DeepSeek is hunting for a second funding round at a valuation of roughly 500 billion yuan, with plans to raise another 50 billion yuan, according to The Information. New numbers show the company pulled in 475 million yuan (about $70.7 million) in revenue between January and July 2026 — roughly 10 times its total for all of 2025 — while its net loss narrowed to 715 million yuan, down from 935 million yuan last year.
On the profitability front, DeepSeek's overall gross margin now sits at 44.6%, with the API business alone posting an 82.9% gross margin. That's the payoff from AI infrastructure optimization, which lets the lab run models on fewer chips and keep inference costs down. And with the V4 family — especially V4-Flash — winning over developers globally, revenue growth is accelerating; last month, annualized revenue reportedly hit $400 million to $500 million.
DeepSeek has also been raising API prices to push commercialization. V4-Pro peak output pricing jumped from $0.87 to $3.96, with off-peak pricing at $1.98 — still well under Kimi K3's $15 and Claude Opus 5's $25. At the same time, AI infrastructure spending has ballooned from 1.2 billion yuan in 2025 to 11 billion yuan in the first seven months of this year, mostly on server rentals, AI chips, and compute equipment.
The company completed its first external funding round of roughly 50 billion yuan this past June, and has already hired investment banks to prepare for a Shanghai IPO next year.
https://twitter.com/jukan05/status/2092510606763180513