OLIX光子AI晶片評測：3.12億美元挑戰輝達GPU霸權 | OLIX Photonic AI Chip Review: $312M Bet Against Nvidia GPUs
By Kit 小克 | AI Tool Observer | 2026-08-09
🇹🇼 OLIX光子AI晶片評測：3.12億美元挑戰輝達GPU霸權
光子AI晶片新創OLIX宣布完成3.12億美元B輪融資，公司估值一舉衝上33億美元，成為近期AI晶片產業繼輝達之外最受矚目的資金動向。這輪融資由多家新舊投資人共同加碼，反映市場開始認真押注「用光取代電來跑AI運算」這條路線，能不能真的撼動GPU的推論市場地位。
什麼是光子AI晶片（OTPU）？
光子AI晶片全稱Optical Tensor Processing Unit（OTPU），用光子而非電子訊號執行AI運算核心的矩陣乘法。理論優勢很直白：光速傳輸更快、發熱更低、耗電也更省，這正好打中目前GPU資料中心最頭痛的散熱與電費問題。OLIX 把賭注全押在這套架構上，目標是切入輝達GPU目前壟斷的AI推論市場。
OLIX這輪融資規模與背景
本次3.12億美元Series B吸引了不少重量級新股東進場，包括晶片IP大廠Arm、量化交易巨頭Hudson River Trading，原有投資人Hummingbird Ventures、Crane、Plural、Creandum也持續加碼，Netflix共同創辦人Reed Hastings則以天使投資人身分入股。更值得注意的是，OLIX背後有英國政府主權AI創投基金撐腰，顯示英國正試圖在美中晶片競賽之外，走出自己的AI硬體路線。
- 融資金額：3.12億美元 Series B
- 公司估值：33億美元，較今年2月約10億美元大幅跳升
- 公司前身：原名Flux Computing，2026年1月才改名為OLIX
DX-1晶片為何要「拋棄HBM」？
OLIX計畫用這筆資金開發首款推論加速晶片DX-1，主打不依賴高頻寬記憶體（HBM）的架構。HBM正是目前輝達GPU供應鏈最大的瓶頸之一，長期缺貨又昂貴，也是AI伺服器成本居高不下的主因之一。如果光子AI晶片真能繞開HBM依賴，對整條AI硬體供應鏈會是不小的鬆綁。不過DX-1要等到2027年下半年才會交給第一批客戶測試，離真正量產出貨還有一段距離。
光子AI晶片能打敗GPU嗎？
光子運算的物理優勢（光速傳輸、低發熱）不是新概念，過去十幾年一直卡在「實驗室能展示、量產做不出來」的老問題上。OLIX面對的挑戰是要在2027年前把技術從展示品變成能穩定出貨的商用晶片，同時說服客戶放棄已經高度成熟、生態系完整的GPU方案。對開發者跟企業而言，光子AI晶片目前仍是「值得追蹤、但先別急著壓身家」的技術方向，實際性能與良率，還是得等DX-1真正出貨才知道。
好不好用，試了才知道。
🇺🇸 OLIX Photonic AI Chip Review: $312M Bet Against Nvidia GPUs
UK photonic AI chip startup OLIX just closed a $312 million Series B, pushing its valuation to $3.3 billion — one of the biggest funding stories in AI hardware this month, second only to Nvidia headlines. The round pulled in a mix of new and returning investors, a sign the market is starting to take seriously the bet that light, not electricity, could run AI inference workloads.
What Is a Photonic AI Chip (OTPU)?
A photonic AI chip, or Optical Tensor Processing Unit (OTPU), runs the matrix multiplications behind AI inference using light instead of electrical signals. The pitch is simple: photons move faster, generate less heat, and use less power — directly attacking the cooling and electricity costs that plague today's GPU data centers. OLIX is betting its entire roadmap on this architecture to challenge Nvidia's grip on AI inference.
How Big Is OLIX's Series B?
The $312 million round brought in heavyweight new backers including chip IP giant Arm and quant trading firm Hudson River Trading, alongside existing investors Hummingbird Ventures, Crane, Plural, and Creandum doubling down. Netflix co-founder Reed Hastings joined as an angel investor. Notably, OLIX is also backed by the UK government's Sovereign AI venture fund — a signal that Britain wants its own AI hardware play outside the US-China chip race.
- Round size: $312M Series B
- Valuation: $3.3B, up sharply from roughly $1B in February 2026
- Backstory: OLIX was formerly known as Flux Computing before rebranding in January 2026
Why Does the DX-1 Chip "Ditch HBM"?
OLIX plans to use the fresh capital to build its first inference accelerator, DX-1, designed to run without dependence on High Bandwidth Memory (HBM) — currently the biggest bottleneck in Nvidia's GPU supply chain, chronically short and expensive. If photonic AI chips can genuinely sidestep HBM, it could meaningfully loosen a chokepoint across the whole AI hardware supply chain. But DX-1 won't reach its first customer test systems until the second half of 2027, so mass production is still a way off.
Can Photonic AI Chips Actually Beat GPUs?
The physics advantages of optical computing — light-speed transmission, low heat — aren't new; the field has spent over a decade stuck between "works in the lab" and "can't be manufactured at scale." OLIX's real challenge is turning a demo into a shippable commercial chip by 2027, while convincing customers to walk away from Nvidia's mature GPU ecosystem. For now, photonic AI chips are worth watching, not betting the farm on — actual performance and yield will only be clear once DX-1 ships.
好不好用，試了才知道。
Works or not — you won't know till you try it.
Sources / 資料來源
- Olix Raises $312M for Photonic AI Chip That Ditches HBM (Tech Times)
- Chip startup Olix raises $312m at $3.3bn valuation, backed by UK gov’t Sovereign AI venture fund (DCD)
- OLIX raises $312m at a $3.3bn valuation and appoints Professor Nick McKeown to its board (OLIX official)
常見問題 FAQ
光子AI晶片跟一般GPU差在哪？
光子AI晶片用光子而非電子訊號執行矩陣運算，理論上速度更快、發熱更低、耗電更省，但目前仍處於早期商業化階段，還沒有大規模量產的成熟案例。
OLIX的DX-1晶片什麼時候可以買到？
DX-1預計2027年下半年才會交付第一批客戶進行測試，目前還沒有正式量產與上市時間表。
OLIX融資背後有政府支持嗎？
有，這輪融資包含英國政府主權AI創投基金的資金，顯示英國政府有意扶植本土AI硬體產業。
光子運算未來會取代GPU嗎？
目前還言之過早。光子運算的物理優勢明確，但量產良率與生態系成熟度都還在早期，短期內GPU仍是AI運算主流。
延伸閱讀 / Related Articles
- OpenAI駭進Hugging Face評測：AI代理自主搞出資安事故 | OpenAI x Hugging Face Breach: AI Agents Went Rogue
- AI Token黑市評測：Poison Claude賤賣你的AI帳號 | AI Token Black Market Review: Poison Claude Sells Cheap Access
- Atlassian Rovo漏洞評測：AI代理零點擊外洩機密資料 | Atlassian Rovo Flaw: AI Agent Leaks Data With Zero Clicks
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
張貼留言