cloud. We inherit Google’s security model.

Can we talk about churn? How long do you hold on to customers if the best prototypes end up being rebuilt into the company’s existing stack?

Churn is very low, and overall retention is incredibly high – 300% in some cases. What we really hear from customers is that when developers panic and try to rebuild an application into their stack, they often make it worse. Once businesses are comfortable with the full Replit stack – especially when setting up a single-tenant environment – ​​they keep the applications in Replit. Bain & Company, for example, replaced Replit and Databricks with Tableau and Power BI.

There is a growing concern about AI bloat – non-technical users generating too much code and burning too many tokens. That’s good for you [given your usage-based fees]. What about your customers?

We don’t have much money sadly. Businesses are very aware of ROI, and they tell us about the returns they get. For the most part they feel the investment is absolutely worth it – usually one, two, three orders of magnitude. If they spend $100,000 a month with Replit, they usually generate $2 million, $3 million, $10 million in some kind of return.

Let’s talk about Apple. Another competitor, Lovely, just got its app-building app approved by the App Store this week. Repetition was in App Store purgatory, with Apple blocking its updates for months. How much does that hurt you?

It’s not life or death — we might lose the app and it won’t do anything meaningful for our business. But it’s an app that people genuinely love. We’ve been on the App Store for four years. Children in disadvantaged communities learn to code in Replit on their Android devices. Managers use it in meetings.

The reason Replit is blocked while others are not, we believe, is that Replit makes iOS apps. When we introduced that capability in December, there were moving charts showing how many apps were entering the App Store for us. We think Apple feels threatened by that.

Apple’s stated reason is that you are downloading new code to the device [after the approval process]which violates their guidelines.

That’s a lie. And we can prove it in court if necessary.

Will that happen?

I hope not. I’m an Apple fan, and I’d love to collaborate and build something great together. We are happy to send customers to Xcode [Apple’s own development environment]. But you can’t manage a marketplace that billions of people can access and make decisions that are biased or based on opinions.

Just wondering if, like Nvidia, OpenAI and others, you are considering investing in your customers to get equity.

We thought a lot about it, and it is considered. I personally invested in a few startups that started on Replit before I made any money. Some of them, like School of Magic – a teacher decided to take his time during the COVID to learn a little bit about vibe writing and build an AI app for other teachers. He found the problem that in America, we burn a lot of teachers. He wanted to use AI to reduce work. He did that, and made $20 million in the first year. Some of the companies that started on Replit are, I think, worth half a billion dollars. The trading going on in Replit right now is really exciting. We merged with Stripe a few months ago, and the transactions flowing through Replit are growing triple digits month over month. Soon, our customers will make more money than we do.

You can watch our full interview with Masad below: