How an AI search startup went from a small team with a simple idea to a $20 billion company, and why its hardest chapter may still be ahead.
In 2022, Google search was one of those products that seemed impossible to challenge.
It had billions of users, decades of search data, enormous infrastructure, Chrome, Android, advertising, and an ecosystem built around the simple act of typing something into a box.
Then Perplexity appeared.
Its idea was almost ridiculously simple:
Instead of giving you ten blue links, just answer the question.
Ask a question.
Get an answer.
See the sources.
Ask another question.
That small change in interface turned out to be incredibly powerful.
Within a few years, Perplexity went from a tiny startup to one of the most heavily funded AI companies in the world, attracting investors such as NVIDIA, Jeff Bezos, SoftBank and major venture firms.
Its valuation eventually reached $20 billion.
But something interesting happened along the way.
The company that wanted to disrupt search discovered that search was only the beginning.
And the further Perplexity moved, the more complicated the business became.
Copyright lawsuits.
Expensive AI inference.
Dependence on other companies' models.
Competition from Google and OpenAI.
A pivot toward AI agents.
A new browser called Comet.
And the fundamental question:
Can an AI company actually build a sustainable business on top of the internet without owning the internet?
That is the real Perplexity story.
The rise started with a very simple idea
Perplexity was founded in 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho and Andy Konwinski.
Srinivas became the company's CEO.
His background was unusual for a search startup founder. He had worked as a researcher at DeepMind and later at OpenAI before starting Perplexity.
But the company's biggest advantage wasn't simply its founders.
It was timing.
ChatGPT had just demonstrated that people were willing to interact with computers conversationally.
Google had spent decades training users to search using keywords.
Perplexity asked:
What if search worked like a conversation instead?
Instead of:
"best laptops under 1000"
You could ask:
"I'm a computer science student who does machine learning and travels frequently. Which laptop should I buy under $1,000 and why?"
The system could search the web, synthesize information and provide an answer with citations.
That was a fundamentally different experience.
And it solved a real problem.
Google was extremely good at finding information.
Perplexity was trying to compress the process of finding, reading and synthesizing information into a single interaction.
That distinction mattered.
The product loop was almost addictive
Perplexity's early product had another advantage.
It wasn't trying to completely replace Google on day one.
Instead, it became useful for things Google search was increasingly bad at.
Research.
Comparisons.
Technical questions.
Follow-up questions.
Summarizing multiple sources.
Exploring unfamiliar subjects.
Once users experienced an answer engine, traditional search could feel unnecessarily slow.
This created a powerful product loop:
Question → answer → sources → follow-up → deeper research.
And unlike a traditional search engine, the product itself became the destination.
That was important.
Because if Perplexity could own the interface through which people accessed information, it didn't necessarily need to own every piece of information underneath.
Then the money arrived
Investors noticed.
In January 2024, Perplexity announced a $73.6 million round led by IVP, with participation from investors including NVIDIA and Jeff Bezos. The company was valued at roughly $520 million.
That number looks almost comically small compared with where the company eventually ended up.
Because the next stage was explosive.
Perplexity's valuation moved from roughly $1 billion to $3 billion and then to $9 billion during 2024, with SoftBank becoming a significant investor. By December 2024, the company had raised a $500 million round at a reported $9 billion valuation.
Then the numbers became even more difficult to comprehend.
In 2025, Perplexity's valuation was reported at $18 billion and later $20 billion.
Think about that for a moment.
A company founded in 2022 had gone from an experimental AI search product to a $20 billion private company in only a few years.
This wasn't just venture capital hype.
There was real product demand behind it.
But there was also a dangerous assumption hiding underneath the excitement.
Investors were betting that AI search wasn't merely a feature.
They were betting it could become a new computing layer.
And then came the problem with the internet
This is where the Perplexity story becomes much more interesting.
Perplexity's product depends on the web.
News organizations publish articles.
Researchers publish papers.
Companies publish documentation.
Bloggers create tutorials.
Reddit users create discussions.
Websites create databases.
Perplexity searches this enormous ecosystem and turns pieces of it into answers.
From a user's perspective, that is fantastic.
