DeepSeek nears $12 billion funding round ahead of planned 2027 IPO
Chinese artificial intelligence startup DeepSeek is close to raising at least $12 billion—and potentially up to $15 billion—in a Tencent- and CATL-backed funding round ahead of a targeted early 2027 public listing.
Chinese artificial intelligence startup DeepSeek is nearing at least 80 billion yuan (roughly $12 billion) in a new financing round, blowing past its initial capital-raising target of 50 billion yuan, Bloomberg reported. Sourced reports from CNBC and Cointelegraph note the company is considering expanding the round to between $14.9 billion and $15 billion as corporate and state-backed investors seek stakes following the release of its latest AI model.[7][4][1][5][2][9][8]
Tech conglomerate Tencent and battery manufacturer CATL are among the largest investors in the round, having also participated in DeepSeek's first external raise of more than $7 billion earlier in the year. DeepSeek had originally targeted a valuation of approximately 500 billion yuan, or around $75 billion, for the financing.[4][1][10]
Once the financing closes, the startup is expected to begin corporate restructuring ahead of an initial public offering targeted for early 2027, following earlier reports from Reuters of a potential mainland China listing. The company plans to use the new capital to expand its computing infrastructure, including deploying at least 160,000 of Huawei's highest-end AI accelerators at a data center in Inner Mongolia.[4][7][5]
Key facts
- DeepSeek is close to raising at least 80 billion yuan (about $12 billion), surpassing its initial target of 50 billion yuan ($7.5 billion).
- Signed term sheets and investor demand could expand the total raise to approximately 100 billion yuan, or between $14.9 billion and $15 billion.
- Tencent and CATL are among the largest backers in the funding round.
- DeepSeek initially targeted a valuation of around 500 billion yuan ($75 billion) for the round.
- DeepSeek plans to restructure for an initial public offering scheduled for early 2027.
- The company intends to deploy at least 160,000 Huawei AI accelerators at a data center in Inner Mongolia.
- DeepSeek raised more than $7 billion in its first external funding round earlier in the year.
Sources · 10 sources
- 阳明
阳明AI@x_autonomyPost on X ·
【AI热点】14:00-16:00 更多详细信息→https://t.co/Q0KLW1XS6S 1. 消息称 DeepSeek 接近完成至少 800 亿元融资，腾讯、宁德时代重金参与 [影响大·可执行低]据彭博社报道，DeepSeek 新一轮融资接近敲定，规模至少 800 亿元，超出原定约 500 亿元目标，最终总额可能逼近 1000 亿元，腾讯和宁德时代是出资规模最大的投资方之一。（补充：Bloom 2. Anthropic 5180亿美元计算力支出中约4137亿美元无论使用与否都需支付 [影响大·可执行低]Anthropic 计划未来数年在云计算和计算力上支出 5180 亿美元，其中约 4137 亿美元为不可撤销承诺，即使容量闲置也需支付，平均每年约 410 亿美元。 3. 索尼音乐要求下架 26 万首 AI 深度伪造歌曲，阿黛尔等艺人被冒充 [影响大·可执行低]索尼音乐要求各流媒体平台在 9 月底前下架超过 26 万首未经许可冒充其旗下艺人的 AI 生成曲目，数量几乎是 3 月底 13.5 万首的两倍。