Start up company Cerebras launches new AI giant chip, challenging Nvidia
Cerebras Systems is a startup that aims to challenge Nvidia in the field of artificial intelligence (AI) computing. The company has launched a new chip that it claims will outperform its competitors in running AI models and generating responses. Cerebras has also launched a tool for AI developers that allows them to access ultra large chips to run applications.
Accessing NVIDIA GPUs to train and deploy large AI models is difficult and costly, and developers refer to this process as inference. Cerebras offers the chip as part of an AI computing system, which data center operators can purchase and operate on their own.
Cerebras has submitted a confidential plan to sell its shares in an initial public offering (IPO) in an attempt to gain a share of the booming market. Tech giants are investing billions of dollars in AI computing. So far, the biggest beneficiary is Nvidia, whose graphics processing unit (GPU) is an important component of this new infrastructure.
However, Andrew Feldman, founder and CEO of Cerebras, said that the company's computers would subvert the entire industry by improving the response speed of AI systems - he compared this transformation to the transformation to high-speed Internet.
Cerebras' method relies on giant chips made from a single silicon wafer, each of which is the size of a tray and is called a wafer level chip, thus avoiding a problem in AI data processing: data processed by large models that support AI applications often cannot be placed on a single chip and may require hundreds or thousands of chips to be connected in series.
Cerebras stated that this new technology makes chips more powerful than traditional chips. But the company must provide computers specifically designed to accommodate ultra large chips, as traditional hardware cannot accommodate them.
According to Andrew Feldman, a key advantage comes from the way their product uses memory. This capability is built into the Cerebras chip, unlike GPUs and other processors, which need to be connected to memory through interfaces to access information.
It can be certain that Nvidia has a huge lead in AI infrastructure, while other competitors including Intel have been striving to compete. Cerebras must prove to the computing industry that it can reliably produce and deploy its technology.
Cerebras is building its own data center to provide AI computing services. It also attempted to sell its chips to cloud providers, including Microsoft and Amazon. This startup has already approached these tech giants, but has not yet acquired these clients.
When asked how much market share Cerebras can take away from Nvidia, Andrew Feldman said, "It's enough to make them angry
Cerebras has submitted a confidential plan to sell its shares in an initial public offering (IPO) in an attempt to gain a share of the booming market. Tech giants are investing billions of dollars in AI computing. So far, the biggest beneficiary is Nvidia, whose graphics processing unit (GPU) is an important component of this new infrastructure.
However, Andrew Feldman, founder and CEO of Cerebras, said that the company's computers would subvert the entire industry by improving the response speed of AI systems - he compared this transformation to the transformation to high-speed Internet.
Cerebras' method relies on giant chips made from a single silicon wafer, each of which is the size of a tray and is called a wafer level chip, thus avoiding a problem in AI data processing: data processed by large models that support AI applications often cannot be placed on a single chip and may require hundreds or thousands of chips to be connected in series.
Cerebras stated that this new technology makes chips more powerful than traditional chips. But the company must provide computers specifically designed to accommodate ultra large chips, as traditional hardware cannot accommodate them.
According to Andrew Feldman, a key advantage comes from the way their product uses memory. This capability is built into the Cerebras chip, unlike GPUs and other processors, which need to be connected to memory through interfaces to access information.
It can be certain that Nvidia has a huge lead in AI infrastructure, while other competitors including Intel have been striving to compete. Cerebras must prove to the computing industry that it can reliably produce and deploy its technology.
Cerebras is building its own data center to provide AI computing services. It also attempted to sell its chips to cloud providers, including Microsoft and Amazon. This startup has already approached these tech giants, but has not yet acquired these clients.
When asked how much market share Cerebras can take away from Nvidia, Andrew Feldman said, "It's enough to make them angry