跳到主要內容

DeepSeek融資評測:估值兩個月翻倍衝上750億美元 | DeepSeek Funding Review: Valuation Doubles to $74B in Two Months

By Kit 小克 | AI Tool Observer | 2026-08-31

🇹🇼 DeepSeek融資評測:估值兩個月翻倍衝上750億美元

DeepSeek 傳出正在進行新一輪融資,金額約 74 億美元(人民幣 500 億元),投前估值來到 740 億美元,投後估值約 810 億美元,預計 2026 年 8 月底完成交割。這距離 6 月那輪外部募資、估值僅約 500 億美元,才過了兩個月,估值幾乎翻倍——這是這家開源模型黑馬第一次真正意義上的「創投化」,也是這波 AI 新聞裡最值得拆解的一則。

從避險基金自籌到外部創投

DeepSeek 過去主要靠創辦人梁文鋒的避險基金 High-Flyer Quant 自掏腰包養算力,但基礎設施成本一路飆升,光是這輪就打算加碼約 1GW 運算容量,用來訓練更大模型、搶人才。這次名單上出現 Monolith、世芯資本(Shixiang Capital)、寧德時代(CATL)等既有股東,加上 CPE、聯想之星(Legend Capital)、合肥的地方國資基金等新面孔,等於把「DeepSeek 融資」從創辦人個人操盤,正式轉為一場多方參與的資本局。

估值背後的誠實提問:營收撐得起 740 億嗎?

目前流傳的數字是 DeepSeek 年化營收(ARR)約 5 億美元,對應 740 億估值,大約是 150 倍營收的天價倍數。要注意的是,這輪條款尚未經 DeepSeek 或 High-Flyer 官方證實,外媒說法都來自「知情人士」。換句話說,這是一則值得關注、但還不能全信的融資傳聞——AI 圈的估值故事,這幾年已經看過太多先喊價、後打折的案例。

對用戶與開發者實際上代表什麼

老實說:這輪融資不會讓你手上的 DeepSeek R1 或 V3 模型變得更聰明,開源權重也不會因為估值翻倍而改版。真正該注意的是 DeepSeek 先前已經調漲過 API 價格——算力便宜了,但晶片、電費、留才成本沒有變便宜,這才是融資背後真正的商業邏輯。如果你在生產環境用 DeepSeek API,該看的是計費頁面,不是估值新聞。

DeepSeek 計畫最快 2026 年底遞交 IPO 申請,2027 年在上海科創板掛牌。這條路線清楚說明中國把 DeepSeek 當成 AI 國家隊在扶植,但估值是否經得起公開市場檢驗,還是未知數。

好不好用,試了才知道。


🇺🇸 DeepSeek Funding Review: Valuation Doubles to $74B in Two Months

DeepSeek is reportedly closing a new funding round of roughly $7.4 billion (50 billion yuan), at a pre-money valuation of $74 billion — about $81 billion post-money — with the deal expected to close by the end of August 2026. That's barely two months after its June round, which valued the company around $50 billion. A near-doubling in two months is the headline, and it's worth unpacking honestly rather than just repeating the number.

From Self-Funded Hedge Fund to VC-Backed AI Lab

DeepSeek has largely been bankrolled by founder Liang Wenfeng's hedge fund, High-Flyer Quant, but compute costs have outgrown what one hedge fund can cover. This round targets roughly 1 gigawatt of additional compute capacity for bigger training runs and talent retention. Existing backers Monolith, Shixiang Capital, and CATL are joined by new names including CPE, Legend Capital, and state-backed investment vehicles from Hefei — turning DeepSeek funding from a founder-run side project into a proper multi-investor capital structure.

The Honest Question: Does Revenue Justify $74B?

Reported annualized revenue sits around $500 million. Against a $74B valuation, that's roughly a 150x revenue multiple — an eye-watering number even by AI-bubble standards. Neither DeepSeek nor High-Flyer has officially confirmed the terms; every figure traces back to unnamed sources. Treat this as a credible but unconfirmed report, not a done deal — the AI funding space has seen plenty of headline numbers get quietly revised down.

What This Actually Changes for Users and Developers

To be blunt: this funding round won't make your DeepSeek R1 or V3 weights any smarter, and the open-source models aren't getting a valuation-triggered upgrade. What actually matters is that DeepSeek already raised its API pricing — inference may be getting cheaper industry-wide, but chips, power, and talent are not, and that's the real economics behind the raise. If you're running DeepSeek in production, watch the pricing page, not the funding headlines.

DeepSeek is reportedly targeting an IPO filing by end of 2026, with a Shanghai STAR Market listing in 2027. That timeline signals China is grooming DeepSeek as a national AI champion — but whether a public market validates a $74B tag is a separate, unanswered question.

好不好用,試了才知道 — you won't know if it's actually good until you try it yourself.

Sources / 資料來源

延伸閱讀 / Related Articles


AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends

留言

這個網誌中的熱門文章

Google Ironwood TPU v7 推理專用晶片解析:效能追平 NVIDIA、成本低 44%,AI 晶片戰爭正式開打 | Google Ironwood TPU v7 Explained: Matching NVIDIA Performance at 44% Lower Cost — The AI Chip War Heats Up

Claude Code 實測:AI 幫你寫程式到底行不行? | Claude Code Review: Can AI Really Code for You?

Cursor vs GitHub Copilot vs Claude Code:AI 程式助手大比拼 | AI Coding Assistants Compared: Cursor vs GitHub Copilot vs Claude Code