被冒充的艺人包括阿黛尔、布兰妮·斯皮尔斯、皇后乐队和迈克 4. Strata 开源引擎让 12GB 显存显卡运行 125B Qwen3.8-Flash-Next，RTX 5070 达 94 词元/秒 [影响中·可执行高]开发者 Niko1221 开源推出 Strata 引擎，可在 12GB 及以上显存的消费级显卡上运行量化的 125B 参数 Qwen3.8-Flash-Next 模型。 5. OpenAI 将 ChatGPT Auto-review 免费开放给所有 ChatGPT 账号用户 [影响大·可执行低]OpenAI 宣布 Auto-review 免费向所有通过 ChatGPT 账号登录的用户开放，可在 settings > permissions > auto-review 中开启。该功能由第二个智 6. OpenAI 与 Anthropic 向澳大利亚议会表态支持强制通报智能体数据泄露 [影响大·可执行低]据路透社报道，OpenAI 与 Anthropic 于周二向澳大利亚议会表态，支持立法强制企业通报其 AI 智能体造成的数据泄露事件，并坦承目前是否通报监管当局完全由企业自主决定。 7. Google 让 Google Drive 和 Google Docs 原生支持 Markdown 文件 [影响大·可执行低]Google 宣布让 Markdown 文件在 Google Drive 和 Google Docs 中获得原生支持。用户可在 Drive 中预览 .md 文件、在 Docs 中打开、编辑、评论并协作 8. JEPA-Anything：用一套配方为 7 个领域构建世界模型 [影响中·可执行高]PhAI Labs、CUHK、复旦、斯坦福、牛津与普林斯顿的研究者发布 JEPA-Anything，一个领域无关的世界模型框架，用共享学习配方覆盖视觉、生物、临床轨迹、控制、分子动力学、物理场和天气 9. 为什么 Common Lisp 如今是最好的编程语言 [影响中·可执行高]有观点认为，LLM 写代码极快，瓶颈已从写代码转为验证代码，而 Common Lisp 的 image-based 特性让函数可即时替换、无需重启，反馈循环几乎消失。 10. OpenAI Web Search 首登搜索指数排第五 [影响大·可执行低]OpenAI Web Search 首次进入 Artificial Analysis 搜索指数，得分 74，位列 26 款产品第 7、提供商第 5，落后 Perplexity、Octen、Parall 11. Kandinsky 6.0 上线 fal [影响中·可执行高]Kandinsky 6.0 现已在 fal 上线 - Pro 用于电影级、高保真生成- Lite 用于快速、低成本迭代- 原生同步音频- 内置放大，另有独立的 VSR 和 VSR Lite 用 12. Falcon-Emirati-7B：让 LLM 学会阿联酋方言、文化与语感 [影响中·可执行高]Falcon-Emirati-7B 发布，这是一个基于 Falcon-H1-Arabic 的 7B 方言专用模型，专注理解与生成阿联酋阿拉伯语。它沿用 Falcon-H1 混合架构（Mamba 与 T 13. 韩国拟推 4.7 万亿韩元专项计划，2027 年 3 月起研发前沿 AI 大模型 [影响大·可执行低]韩国科学技术信息通信部拟推出总规模 4.7 万亿韩元的专项计划，自 2027 年 3 月起研发前沿人工智能大模型。该项目将在韩国国会 12 月审议通过 2027 财年预算后通过竞争性招标遴选牵头开发主 14. Cohere 推出 North 2：可对接任意模型的企业级 AI 控制中心 [影响大·可执行低]Cohere 发布企业平台 North 2，可自主处理多步工作流并跨会话保留上下文，支持 Slack、SharePoint、Jira 等工具接入，还能从聊天提示词直接生成演示文稿、仪表盘和简单应用。该 15. 面向扩散语言模型的表示空间 MMD 后训练方法 [影响中·可执行高]研究者提出一种扩散语言模型（DLM）后训练方法，在冻结预训练 DLM 的特征空间中最小化生成分布与参考分布的 Maximum Mean Discrepancy（MMD）。 16. TextReg：通过正则化文本空间优化缓解提示词分布过拟合 [影响中·可执行高]TextReg 是一个正则化框架，通过正则化文本梯度实现软惩罚目标，缓解 LLM 提示词优化中的分布过拟合问题。 17. 谷歌 Googlebook 解锁 BL 指南上线，Magisk 确认将适配 [影响大·可执行低]谷歌已在官网发布 Googlebook 笔记本解锁 Bootloader 指南，用户需在设置中连点 build 号 7 次开启 OEM 解锁，再进入 Recovery 模式的高级选项解锁引导加载程序。 18. Sam Altman 谈 AI 安全与用户自主权 [影响小·可执行中]Sam Altman 在 POLITICO 采访中表示： 19. AI 替代工人与复印机维修技术员 [影响小·可执行中]当人们谈论 AI 轻易取代工人时，我想起这篇关于 1990 年代复印机维修技术员的民族志研究。 你会看到工作是复杂的、由工人即兴发挥的、非正式的，而且记录得很差。 （但最终技术确实在很大程度上消灭 20. 马东锡NLP感叹一丝恐惧 [影响小·可执行中]但是，怎么感觉到了一丝 fear？
Open source - CO
Cointelegraph@CointelegraphPost on X ·
🇨🇳 JUST IN: China’s DeepSeek is considering doubling its latest funding round to $14.9 billion as state-backed funds and corporate investors seek stakes. https://t.co/6JdoqqRnzf
Open source - R�
RuntimeWire 🏴☠️@runtimewirePost on X ·
DeepSeek is seeking at least 80 billion yuan after investor interest pushed its fundraise past an initial 50 billion yuan target. https://t.co/WC5GWx1UlA
Open source - WS
Wall St Engine@wallstenginePost on X ·
