Nvidia投資Perplexity評測：300億美元背後的晶片布局 | Nvidia-Perplexity Investment Review: $30B Chip Compute Bet

8月 27, 2026

By Kit 小克 | AI Tool Observer | 2026-08-27

🇹🇼 Nvidia投資Perplexity評測：300億美元背後的晶片布局

Nvidia投資Perplexity的傳聞在2026年8月底炸開：據The Information報導，Nvidia正洽談以超過300億美元的估值投資AI搜尋新創Perplexity，同時考慮更廣泛的技術授權合作。這個數字比Perplexity去年9月那輪的200億美元估值暴漲超過五成，也讓這樁交易成為本週AI圈討論度最高的話題之一。

為什麼是現在？Perplexity營收正在暴衝

撐起這波估值的不是願景簡報，是實打實的營收數字。Perplexity的年化營收已經衝上7.5億美元，比今年初不到2.5億美元成長超過三倍。推手之一是「Perplexity Computer」——一個幫專業用戶自動處理電腦任務的AI代理產品，把Perplexity從單純的搜尋引擎往AI代理公司轉型。

Nvidia的「晶片房東」打法，不是第一次

這不是Nvidia第一次用鈔票綁住下游客戶。同樣的劇本近期在多個案子上演：

Groq：Nvidia領投，估值來到200億美元
Poolside：60億美元授權其AI模型，再加碼10億美元投資，估值120億美元，還挖走上百名員工
Enfabrica：投資估值9億美元

報導指出，Nvidia原本考慮對Perplexity採取類似Poolside的「授權加投資」結構，付錢換技術授權、順便挖角工程師，最後才轉向純股權投資。邏輯很直白：AI推理（inference）是吃算力的大戶，Nvidia身為晶片供應商，與其等別人來搶市場，不如直接入股，確保搜尋層的架構持續跟自家硬體生態綁在一起。

對開發者跟用戶意味著什麼？

老實說，現階段這對Perplexity的產品體驗不會有任何直接改變——這是一筆財務與策略布局，不是功能更新。值得留意的兩點：第一，交易目前仍未定案，雙方都沒有正式證實，AI圈這類「洽談中」的新聞常常最後條件會變或直接告吹；第二，即使成交，Perplexity依然是靠外部資金撐算力燒錢的公司，300億美元估值不代表獲利能力，只代表投資人（尤其是自己的晶片供應商）願意用真金白銀押注它能撐到下一輪。對於在用Perplexity或考慮串接其API的開發者，這筆交易更像是「Nvidia幫你的供應商上了保險」，能不能反映在產品穩定性或定價上，還要看後續。

好不好用，試了才知道。

🇺🇸 Nvidia-Perplexity Investment Review: $30B Chip Compute Bet

The Nvidia Perplexity investment talk blew up in late August 2026: The Information reports Nvidia is in discussions to back AI search startup Perplexity at a valuation exceeding $30 billion, while also exploring a broader technology licensing partnership. That number marks a jump of more than 50% over Perplexity's roughly $20 billion valuation from last September, making it one of the most-discussed AI stories this week.

Why Now? Perplexity's Revenue Is Surging

The valuation bump isn't riding on a pitch deck — it's backed by real revenue growth. Perplexity's annualized revenue has climbed above $750 million, up more than 3x from under $250 million at the start of the year. Part of that growth comes from Perplexity Computer, a cloud-based AI agent that automates computer tasks for professional users, pushing Perplexity beyond a pure search engine into agent territory.

Nvidia's “Compute Landlord” Playbook, Again

This isn't Nvidia's first time using its balance sheet to lock in downstream demand. The same pattern has played out recently across several deals:

Groq: Nvidia-led round valuing the company at $20 billion
Poolside: a $6 billion model licensing deal plus a $1 billion investment at a $12 billion valuation, along with job offers to more than 100 employees
Enfabrica: an investment at a $900 million valuation

Reports say Nvidia initially considered a Poolside-style structure for Perplexity — paying to license its technology while pulling in engineering talent — before settling on straight equity. The logic is simple: AI inference is a massive compute consumer, and as the primary chip supplier, Nvidia would rather own a stake in the search layer than watch its architecture drift away from Nvidia hardware.

What This Actually Means for Developers and Users

Honestly, none of this changes the Perplexity product you use today — it's a financial and strategic maneuver, not a feature update. Two things worth watching: first, the deal is still unconfirmed by either side, and “in talks” AI funding stories frequently change terms or fall apart entirely before closing. Second, even if it closes, Perplexity remains a company burning outside capital to pay for compute — a $30 billion valuation reflects investor conviction (especially from its own chip supplier), not profitability. If you're building on Perplexity's API or relying on its product, this deal is best read as “Nvidia insuring its own supplier” — whether that translates into better uptime or pricing for you is still an open question.

好不好用，試了才知道。

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

搜尋此網誌

AI小貼士