DeepSeek built its reputation on doing more with less — training competitive models for a fraction of what Silicon Valley spends. Turns out "less" still has a limit, and that limit is apparently $7.4 billion.
A Very Large Check for a Company Famous for Small Ones
DeepSeek is closing in on a funding round worth roughly 50 billion yuan, about $7.4 billion, at a pre-money valuation near $74 billion, with the round expected to finalize by the end of August. Existing backers Monolith and Shixiang Capital are in, alongside a notable new name: CATL, the Chinese battery giant, joining the round. The money is earmarked for model research and, unsurprisingly, a serious compute buildout.
The raise is also a stepping stone toward a public listing — DeepSeek is reportedly eyeing an IPO filing as soon as late 2026, with a debut on Shanghai's STAR Market targeted for 2027. The company's annualized recurring revenue has climbed to around $500 million, but it's still burning cash fast, plowing money into training runs, chips, data centers, and talent.
Efficient AI Still Isn't Cheap AI
DeepSeek's entire origin story was proving that frontier-quality models didn't require Big Tech-sized budgets. This round quietly complicates that narrative: staying competitive at the frontier, even efficiently, now apparently costs tens of billions of dollars in compute and talent. "Efficient" bought DeepSeek a seat at the table. It didn't exempt the company from the table's stakes.
There's a second story here too — a Chinese AI lab edging toward a public listing while most Western frontier labs, OpenAI and Anthropic included, remain privately held and famously cagey about their books. If DeepSeek actually files, it'll be one of the first times outsiders get an audited look at what running a frontier model lab really costs.
Cheap AI got expensive to build. It's still, notably, not expensive to use — which is the part that actually matters if you're the one deciding which model to plug into your business.
Picking the right AI model for your business is about more than the price per token — it's about which vendor's roadmap you can actually trust a year from now, and that's exactly the kind of call we help clients make — let's figure out yours.
Source: Tech Startups