NVIDIA (semiconductor company)Yīngwěidá英伟达atzài在GTC 2026 (NVIDIA annual AI conference)GTC 2026GTC 2026conference / conventiondàhuì大会to launch / releasetuīchū推出Vera Rubin (NVIDIA AI GPU platform)Vera RubinVera Rubinplatformpíngtái平台，strategic focuszhànlüè zhōngxīn战略重心shifting towardzhuǎnxiàng转向AI (artificial intelligence)AIAIinferencetuīlǐ推理。
March 18, 2026 at 2:00 PM
New Words
News in English
Science & TechnologyNVIDIA Unveils Vera Rubin Platform at GTC 2026, Shifting Focus to AI Inference
NVIDIA unveiled the Vera Rubin GPU platform at GTC 2026, positioning it as the core infrastructure for next-generation agentic AI inference. The company also revealed the Feynman architecture roadmap for 2028, which will feature 3D chip stacking and TSMC A16 process technology.
In mid-March, NVIDIA officially launched the next-generation Vera Rubin GPU platform at the GTC 2026 annual conference.
This platform takes the form of a rack-scale supercomputer, deeply integrating the self-developed Vera CPU with the Rubin GPU architecture.
NVIDIA CEO Jensen Huang stated that this platform is specifically designed for large-scale agentic AI inference tasks. Inference performance has improved several times over compared to the previous generation.
The conference also released signals of the company's strategic transformation. NVIDIA clearly indicated that it will shift its business focus from AI training toward large-scale inference and agentic workflows.
Regarding the hardware roadmap, NVIDIA disclosed the next-generation Feynman GPU architecture, expected to be launched in 2028. Feynman will adopt 3D chip stacking design, with the process technology upgraded to TSMC's 1.6 nanometer, marking NVIDIA's first sub-1-nanometer class product.
Analysts believe that this launch further consolidated NVIDIA's leading position in the artificial intelligence infrastructure market.
More in Science & Technology
former / previousqián前Groq (AI chip startup)GroqGroqengineersgōngchéngshī工程师regarding / over (a matter)jiù就$20 billion USD200 yì měiyuán200亿美元Nvidia (US semiconductor company)Yīngwěidá英伟达licensing / authorizationshòuquán授权deal / transactionjiāoyì交易to file / to initiatetíqǐ提起a lawsuit / legal actionsùsòng诉讼。
Former Groq Engineers Sue Board Over $20 Billion Nvidia Licensing Deal
Qualcomm (US semiconductor company)Gāotōng高通for the first timeshǒucì首次to / towardxiàng向Huawei (Chinese tech company)Huáwéi华为to pay / to make a paymentzhīfù支付patent royalties / licensing feezhuānlì fèi专利费，both partiesshuāngfāng双方to sign / to concludeqiānshǔ签署multi-year periodduō nián qī多年期agreement / dealxiéyì协议。
Qualcomm Pays Huawei for First Time in Broad Multi-Year Patent Licensing Agreement
Tencent (Chinese tech company)Téngxùn腾讯from / towardxiàng向Oracle (US tech corporation)Jiǎgǔwén甲骨文to lease / to rent for usezūyòng租用one hundred thousandshí wàn十万measure word for small round objectskē颗artificial intelligenceréngōng zhìnéng人工智能semiconductor chip / microchipxīnpiàn芯片。
Tencent Leases 100,000 AI Chips From Oracle in $7 Billion Deal
Apple Inc.Píngguǒ Gōngsī苹果公司to announcexuānbù宣布will (future marker)jiāng将on / at (preceding a date)yú于October 13thshíyuè shísān rì十月十三日to hold (an event)jǔxíng举行Welcome Home (name of Apple event)Welcome HomeWelcome Homeproduct launch eventfābùhuì发布会，expected toyùjì预计to launch / to introducetuīchū推出HomePad mini (Apple smart home display)HomePad miniHomePad miniand others / etc.děng等new-modelxīnkuǎn新款productschǎnpǐn产品。
Apple Announces October 13 Welcome Home Event, Expected to Reveal HomePad Mini and Other New Products