Perplexity sees 50% revenue jump with strategic pivot from search engine to AI agents
Perplexity just turned some serious heads in the artificial intelligence market.
In March 2025, the startup posted a 50% jump in monthly revenue, pushing its estimated ARR past $450 million annually. The data was obtained by the Financial Times and shows a significant acceleration in the company’s financial results.
That’s no small feat, especially for a company many people still thought of as just an AI-powered search engine.
So what changed? Basically everything about the core business strategy.
Perplexity is moving away from its role as a conversational alternative to Google and going all in on AI agents — systems capable of autonomously executing tasks on behalf of the user. The company also adopted a new usage-based pricing model, which accounts for a big chunk of the accelerated growth the market has been watching.
But the road here wasn’t easy, and what lies ahead is even more challenging. Here’s what’s behind that number and what it means for the company’s future. 👇
From search engine to autonomous agent platform
For a long time, Perplexity was pitched to the public as a reimagined Google — a tool that answered questions with cited sources and a cleaner, more straightforward interface. It worked well, had a loyal user base, and was growing consistently. It was even considered the biggest challenge to Google in two decades. But growing as a search engine has a clear ceiling, and the company seems to have spotted that before hitting it.
The shift didn’t happen overnight, but the March 2025 financial results made it crystal clear that the bet on AI agents wasn’t just a marketing play. It was a strategic decision already generating real, measurable returns for the business.
AI agents work very differently from a chatbot or traditional search engine. Instead of just answering a question or listing results, they execute sequences of tasks autonomously — accessing external tools, browsing the web, filling out forms, making reservations, compiling reports, and much more. All without the user needing to step in at every stage. This level of autonomy completely changes the perceived value of the product, because the user isn’t paying for an answer anymore — they’re paying for a completed action. And completed actions have a commercial value that’s much easier to justify, both for consumers and businesses.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
This transition also positions Perplexity much more competitively within the AI ecosystem. While many players are still fighting over conversational assistant territory, the company is trying to occupy a layer above that, where AI doesn’t just inform — it acts.
The products powering Perplexity’s new chapter
Perplexity’s strategic pivot didn’t stay at the talking-points level. It came with concrete product launches that are already shaping the user experience on the platform.
In February 2025, the company launched Computer, its flagship autonomous agent tool. Computer can operate directly within the user’s digital environment, executing complex tasks based on voice and text commands. With this launch, the company also introduced a credits model, where premium subscribers receive a set number of credits and, once they run out, pay for additional usage.
Before that, last year, Perplexity had already made a bold move by launching Comet, its AI-integrated web browser — one of the first of its kind to hit the market. Comet works as a full-fledged agent inside the browser, following voice and text commands to handle tasks like shopping, summarizing social media feeds, and even sending emails on the user’s behalf.
Another notable release was Model Council, a feature that lets you run the same query across different AI models simultaneously and view the results side by side. This gives users the ability to compare responses and pick the one that best fits their needs — something especially useful in corporate environments where precision and context make all the difference.
The platform offers access to a variety of models, including OpenAI’s GPT, Anthropic’s Claude, and open-source options from Chinese companies like DeepSeek’s R1 and Moonshot’s Kimi. This multi-model approach is one of Perplexity’s most interesting advantages, because it allows each type of task to be routed to the most efficient model. A practical example: using OpenAI’s Codex or Anthropic’s Claude Code for programming tasks, GPT-5 for text generation, and Anthropic’s Opus for complex reasoning.
The new pricing model that changed the game
One of the main drivers behind Perplexity’s revenue jump was the adoption of a usage-based pricing model. Previously, the company operated on fixed monthly subscriptions ranging from $20 to $200 depending on the plan — for both individual consumers and enterprise clients.
With the arrival of Computer, Perplexity added a consumption-based pricing layer. Under this format, top-tier subscribers receive a set number of credits and, once they exceed that limit, pay for each additional use. This change might seem subtle, but it has a massive impact on the revenue growth curve, because it decouples billing from the number of subscribers and ties it directly to usage volume.
An important note here: unlike subscription revenue, usage-based revenue is a more volatile metric for estimating annual growth. It can inflate ARR during high-usage months and contract during quieter periods. This makes direct month-over-month comparisons less precise. Before this new billing system, Perplexity had grown its ARR from $16 million to $305 million over two years — an already impressive pace on its own.
From the enterprise user’s perspective, the consumption-based model makes a lot of sense. Companies prefer to pay for what they use, especially when they’re testing a new tool and don’t yet know what internal adoption volume will look like. This reduces friction in the purchasing decision and lets Perplexity get into smaller organizations or early-stage AI adopters without requiring a hefty financial commitment right out of the gate.
