Nvidia Launches New Chip Platform To Cash In on Generative AI Demand

Nvidia Corp. (NVDA) announced on Tuesday the release of a “groundbreaking” next-generation chip as it seeks to fend off growing competition within the booming AI industry.

The Santa Clara, California-based firm said its next-generation GH200 Grace Hopper platform would rely on the world’s first HBM3e processor, which is 50% faster than current HBM3 technology. The dual configuration will add up to 3.5x more memory capacity and 3x more bandwidth than the current chips on the market, allowing developers to run expanded Large Language Models (LLMs).

Speaking at a conference on Tuesday, CEO Jensen Huang said the new technology would assist the “scale-out of the world’s data centers.” He added that “the inference cost of large language models will drop significantly,” referring to the generative stage of AI computing that follows the training of LLMs.

Nvidia’s latest product launch comes after the buzz over AI technology rocketed the company to a $1 trillion valuation in May. Surging demand for its GPU chips and a predicted transition in data center infrastructure to accelerated computing saw Nvidia emerge as one of 2023’s brightest market stars.

But its massive success has spurred competition. Fellow chipmaker Advanced Micro Devices (AMD) announced the release of its own GPU chip in June, the MI300X, to challenge Nvidia’s dominance.

Nvidia will release its second-quarter earnings on August 23 and BofA analysts expect “less shock and awe” than last quarter when it raised second-quarter revenue guidance by 50% to $11 billion. However, analysts are still bullish on its longer-term prospects with Mizuho saying the company could generate $300 billion in revenue by 2027, ten times the $27 billion it made last year.