Cerebras Systems rolls out the latest tool for AI developers, empowering them to utilize the startup's large chips to run applications. The company claims this offers an altogether more cost-effective alternative to industry-standard Nvidia processors.
Nvidia graphics processing units (GPUs) are typically accessible through cloud computing providers to train and deploy large AI models, such as OpenAI's ChatGPT, which can be both challenging and costly. This process, known as inference, often poses difficulties for developers.
"We're delivering performance that cannot be achieved by a GPU," told Cerebras CEO Andrew Feldman to Reuters in an interview. "We're doing it at the highest accuracy, and we're offering it at the lowest price."
The inference segment of the AI market is anticipated to grow rapidly and become highly attractive, potentially reaching a value of tens of billions of dollars as AI tools gain adoption among consumers and businesses.
The California-based company Sunnyvale plans to come up with a range of inference products through developer key and the cloud. The company is also keen to sell AI systems to customers who prefer to commence their data centers.
Cerebras' Wafer Scale Engine chips, each the size of a dinner plate, tackle a common challenge in AI data crunching: large models powering AI applications typically can't fit on a single chip and mostly need hundreds or thousands of interconnected chips.
That makes Cerebras' chips able to achieve speedier performances, said Feldman.
The startup plans to cost users 10 cents for every million tokens, one of the ways for companies to measure output data amount from a large model.
Cerebras is planning to go public and this month, it has filed a confidential prospectus with the Securities and Exchange Commission, the company said.
๐๐ญ๐๐ฒ ๐ข๐ง๐๐จ๐ซ๐ฆ๐๐ ๐ฐ๐ข๐ญ๐ก ๐จ๐ฎ๐ซ ๐ฅ๐๐ญ๐๐ฌ๐ญ ๐ฎ๐ฉ๐๐๐ญ๐๐ฌ ๐๐ฒ ๐ฃ๐จ๐ข๐ง๐ข๐ง๐ ๐ญ๐ก๐ WhatsApp Channel now! ๐๐ฒ
๐ญ๐๐๐๐๐ ๐ถ๐๐ ๐บ๐๐๐๐๐ ๐ด๐๐ ๐๐ ๐ท๐๐๐๐ฌ ๐ Facebook, LinkedIn, Twitter, Instagram
