OpenAI 同意使用 AMD 的電腦晶片

Recently the New York Times reported thee following:

OpenAI Agrees to Use Computer Chips From AMD

Weeks after a $100 billion agreement to use computer chips from Nvidia, OpenAI has signed a similar deal with a rival chipmaker.

The NYT - By Cade Metz Reporting from San Francisco

Oct. 6, 2025

Late last month, OpenAI announced a $100 billion agreement to use computer chips from Nvidia, the world’s most valuable publicly traded company.

Now, OpenAI has entered a similar agreement with AMD, one of the many chipmakers hoping to challenge Nvidia as the dominant supplier of chips used to power artificial intelligence technologies like OpenAI’s ChatGPT.

On Monday, OpenAI said it would begin using AMD chips in the second half of next year as it built computer data centers. The new facilities would be separate from the data centers that OpenAI has committed to building in Texas, New Mexico, Ohio and an unnamed site in the Midwest.

Over several years, OpenAI plans to deploy enough AMD chips to consume six gigawatts of power, an amount that could supply all the households in Massachusetts. As part of its recent agreement with Nvidia, OpenAI agreed to deploy enough chips to consume 10 gigawatts.

AMD is not investing in OpenAI. But the agreement allows OpenAI to buy up to 160 million shares in the chipmaker at a penny per share, which would give OpenAI a 10 percent stake in the company. It could also supply OpenAI with additional capital as it works to build computing facilities over the next several years.

The unorthodox arrangement — the latest in a series of unusual financing deals that OpenAI has made in recent months to support its plans to build data centers across the world — could help AMD gain a foothold in the market for A.I. chips that has been dominated by Nvidia. By some estimates, Nvidia sells 90 percent of graphics processing units chips that are vital to the development of A.I. technologies.

By essentially giving away shares to OpenAI, AMD has closely aligned itself with the company that is leading the A.I. boom. And it now has a major customer for its own GPUs.

Until now, the only limitation on Nvidia’s growth has been its ability to produce enough chips to meet demand. But several other companies, including AMD, tech giants like Google and Amazon and start-ups like Cerebras and Groq, are racing to challenge its dominance.

AMD’s shares jumped more than 20 percent in premarket trading on Monday, as did the share price of several other chipmakers. Nvidia’s stock was down slightly.

“This partnership is a major step in building the compute capacity needed to realize A.I.’s full potential,” OpenAI chief executive, Sam Altman, said in a statement.

The agreement is part of a wider effort among tech companies to spend hundreds of billions of dollars on the construction of data centers. OpenAI, Amazon, Google, Meta and Microsoft plan to spend more than $325 billion combined on these facilities by the end of this year alone.

Tech giants, which pull in tens of billions of dollars in profits each year, have been able to finance data center construction with cash. But as newer and smaller companies like OpenAI have built computing facilities, they have been forced to raise or borrow tens of billions of dollars.

Through its Stargate Project, OpenAI previously said it was working with the cloud computing company Oracle and the Japanese conglomerate SoftBank to spend more than $400 billion on new data centers in the United States. But the start-up and its partners do not have the money needed to pay for these data centers, so it has looked for creative ways of bridging the financial gap.

In its recent deal with Nvidia, OpenAI agreed to deploy Nvidia’s chips. It also received a $100 billion investment from Nvidia. After investing an initial $10 billion in OpenAI, the chipmaker plans to invest an additional $90 billion in the company over the next several years. The agreement was the latest example of OpenAI’s raising money from the companies it relies on for products and services.

