DeepSeek seeks $1.5B at $71B valuation with IPO plans, threatening to set the first public-market price for open-source AI ahead of OpenAI and Anthropic

The $71 Billion Question: What Happens When Open-Source AI Gets a Ticker Symbol

DeepSeek is about to do something neither OpenAI nor Anthropic has done: put a ticker symbol on open-source AI.

The lab, which already handles nearly a quarter of enterprise AI token traffic, is in talks to raise $1.5 billion at a roughly $71 billion valuation, Bloomberg reports, with an IPO targeted for 2027 that could arrive as early as late 2026 1. That would make it the first major open-source AI company to trade on a public exchange. And the implications are bigger than the number.

Consider the math. A month ago, DeepSeek closed its first-ever outside funding round: $7 billion at around a $50 billion valuation 1. Now it's chasing $71 billion. That is a 42% jump in 30 days. In a year where OpenAI's valuation negotiations and Anthropic's private rounds dominate headlines, a company most Americans couldn't name a year ago is quietly building a case that open-source AI is worth nearly as much as the closed-source crown jewels.

Here's why that matters for anyone building, investing in, or using AI.

The first price tag for open source

Right now, if you want to know what "the AI industry" is worth, you're stuck with private valuations negotiated behind closed doors. OpenAI's number comes from a funding round. Anthropic's comes from another. Neither trades publicly. Neither has a ticker symbol. Neither faces daily market pressure to justify its price.

If DeepSeek IPOs, that changes overnight. For the first time, an open-source AI company would have a real-time, publicly discoverable valuation. Every analyst, every journalist, every competitor could point to a stock price and say: this is what the market thinks open-source AI is actually worth.

And it's not just any open-source company. DeepSeek runs its cloud service on chips made by Huawei 1. Not Nvidia. Not AMD. Huawei. The same Huawei that sits at the center of U.S. export controls designed to choke China's access to advanced AI silicon.

So here's the uncomfortable question: if a company running on sanctioned chips can command a $71 billion valuation, what exactly are those export controls controlling?

The usage is real

This isn't a speculative bet on future potential. DeepSeek is already handling serious volume.

In June, DeepSeek accounted for nearly 23% of all tokens processed by enterprise AI gateway Vercel 1. Anthropic took 32%. That gap is narrower than you might think when you remember that Anthropic runs on the most expensive chips money can buy. DeepSeek is backed by Tencent and Beijing's National AI Industry Investment Fund, and it's running on hardware the U.S. government tried to block 1.

The model itself backs up the usage. DeepSeek's V4 Pro, previewed in April, is a mixture-of-experts model with 1.6 trillion total parameters and a 1-million-token context window 2. The company claims its V4-Pro-Max model outperforms open-source peers on reasoning benchmarks and outstrips OpenAI's GPT-5.2 and Gemini 3.0 Pro on some tasks, though it trails frontier models on knowledge tests by roughly three to six months 2. And it does this at a fraction of the cost: the smaller V4 Flash model charges $0.14 per million input tokens, undercutting every comparable frontier model on the market 2.

This is the financial proof point for a thesis TechCrunch recently framed this way: the real AI race is no longer at the frontier 1. It's about who can deliver capable models cheaply enough to run at scale. If DeepSeek IPOs at $71 billion, the market is essentially validating that thesis with real money.

The complication

There's a wrinkle. Anthropic has accused DeepSeek, along with Moonshot AI and MiniMax, of setting up more than 24,000 fake accounts to extract data from Claude through a technique called "distillation" 3. OpenAI separately sent a memo to House lawmakers accusing DeepSeek of using distillation to mimic its products 3. Distillation is a common training method where AI labs create smaller, cheaper versions of their own models. When applied to a competitor's model, TechCrunch notes, it can be used to "essentially copy the homework of other labs" 3.

A company facing those accusations going public at $71 billion would create a test case for how public markets price that kind of risk. And the backdrop is sharpening: the U.S. government has accused China of stealing American AI labs' intellectual property "on an industrial scale" 2.

And for the engineering leaders reading this: if open models like DeepSeek's can handle nearly a quarter of enterprise token volume at a fraction of frontier-model pricing, then per-engineer token budgets stop being about cost control. They become about competitive survival. You're not optimizing to save money. You're optimizing because the team down the street, or across the Pacific, is running the same workloads for pennies on your dollar.

The bottom line

DeepSeek's potential IPO isn't just a China story or an open-source story. It's the moment where the market gets to vote, in real time, on what AI is actually worth. And the company asking that question runs on chips the U.S. tried to ban, faces distillation accusations from its biggest competitors, and is backed by the Chinese government.

The $71 billion question isn't whether open-source AI is valuable. It's whether the market will price the future of AI, or the cost of pretending export controls work.

Either answer reshapes the landscape.