(Reuters) .- Alphabet, Google Matrix, prepares to partner with Taiwan MediaTk about the next version of its artificial intelligence (AI) chips, processing processing units, which will be manufactured next year, The Information reported to name people involved in the project.
However, Google has not cut ties with Broadcom, the chips designer with whom he has worked exclusively on AI chips in recent years, the report points out, citing an employee of the company based in San José.
Like Nvidia, Google also designs its own artificial intelligence server chips, which uses internal research and development and also rents cloud customers.
This approach gives Google a competitive advantage in the AI race by reducing its Nvidia dependence, even when rivals as Microsoft OpenAi and target have seen an increase in the demand for Nvidia chips.
Lee: ‘The role of artificial intelligence will be crucial’: CMO de Banamex
At the end of last year, Google launched its sixth generation TPU in an attempt to offer itself its cloud customers an alternative to the Nvidia chips, which are the most wanted processors in the industry.
Google chose Mediatek in part because the Taiwanese firm has a strong relationship with TSMC and charges you less for chip compared to Broadcom, added the information report.
Google spent between 6,000 and 9,000 million dollars on TPU last year, according to the OMDIA research firm, based on Broadcom’s goal for revenue from IA semiconductors last year.
Follow us on Google News to always keep you informed