From a publisher's perspective, it creates an uncomfortable question:
Why would the user visit my website if Perplexity already gave them the answer?
That is the fundamental conflict.
Traditional search generally sends traffic to websites.
AI search can potentially consume the information and return the synthesized answer directly to the user.
That changes the economics of the internet.
And publishers noticed.
Perplexity faced criticism over how it used and cited publisher content, including accusations that some generated material closely resembled existing reporting.
The conflict eventually moved into the courts.
The New York Times, Dow Jones and other publishers have pursued legal action against Perplexity, while additional publishers have also filed copyright-related cases. CNN sued Perplexity in 2026, alleging unauthorized use and distribution of its copyrighted material. Perplexity has disputed aspects of the publishers' claims.
This isn't just a legal headache.
It could become a business-model problem.
Because if AI search becomes successful enough, publishers may demand licensing fees.
And suddenly the economics change.
Perplexity isn't simply paying for servers and GPUs.
It may also need to pay for the information that makes the product useful.
The second problem: Perplexity doesn't own the intelligence
There is another uncomfortable part of Perplexity's story.
Its strength is partly that it can use different AI models.
Perplexity has historically offered access to models from companies including OpenAI, Anthropic and others.
That is a fantastic product strategy.
If one model gets better, Perplexity can potentially integrate it.
But strategically, there is a question:
Where is the moat?
Imagine that Google improves Gemini.
OpenAI improves ChatGPT search.
Anthropic improves Claude.
Meta improves its models.
And open-source models become increasingly capable.
If the underlying intelligence becomes widely available, what exactly makes Perplexity irreplaceable?
The answer has increasingly become:
The product layer.
The search infrastructure.
The user experience.
The index.
The agents.
The distribution.
The workflows.
And eventually, perhaps, the browser.
That is why Perplexity's strategy started changing.
Search wasn't enough
Perplexity gradually moved beyond answering questions.
It started building toward something much larger.
An AI that could actually do things.
This is where Comet becomes important.
Instead of simply asking:
"Find me the best flight."
An agent could potentially:
Search for flights, compare them, fill out information, navigate websites and perform actions.
That is a completely different product category.
Search answers questions.
Agents execute intentions.
And the difference could be enormous.
If search is worth billions of dollars, an agent that sits between the user and thousands of websites could potentially capture much more value.
That's why the browser matters.
The browser is the real bet
Google doesn't just own a search engine.
It owns Chrome.
It owns Android.
It owns YouTube.
It owns advertising infrastructure.
It owns enormous amounts of distribution.
Microsoft has Windows.
Apple has Safari and iOS.
OpenAI has ChatGPT.
Perplexity doesn't have an operating system.
So instead, it is trying to create another interface through which people interact with the internet.
Comet.
The idea is not simply:
"Here's another browser."
It is closer to:
"What if the browser itself had an AI agent living inside it?"
That changes the competitive landscape.
The browser becomes the interface.
The agent becomes the operator.
The web becomes the environment.
And Perplexity becomes the intelligence sitting between the user and that environment.
That's a much bigger ambition than search.
But bigger ambition means bigger costs
There is a misconception about AI companies.
People see revenue growth and assume profitability follows.
It doesn't necessarily work that way.
Every Perplexity query can involve expensive model inference, retrieval infrastructure and other computational costs.
And if Perplexity uses third-party frontier models, it may also have model-related costs.
So the equation isn't simply:
Users × subscription price = profit.
It is closer to:
Revenue − inference − model costs − infrastructure − employees − distribution − content/licensing costs = actual economics.
And that equation is still being figured out across the AI industry.
Perplexity has reported substantial revenue growth, but the company has also remained unprofitable as it invests heavily in models, infrastructure and its agentic products. Recent estimates and reports put its annualized revenue in the hundreds of millions of dollars, while user numbers have grown into the tens of millions and beyond.
That is impressive.
But it also shows the scale of the challenge.
A $20 billion valuation creates very different expectations from a $500 million startup.
So did Perplexity fall?
Not really.
At least not in the traditional sense.
There is no obvious collapse story here.
The company has more users, more revenue, more products and more capital than it had a few years ago.