Interestingly, Nvidia CEO Jensen Huang, during the company’s annual conference last month, encouraged the audience to subscribe to Perplexity’s Computer tool and pay the maximum amount. Aravind Srinivas, Perplexity’s CEO, responded directly: there’s no limit — you can spend as much as you want. An exchange that pretty much sums up the philosophy behind the new monetization model. 😄
What the numbers say about the company’s growth
A 50% jump in monthly revenue in a single month is the kind of number that grabs the attention of any investor or market analyst. But the most relevant data point here isn’t just the percentage itself — it’s what it represents within the broader context of the company and the sector.
Perplexity went from a solid but predictable growth trajectory to a leap that signals a qualitative shift in the type of product it offers and the customer profile it’s attracting. The startup now has over 100 million monthly active users across its search and agent tools, including tens of thousands of enterprise clients.
With an estimated ARR surpassing $450 million, Perplexity enters a different league within the AI ecosystem. That level of annualized recurring revenue puts the company in a more comfortable position to invest in infrastructure, hire talent, and develop new products without relying solely on fundraising rounds.
That said, these numbers need to be put in perspective. Even with this impressive growth, Perplexity’s trajectory is still considerably smaller compared to other standout players in the space:
- Cursor, the AI-powered coding company, grew to $2 billion in ARR from less than $100 million in 2024.
- Anthropic reported an ARR of $19 billion at the end of February.
- OpenAI generated $20 billion in revenue last year.
These are numbers that show the colossal scale of the AI market and reinforce that, despite real progress, Perplexity still has a long road ahead to establish itself among the sector’s biggest players.
On the investor side, the company was last valued at $20 billion in September 2024, a sharp increase from $500 million at the beginning of that year. Its investors include heavyweights like Nvidia, SoftBank’s Vision Fund 2, venture capital firms New Enterprise Associates and IVP, along with Jeff Bezos, Amazon’s founder, and Yann LeCun, former head of AI at Meta.
The controversies that have marked the company’s journey
Perplexity’s path hasn’t been free of turbulence. The attempt to position itself as an AI-powered search engine generated significant controversy. The company has faced lawsuits from publications like The New York Times and Britannica, which allege copyright infringement and plagiarism, claiming the platform copies content illegally. On top of that, a recent privacy lawsuit accuses the company of sharing user data with Google and Meta without consent. Perplexity denies wrongdoing in all of these allegations.
It’s also worth remembering that last year, Perplexity positioned itself as a potential buyer of TikTok’s U.S. operations. The proposal didn’t gain traction, though, and control of the operation ended up going to a group of existing investors and individuals connected to the Trump administration.
These episodes haven’t stopped the growth, but they’ve created an extra layer of legal and reputational attention that the company will need to manage carefully as it scales its products.
The challenges that still lie ahead
Despite the impressive numbers, Perplexity still faces important questions that will determine whether this growth is sustainable or was a one-time spike driven by the novelty of the credits model.
The biggest challenge is probably the reliability of AI agents at scale. Autonomous agents still make mistakes, and when those mistakes happen on tasks with real-world consequences — like an incorrect purchase or information sent to the wrong person — the impact on user trust can be severe. The company will need to invest heavily in verification mechanisms, controls, and transparency so that agents operate at a reliability level that justifies large-scale enterprise use.
Another area to watch is the financial picture. Perplexity operates in the red. The company pays model providers like OpenAI and Anthropic to use their models, on top of bearing the inference costs to run user queries. A Perplexity executive told the Financial Times that revenue retention is strong, but didn’t provide specific numbers. The ability to route each request to the most efficient model — the so-called triage model — is cited as one of the company’s advantages for controlling costs, but the path to profitability still isn’t clear.
The regulatory question also weighs heavily. As AI agents start executing tasks with real-world impact — financial transactions, communications on behalf of users, access to external systems — discussions around liability, privacy, and compliance will inevitably intensify. Europe already has a robust regulatory framework with the AI Act, and the United States is advancing on multiple legislative fronts. Navigating that landscape without compromising development speed will be a delicate balancing act.
Finally, there’s the challenge of long-term differentiation. The large language models powering Perplexity’s agents are also available to competitors. That means the company’s competitive advantage needs to live in layers beyond the model itself — in the user experience, in the quality of integrations, in the reliability of agent workflows, and in the ability to quickly adapt to the needs of different market segments. These are harder assets to build, but they’re also much harder to copy, and that’s exactly where Perplexity will need to solidify its position in the next chapters of this story. 🚀
