DeepSeek permanently reduces V4-Pro API price to promotional level, with JD.com, NetEase, and CATL a...
The AMW Read
First external funding for DeepSeek overturns §4.4 baseline of self-funded-only status; ~$10B round + permanent price cut is cross-segment structural event.
DeepSeek permanently reduces V4-Pro API price to promotional level, with JD.com, NetEase, and CATL adopting the model; CEO Liang Wenfeng reaffirms AGI goal
DeepSeek announced that starting June 1, its V4-Pro API pricing will be permanently set at the current promotional level — roughly one-quarter of the original rate: RMB 0.025 per million tokens for cached hits, RMB 3 for cache misses, and RMB 6 per million output tokens. The move comes alongside a reported ~RMB 70 billion (~$10B) fundraising round that values the company at approximately $45 billion pre-money. Confirmed investors include China's National AI Industry Fund (~RMB 10B), Tencent, IDG, and Monolith Capital, with battery giant CATL (宁德时代), JD.com, and NetEase also in active discussions. CEO Liang Wenfeng is reportedly contributing ~RMB 20B of his own capital and has told investors that DeepSeek's primary goal is advancing toward AGI, not near-term commercialization, while reaffirming the company's open-weight commitments.
Why it matters: This event collapses two distinct market signals into one moment. The permanent API price cut — reducing inference costs to roughly 25% of previous levels — extends DeepSeek's strategy of commoditizing frontier-model access, a move that pressures competitors toward cost-compression and rewards high-volume enterprise adoption. Simultaneously, the ~$10B capital raise marks DeepSeek's first-ever external funding and its entry into the elite capital tier typically reserved for labs like OpenAI and Anthropic. The participation of CATL — a battery manufacturer pivoting toward data-center energy infrastructure — signals a deepening cross-sector bet that AI model growth will drive physical-energy demand. For the broader substrate, this updates the player map of China's foundation-model segment while sharpening an open debate: whether the fastest path to AGI runs through deliberate loss-leading on price to maximize user feedback, or through premium-margin capture that funds frontier research.
Expert take: This is a capital-compression arc in fast-forward. DeepSeek is simultaneously raising the largest fundraising round in Chinese AI history and slashing inference prices to near zero-margin levels — a combination that can only work if (1) compute costs continue to drop faster than the price cuts, and (2) the investor base treats the round as a long-duration AGI bet rather than a near-term commercial return. The CATL involvement is particularly structural: it binds a model lab to a physical-energy partner, creating a hedge against the coming data-center power crunch. Liang Wenfeng's personal contribution of ~$3B underscores an ownership structure that remains founder-controlled despite the influx of sovereign and corporate capital — rare in this capital tier.