Calling it a failed startup would be misleading.
The more interesting "fall" is different.
Perplexity fell from the simplicity of its original idea.
At the beginning, the mission looked almost elegant:
Build a better search engine.
Now it looks more like:
Build an AI search engine, an agent platform, a browser, an enterprise product, a consumer subscription business, a model-agnostic intelligence layer and a sustainable relationship with the publishers whose information powers the system.
That is a much harder company to build.
And that is where the real test begins.
Three possible futures
Nobody knows exactly where Perplexity ends up.
But there are several plausible directions.
1. Perplexity becomes an AI interface to the web
This is the most ambitious version.
Google owns traditional search.
Perplexity owns the AI interaction layer.
Users stop thinking about "searching."
They simply tell an AI what they want.
The agent figures out which websites, tools and services to use.
In this world, Comet isn't just a browser.
It becomes an operating interface for the internet.
If Perplexity can monetize transactions, subscriptions and enterprise workflows, the business could become significantly larger than an AI search subscription product.
2. Perplexity becomes a durable AI search company
The second possibility is less dramatic.
Perplexity doesn't replace Google.
It doesn't need to.
It becomes the leading independent AI search product for people who prefer conversational research.
Google continues to dominate traditional search.
OpenAI dominates conversational AI.
Perplexity owns a meaningful niche somewhere between them.
This could still be a huge business.
Not every successful company has to destroy the incumbent.
3. The economics catch up
There is also a harder scenario.
AI search becomes increasingly commoditized.
Google improves.
OpenAI improves.
Anthropic improves.
Open-source models improve.
Browsers become increasingly agentic.
The underlying models become cheaper and more interchangeable.
At the same time, publishers demand licensing revenue.
Inference remains expensive.
Customer acquisition becomes harder.
And the market begins questioning whether a search company with large AI costs deserves a $20 billion valuation.
In that world, Perplexity could still survive.
But the company might have to become much more disciplined about where it spends money and where it can create defensible value.
The biggest question isn't Google
It is easy to frame the Perplexity story as:
Perplexity vs Google.
I don't think that is the most interesting way to look at it.
The real competition is much broader.
Perplexity is competing for the layer between humans and computers.
Google wants to own that layer.
OpenAI wants to own that layer.
Microsoft wants to own that layer.
Anthropic is moving toward that layer.
Apple is building toward that layer.
Every major AI company increasingly wants the same thing:
Become the interface through which people get things done.
That is why AI search is becoming AI agents.
And that is why browsers suddenly matter again.
The irony of Perplexity
There is something almost poetic about Perplexity's journey.
The company started by trying to make search simpler.
Instead of giving people hundreds of pages, give them one answer.
But the success of that idea created a much more complicated problem.
If you become the answer layer of the internet:
Who owns the information?
Who gets paid?
Who controls distribution?
Which model powers the answer?
Who pays for the computation?
Who owns the user?
And ultimately:
Who owns the relationship between humans and the web?
Those are much bigger questions than search.
Perplexity's next chapter
I don't think Perplexity's future will be decided by whether it "beats Google."
That's the wrong scoreboard.
The more important question is whether Perplexity can build something that remains valuable even when AI models become cheaper, better and more interchangeable.
Its opportunity is enormous.
It already has a recognizable consumer product.
It has attracted some of the world's largest technology investors.
It has built an AI-native search experience.
It is moving into agents and browsers.
It has partnerships that can expand distribution.
And its CEO has managed to turn a relatively small startup into one of the most talked-about AI companies in the world.
But the next stage is fundamentally different.
The company has already proven that people want AI-powered search.
Now it has to prove something harder:
that AI search can become a durable business.
Perplexity's biggest challenge may not be killing Google.
It may be surviving long enough to become something Google cannot simply copy.
Because features can be copied.
Models can be replaced.
Interfaces can be redesigned.
But if Perplexity can own the workflow between a human, an AI agent and the internet, it could create something much more difficult to displace.
The rise was spectacular.
The "fall" is not a collapse.
It is the moment when the easy part ended.
Now comes the hard part.
And perhaps that is the most interesting chapter of the Perplexity story.
Top comments (0)