In a spacious room within a one-story building in Santa Clara, California, a new supercomputer powered by specialized chips created by Cerebras, a Silicon Valley start-up, hummed behind sleek white cabinets. The unveiling of this impressive supercomputer, which took place on Thursday, marks another significant development in the field of artificial intelligence (A.I.). Each of Cerebras’ chips, about the size of a dinner plate and 56 times larger than a typical A.I. chip, packs the computational power equivalent to hundreds of traditional chips.
This supercomputer was custom-built for G42, an A.I. company, which plans to leverage its capabilities to develop and drive A.I. products specifically designed for the Middle East market. Andrew Feldman, the CEO of Cerebras, expressed the team’s intention to demonstrate that it is possible to create a dedicated A.I. supercomputer that is not only faster but also consumes less energy and is more cost-efficient.
The demand for computing power and A.I. chips has skyrocketed due to the global A.I. boom. Tech giants like Microsoft, Meta, and Google, as well as numerous start-ups, have rushed to introduce A.I. products in recent months following the viral success of the A.I.-powered chatbot ChatGPT. However, the development of A.I. products typically requires a substantial amount of computing power and specialized chips, triggering a fierce competition to secure these technologies. In fact, Nvidia, a leading chip maker for A.I. systems, announced in May that its sales would exceed Wall Street estimates by more than 50% due to the strong demand for its graphics processing units (GPUs). This surge in demand has prompted a race among companies like Cerebras, Graphcore, Groq, and SambaNova to break into the market long dominated by Nvidia.
Chips are poised to play a pivotal role in the A.I. landscape and can potentially reshape the power dynamics among tech companies and even nations. In fact, the Biden administration has considered imposing restrictions on the sale of A.I. chips to China, with concerns over China’s A.I. advancements posing national security risks for the United States. While A.I. supercomputers have been constructed in the past, it is uncommon for start-ups like Cerebras to undertake such endeavors.
Cerebras, headquartered in Sunnyvale, California, was founded in 2016 by Andrew Feldman and four other engineers. The company’s primary goal is to develop hardware that accelerates A.I. development. Over the years, Cerebras has secured $740 million in funding, including investments from Sam Altman of OpenAI and venture capital firms such as Benchmark. Notably, the company has been valued at $4.1 billion.
Unlike the small chips typically utilized to power A.I., often the size of a postage stamp, Cerebras unveiled the largest computer chip ever created in 2019. Feldman has stated that Cerebras’ chips can accelerate A.I. system training between 100 and 1,000 times faster than existing hardware. In 2021, G42, an Abu Dhabi-based company, commenced collaborating with Cerebras. They employed a Cerebras system to train an Arabic version of ChatGPT with remarkable success.
In May, G42 requested that Cerebras create a network of supercomputers in different locations worldwide. Talal Al Kaissi, CEO of G42, highlighted the cutting-edge technology’s capability to facilitate the development of chatbots and aid in the analysis of genomic and preventive care data using A.I. However, due to the high demand for GPUs, it was challenging to acquire enough to construct a supercomputer. Cerebras’ technology presented an available and cost-effective alternative, enabling them to build the supercomputer for G42 within a mere 10 days. Cerebras plans to construct two additional supercomputers for G42 in Texas and North Carolina over the next year, followed by six more distributed worldwide. This network of supercomputers has been named Condor Galaxy.
Chris Manning, a computer scientist at Stanford University specializing in A.I., suggests that start-ups may face challenges when competing against Nvidia. This is primarily because those building A.I. models are accustomed to utilizing software compatible with Nvidia’s A.I. chips. He remarks that other start-ups attempting to enter the A.I. chip market have largely struggled.
Nonetheless, Feldman remains optimistic, stating that many A.I. businesses are keen to avoid being solely dependent on Nvidia. Moreover, there is a global demand for other powerful chips like those developed by Cerebras. Feldman hopes that their advancements in A.I. will propel the industry forward.
Denial of responsibility! VigourTimes is an automatic aggregator of Global media. In each content, the hyperlink to the primary source is specified. All trademarks belong to their rightful owners, and all materials to their authors. For any complaint, please reach us at – [email protected]. We will take necessary action within 24 hours.
