DeepSeek raises $7.4 bn, plans staff boom and sees global firms switch to its low‑cost V4 model

Chinese AI firm DeepSeek announced a $7.4 billion external funding round that valued the Hangzhou‑based company at roughly $50‑59 billion. Investors included Tencent, CATL and founder Liang Wenfeng, who contributed about $3 billion himself. The capital will fund a staff expansion that could double the current headcount of 150‑170 employees and support talent‑retention equity grants.

DeepSeek’s V4 model is being adopted for its low inference cost. Startup Lindy migrated all AI‑agent traffic from Anthropic’s Claude to DeepSeek V4, claiming annual savings of several million dollars despite a migration effort that was 100 times larger than expected. Microsoft is also testing V4 as a cheaper alternative for its Copilot suite, seeking to curb the high token costs of Anthropic and OpenAI models.

Access to DeepSeek’s API outside China is hindered by registration and payment requirements; third‑party gateways now provide OpenAI‑compatible endpoints at prices far below those of GPT‑4o (e.g., $0.50 per million input tokens versus $2.50). DeepSeek’s technical approach includes iterative dataset refinement, bias removal, and diverse data curation, contributing to its competitive performance at a fraction of typical AI model costs.