OpenRouter | He sold his 2.2 billion fortune before the NFT crash and quickly positioned himself in the hottest AI sector

Author: Di Yafan

In 2022, a Stanford graduate's net worth once exceeded $2 billion.

The OpenSea he created is the world's largest NFT trading market, valued at $13.3 billion.

Just months before the NFT bubble burst, he made a more critical decision to leave.

Two years later, his new company grew tenfold in 7 months, securing investments from a16z, Sequoia, and Menlo, with a valuation of $500 million.

His name is Alex Atallah. His new company is called OpenRouter.

This is a story about timing and methodology replication.

Who is OpenRouter? What does it do?

If you are an AI application developer, you must know the name OpenRouter, which primarily helps developers solve the pain of model switching:

• Want to use Claude to write code, but find it often lacks capacity
• Want to use GPT for analysis, but the price makes you wince
• Want to try open-source models, but find you have to rewrite a whole set of API integrations

Each model vendor's API is different. Every time you switch models, you have to change the code.

What OpenRouter does is somewhat like Ctrip, putting all airlines into one app.

One API, access to 300+ models. 60+ providers. Switching models? Change one line of code. OpenRouter as a multi-model aggregation layer

Two startups, the same methodology

Before starting his ventures, Alex Atallah already had a hardcore software background: Stanford Computer Science major, Palantir engineer, co-founder and CTO of OpenSea… OpenSea founders Alex Atallah (left) and Devin Finzer (right)

He explained the commonalities of his two startups in a podcast:

"OpenSea organized this very heterogeneous inventory and put it together in one place… You see a lot of those similarities with how AI works today."

What is his methodology?

Find a "fragmented ecosystem" and then create an "aggregation layer."

• NFT era: Metadata standards vary → OpenSea aggregates
• AI era: API standards vary → OpenRouter aggregates

Alex said something that impressed me in a podcast: If it only costs $600 to train a large AI model, there could be tens of thousands or even hundreds of thousands of models in the future. At that point, they will need their own 'market'.

At the beginning of 2023, this was an extremely contrarian judgment. The mainstream narrative at the time was that OpenAI was far ahead, and other models were just following.

But Alex was right.

Today, there are already thousands of open-source models. Claude, Gemini, Llama, Mistral, DeepSeek… new players enter the market every few weeks.

In a world of explosive model growth, a "aggregation layer" is needed. That is exactly where OpenRouter fits in.

An underestimated huge market

The success of OpenRouter comes from the visible trend in the AI market today, where "inference" will replace "training" as the main focus.

The difference between inference and training, and the future trend of this market, was clearly explained yesterday in Groq's interpretation, feel free to check it out.

COO Chris Clark's perspective can be referenced:

"We believe that inference costs will eclipse salaries as the dominant operating expense for most knowledge-based companies over the next five to 10 years."

This can actually be seen from OpenRouter's own data. OpenRouter's token consumption approaches 80 trillion

The well-known AI model "Dazhong Dianping"

As one of the earliest participants in this field, OpenRouter has a unique advantage: leaderboard.

After processing over 100 trillion tokens, they know:

• Which model is the strongest at writing code
• Which model offers the best cost-performance ratio
• Which model suddenly excels at specific tasks

This leaderboard has become an important reference in the industry, with high recognition in the developer community.

What’s even more outrageous? In April 2025, a mysterious model called "Quasar Alpha" will launch on OpenRouter.

A few days later, everyone learned: This is GPT-4.1, and OpenAI chose to exclusively debut it on OpenRouter.

Because OpenRouter has a killer asset: the largest multi-model usage dataset on the internet.

Every day, millions of developers call different models here. OpenRouter knows:

• Which model performs best on what tasks
• Which provider is the most stable
• Which time period is the cheapest

This data has created the most authoritative LLM leaderboard in the industry. According to Menlo Ventures, even Andrej Karpathy (former Tesla AI director, co-founder of OpenAI) has publicly recommended it.

Once the data flywheel starts turning, it’s hard for newcomers to catch up. Andrej Karpathy mentions OpenRouter LLM rankings on X

How does OpenRouter make money?

OpenRouter's business model is relatively simple: If you use a model and spend $100, they take $5.

They charge according to the pricing set by the model vendors. They earn a "toll fee," not a "markup."

This model aligns well with the intermediary business model in Europe and the U.S.:

1. Maintain a neutral position: If OpenRouter has its own models, would you trust its leaderboard?
2. Grow naturally with the market: The larger the AI market, the more they earn.
3. Network effects: More users → more accurate data → more valuable leaderboard → more users

Alex's original words: "We want developers to not feel vendor lock-in. We want them to feel like they have choice and they can use the best intelligence, even if they didn't before."

Financial data (as disclosed)

8 people, annualized nearly $100 million GMV.

This productivity ratio is leading among similar startups.

Big market, small space

Having discussed the highlights, it’s essential to address some issues with this model:

OpenRouter's core advantages are "data" and "community," and the flywheel has begun to turn (more users → more accurate data → more valuable leaderboard), but this model also means it is closely tied to the prosperity and decline of the developer ecosystem.

This business cannot thrive without the emergence of more and more small and medium developers, as they do not have the time to do this kind of aggregation development, nor the scale to negotiate prices with AI vendors, so they need an intermediary to accomplish this.

For large companies, this may have some value during testing, but when it comes to real-scale deployment, they will likely bypass it.

In fact, it’s not just large companies; even medium-sized projects with higher usage will want to avoid it. For example, there is currently an open-source alternative called LiteLLM, which is free and can be self-deployed.

Cost-sensitive developers may ask: "Why should I give you 5%?"

If competition intensifies, this commission rate could drop from 5% to 3% or even 2%.

At that point, whether it can maintain the recently disclosed 100 times valuation is uncertain.

Of course, it is still in its early stages and will continue to grow rapidly; its upper limit is just a question to consider in analysis.

Understand OpenRouter in one minute

Q1: What is OpenRouter?

OpenRouter is an LLM large language model API aggregation platform. Through a single API interface, developers can access over 300 models (including GPT-4, Claude, Llama, etc.) without needing to integrate each model vendor's API separately.

Q2: What is the difference between OpenRouter and LiteLLM?

Both provide LLM API aggregation, but the models are different. OpenRouter is a hosted SaaS that charges a 5% commission; LiteLLM is an open-source project that can be self-deployed at no cost. OpenRouter's unique advantage is its public model leaderboard and broader provider coverage.

Q3: Who is the founder of OpenRouter?

Alex Atallah, a Stanford University Computer Science major, was a co-founder and CTO of OpenSea (the world's largest NFT market). He left OpenSea in 2022 and founded OpenRouter in 2023. His net worth once exceeded $2 billion.

Q4: How much has OpenRouter raised?

In June 2025, OpenRouter completed a total of $40 million in funding (seed round + Series A), led by a16z and Menlo Ventures, with Sequoia participating, and is valued at approximately $500 million.

Q5: Why does OpenAI test new models on OpenRouter?

According to OpenRouter, OpenAI has anonymously tested new models on its platform to obtain unbiased feedback from developers. This indicates that the OpenRouter community has a certain influence in the industry.