Baidu (BIDU): “Embedded Optionality”
One of the clearest examples of embedded optionality I’ve ever seen.
I’ve spent a lot of time lately thinking about embedded optionality—those parts of a business that are either misunderstood, ignored, or not modeled at all. The kind of things that don’t show up in a screen, but end up mattering a lot more than the core operating metrics if they play out.
Baidu is one of the clearest examples I’ve seen in years.
On the surface, Baidu still looks like a legacy search engine—profitable, entrenched, but not particularly exciting. Judging by the current $30 billion market cap, most investors seem content to value it that way.
But if you actually break it down, Baidu holds around $19.5 billion in cash and short-term investments, and carries roughly $12.5 billion in debt, leaving it with just over $7 billion in net cash. On top of that, the company owns $20.5 billion worth of publicly listed equity stakes, including iQIYI and Trip.com.
That means more than $27 billion of Baidu’s $30 billion market cap is covered by cash/investments alone—which implies the market is valuing everything else—the core business, AI investments, and Apollo—for about $2 to $3 billion.
That would already be wild if the company were shrinking or bleeding cash. But Baidu is generating close to $5 billion in EBITDA annually. It’s profitable. It’s reinvesting in AI infrastructure. And it’s actively buying back stock (though not as much as I’d like at these valuations).
And buried inside is Apollo—Baidu’s autonomous driving platform—which could be worth more than the company’s entire market cap.
Yet the market’s assigning Apollo zero value. And by extension, it’s doing the same for the rest of Baidu’s core business too.
Apollo Go
Let’s start with what Apollo actually is.
Apollo Go is China’s largest robotaxi platform—and probably the most scaled autonomous driving network in the world. As of mid-2025, it’s operating in over 15 cities, including Beijing, Shenzhen, Chongqing, and Wuhan. In total, Baidu’s robotaxis have now completed more than 11 million rides, including 1.4 million in Q1 alone—a 75% increase year-over-year.
But the scale in Wuhan is what really stands out.
Today, Baidu has over 400 fully driverless robotaxis on the road in Wuhan. No safety drivers. No test loops. These are commercial rides, booked by customers, running 24/7. And Baidu expects that number to surpass 1,000 vehicles in Wuhan by the end of this year.
It’s a real, operating network and according to the company, it’s already reached breakeven unit economics in Wuhan, which puts it well ahead of the pack globally.
Part of what makes that possible is the cost structure. Apollo’s robotaxi fleet is built around the RT6, an internally developed autonomous vehicle designed specifically for commercial use. Baidu says it can manufacture each RT6 for under $30,000—a massive improvement from earlier generations and a fraction of what competitors spend.
Waymo and Cruise, for example, are estimated to spend $150,000 or more per vehicle, in part due to complex sensor arrays, LIDAR stacks, and redundant computing systems. Apollo went the other way: simplify, standardize, scale. Bring hardware costs down enough that the business model can actually work—one city at a time.
The goal is to operate a true autonomous driving business (profitably) at scale. And that’s exactly what they’re doing.
Apollo vs. The West: Tesla and Waymo
In this space, it’s natural to compare Baidu to Tesla and Waymo. They’re the most well-known names in autonomous driving, and both are pursuing big visions.