这背后是科技巨头们AI硬件竞争白热化，本周先有亚马逊扬言要建全球最大AI计算集群，紧接着Meta宣布建立百亿美元AI数据中心，马斯克旗下的xAI也传来要将其超级计算机项目扩建十倍的消息……The competition among technology giants in AI hardware has intensified. This week, amazon announced plans to build the world's largest AI computing cluster, followed by meta platforms announcing the establishment of a billion-dollar AI datacenter. Additionally, news came from xAI, owned by Musk, about its plans to expand its supercomputing project tenfold.
The AI arms race is in full swing, with giants like meta platforms, xAI, and amazon generously pouring money into it, igniting another round of AI momentum in usa stocks.
On Wednesday, usa chip stocks experienced a widespread rally,$PHLX Semiconductor Index (.SOX.US)$closing up 1.71%, the chip etf $iShares Semiconductor ETF (SOXX.US)$ closed up 1.71%. Chip companies$Marvell Technology (MRVL.US)$It closed up 23.19%, marking the best single-day performance since May 2023;$Micron Technology (MU.US)$It closed up 3.37%,$Broadcom (AVGO.US)$It closed up 1.43%, AMD closed up 1.42%,$Taiwan Semiconductor (TSM.US)$It closed up 0.91%.
Behind this is the intensified competition in AI hardware among tech giants; Amazon has announced plans to build the world's largest AI computing cluster, followed closely by Meta's announcement of a $10 billion AI datacenter, and news from Elon Musk's xAI about expanding its supercomputer project tenfold...
The AI arms race is heating up.
On Wednesday, Facebook's parent company $Meta Platforms (META.US)$ announced that it plans to spend 10 billion dollars to establish the largest AI datacenter in the company's history in northeastern Louisiana, USA.
Meta did not disclose how many GPUs the new facility will be able to support. Kevin Janda, Meta's datacenter strategy director, stated that the datacenter will cover an area of 4 million square feet (0.37 million square meters) and is expected to create 500 operational jobs and 5,000 temporary construction jobs.
On the same day, Elon Musk's AI startup xAI announced plans to expand its supercomputer Colossus in Memphis, Tennessee, to accommodate at least one million GPUs.
This move represents a massive expansion of Colossus, which currently has 100,000 GPUs to train xAI's chatbot Grok.
Meanwhile, on the previous day, amazon also announced a collaboration with Anthropic to build an AI supercomputer named UltraCluster, code-named Rainer. Amazon stated that this computer is expected to be equipped with hundreds of thousands of the latest AI training chips, Trainium 2, and when completed, will be five times larger than Anthropic's current most powerful model cluster, becoming the largest AI supercomputer globally.
For the past two years, amazon has relied on Anthropic's ai and$NVIDIA (NVDA.US)$hardware to drive its cloud sales growth, but its ai ambitions go beyond this. The company is developing more powerful next-generation chips in hopes of challenging nvidia's monopoly.
On Wednesday, Matt Garman, CEO of amazon web services, stated at the Re:Invent conference that amazon's new ai chip, Trainium 2, holds far greater value for ai developers compared to nvidia's flagship H100 chip, and the company plans to launch the next generation of more advanced 3-nanometer training chips, Trainium 3, by the end of next year.
In order to reduce reliance on nvidia, the ai large model giant OpenAI is also diving into chip development. In February this year, there were reports that OpenAI CEO Sam Altman planned to raise 7 trillion dollars to build a chip empire. However, the latest news shows that OpenAI has chosen a more pragmatic path, first bringing in partners to create its own chip, rather than establishing a network of chip factories.
Related reading:The second half of AI has officially begun! How to find the next nvidia? Goldman Sachs has named these 30 stocks that are expected to stand out.
Editor/rice
