DeepSeek is a China AI Labs company founded in 2023 and based in Hangzhou, China. It has raised $22.4B in total funding, most recently a Funding round in 2026.
| Date | Stage | Amount | Valuation | Lead investors |
|---|---|---|---|---|
| Oct 22, 2026 | Funding round | $15.0B | — | Tencent, CATL |
| Jun 16, 2026 | Funding round | $7.4B | — | — |
Model files suggest DeepSeek trained a model on Nvidia Blackwell chips, despite US export restrictions on those chips to China.

Analysis suggests DeepSeek and Moonshot AI model starting weights carry NVIDIA Blackwell traces, implying 3,000 to 4,500 GPU training clusters.

A report on DeepSeek puts Nvidia's chip exports under scrutiny, with no confirmed policy action in the headline.
Investors linked to Alice Huang reportedly took stakes in DeepSeek, per The Information.
Alice Huang-linked firms join DeepSeek's $15.8B funding round, reportedly to gain access to Nvidia Blackwell chips.

Tencent Hunyuan Hy4 scores 35 on the Artificial Analysis v5 Index, ahead of DeepSeek V4.1 Flash 1017.
DeepSeek-V4 is the model series DeepSeek released in 2026 under an MIT licence, in Flash and Pro variants with a one-million-token context window. It succeeds the V3 base model and the R1 reasoning model as the company's main open-weight release.
DeepSeek's flagship open-weight mixture-of-experts base model, released in December 2024, notable for strong benchmark performance at a low reported training cost.
An open-weight reasoning model released in January 2025 that DeepSeek positioned as competitive with leading reasoning systems, distributed with open weights.
Earlier open-weight model families released in late 2023 covering code generation and general language tasks, which established the company's open-weight release approach.
We don't have a live feed for this company's ATS. Their careers page has every open role.
View all careers ↗