Translation

OpenAI 同意使用 AMD 的電腦晶片

在與英偉達簽署價值 1000 億美元的電腦晶片使用協議幾週後，OpenAI 又與另一家晶片競爭對手簽署了類似的協議

上個月底，OpenAI 宣布了一項價值 1000 億美元的協議，將使用全球市值最高的上市公司英偉達的電腦晶片。

現在，OpenAI 又與 AMD 達成了類似的協議。 AMD 是眾多希望挑戰英偉達在人工智能晶片領域主導地位的晶片製造商之一，這些晶片用於支援 OpenAI 的 ChatGPT 等技術。

週一，OpenAI 表示，將在明年下半年開始使用 AMD 晶片建造電腦數據中心。新設施將與 OpenAI 承諾在德克薩斯州、新墨西哥州、俄亥俄州以及中西部未透露名稱的地點建設的數據中心分開。

OpenAI 計劃在未來幾年內部署足夠多的 AMD 晶片，需要耗用 6 千兆瓦的電力，這耗電量足以滿足馬薩諸塞州所有家庭的用電需求。作為其最近與英偉達達成協議的一部分，OpenAI 同意部署足夠多的晶片，需要耗用 10 千兆瓦的電力。

AMD 並未向 OpenAI投資。但該協議允許 OpenAI 以每股 1 美分的價格收購這家晶片製造商最多 1.6 億股股票，這將使 OpenAI 擁有該公司 10% 的股份。這筆交易還可以為 OpenAI 提供額外資金，幫助其在未來幾年內建造運算設施。

這項異常的安排 - 這是 OpenAI 近幾個月為支持其在全球建設數據中心的計劃而達成的一系列不尋常的融資交易中的最新一筆 - 可能有助於 AMD 在一直由英偉達主導的人工智能晶片市場站穩陣腳。據估計，英偉達銷售了對人工智能技術發展至關重要的圖形處理單元 (GPU) 晶片的90%。

實際上AMD是將股份贈予 OpenAI，從而與引領人工智能熱潮的公司建立了緊密的聯繫。現在，AMD 的 GPU 也迎來了一位重要客戶。

到目前為止，限制英偉達成長的唯一因素是其晶片產量能否滿足需求。但包括 AMD、Google和亞馬遜等科技巨頭以及 Cerebras 和 Groq 等初創公司在內的其他幾家公司正在競相挑戰其主導地位。

AMD 的股價在周一盤前交易中上漲了 20% 以上，其他幾家晶片製造商的股價也出現了上漲。英偉達的股價略為下跌。

OpenAI 執行長 Sam Altman 在一份聲明中表示： “此次合作是去建立充分發揮人工智能潛力所需計算能力的重要一步。”

這項協議是科技公司斥資數千億美元建造數據中心的廣泛努力的一部分。 OpenAI、亞馬遜、Google、Meta 和微軟計劃僅在今年底就在這些設施上投入超過 3,250 億美元。

科技巨頭每年賺取數百億美元的利潤，能夠用現金資助數據中心的建置。但隨著像 OpenAI 這樣的新興小型公司紛紛建造運算設施，科技巨頭被迫籌集或借入數百億美元的資金。

OpenAI 先前曾表示，透過其「星際之門」項目，它正與雲端運算公司甲骨文和日本軟銀集團合作，在美國投資超過 4,000 億美元建設新的數據中心。但這家初創公司及其合作夥伴缺乏建造這些數據中心所需的資金，因此一直在尋找有創意的方法來彌補資金缺口。

OpenAI在最近與英偉達達成的協議中， 同意部署英偉達的晶片。此外，它還獲得了英偉達 1000 億美元的投資。英偉達在向 OpenAI 投資 100 億美元後，還計劃在未來幾年內再向該公司投資 900 億美元。這項協議是 OpenAI 從其所依賴的產品和服務公司籌集資金的最新例證。

So, OpenAI previously has announced a $100 billion agreement to use computer chips from Nvidia.  Now, OpenAI enters a similar agreement with AMD, a chipmaker that is hoping to challenge Nvidia as the dominant supplier of chips. While big companies are earning tens of billions of dollars in profits each year and are able to finance data center construction with cash, a start-up and its partners do not have the money needed to pay for these data centers, so it has to look for creative ways to bridge the financial gap.