DEEPSEEK SET TO RAISE AT LEAST $12B IN TENCENT, CATL-LED ROUND AHEAD OF 2027 IPO DeepSeek is nearing at least $12 billion in new funding, significantly above its initial target of roughly $7.5 billion, as investor demand surged following the release of its latest AI model, according to Bloomberg. Tencent and battery giant CATL are among the largest investors in the round. Based on signed term sheets, the final raise could approach 100 billion yuan, or roughly $15 billion. DeepSeek had originally targeted a valuation of about 500 billion yuan, roughly $75 billion, for the financing. Once the round closes, the company is expected to begin restructuring in preparation for an IPO targeted for early 2027. Reuters previously reported DeepSeek was preparing for a potential mainland China listing. The new capital will help fund DeepSeek’s rapidly growing AI infrastructure needs. The company plans to deploy at least 160,000 of Huawei’s highest-end AI accelerators at a massive new data center in Inner Mongolia, while continuing to develop open AI models and pursue artificial general intelligence. DeepSeek previously raised more than $7 billion in its first external funding round earlier this year, with Tencent and CATL also participating. The latest financing would position DeepSeek for one of the most closely watched Chinese technology IPOs in years, alongside rival Moonshot AI, which is also targeting a 2027 public listing. Source: Bloomberg
Open source - AG
AGTP@AGTPinsightsPost on X ·
BREAKING: DeepSeek is close to raising at least 80 billion yuan ($12 billion), well above its 50 billion yuan target. Tencent and CATL are among the largest backers. An IPO is planned for early 2027. https://t.co/lNlj4mOlY6
Open source - TS
The Straits Times@straits_timesPost on X ·
DeepSeek set to raise at least $15 billion in Tencent-backed funding https://t.co/lPOPSKG5ZK
Open source - BL
Bloomberg@businessPost on X ·
DeepSeek is close to securing at least $12 billion in its latest round of funding, blowing past its own capital-raising target ahead of a landmark IPO in early 2027 https://t.co/2oHD36SJmX
Open source - QU
Quartz@qzPost on X ·
DeepSeek's $12B raise just rewrote the AI funding playbook DeepSeek set out to raise $7.5 billion and ended up nearly doubling that figure as investors flooded in after its latest AI model dropped. Backed by Tencent and b https://t.co/JUWbRCtGOZ #DeepSeek #TencentAI #AIFunding https://t.co/LoBYx0bMkz
Open source - CN
CNBC@CNBCPost on X ·
DeepSeek considers doubling latest funding round to up to $15 billion, sources say https://t.co/MUvul37CmC
Open source - TE
Techmeme@TechmemePost on X ·
Sources: DeepSeek is close to raising $12B+ in a round that could reach ~$14.9B, with Tencent and CATL as the biggest contributors, ahead of an early-2027 IPO (Bloomberg) (Visit Techmeme dot com for the link and full context!)
Open source