In the rapidly evolving landscape of the artificial intelligence chip industry, Nvidia continues to distance itself from its competitors. This divide became even more pronounced in the wake of Nvidia’s unexpected sales surge, as rivals scramble to gain ground.
AMD’s foray into AI chips fails to impress
Nvidia’s key competitor, AMD, recently unveiled its latest AI chip, the MI300X, aiming to close the gap with the former. This new offering is a reflection of the industry’s trend towards combining diverse technologies for optimal data processing, crucial in generative AI’s big data requirements.
This chip, armed with a GPU – originally a video gaming innovation, but now the backbone of the company’s success – also comes equipped with a general-purpose CPU and inbuilt memory for both processors.
Despite AMD’s claims of the MI300X surpassing Nvidia’s flagship H100 in various aspects, they fell short of showcasing any prospective customers contemplating this chip.
Moreover, AMD stressed the chip’s capacity for AI inferencing, instead of training, the latter being a significant driver behind Nvidia’s ballooning sales. Furthermore, AMD won’t kickstart the new chip’s production until this year’s end, paving the way for Nvidia to retain its significant lead.
Bernstein analyst Stacy Rasgon opines that when AMD’s new chip hits the market next year, Nvidia’s H100 will have been available for 18 months, further solidifying its advantage.
With AMD considerably behind, Rasgon speculates they might secure the AI market’s leftovers, which might be sufficient to sustain Wall Street’s current bullish outlook on AMD’s shares.
Nvidia maintains dominance amidst Wall Street speculation
Following Nvidia’s astounding sales forecast last month, Wall Street was abuzz with speculation about which chip companies could ride the AI wave.
Companies like AMD, Broadcom, and Marvell saw a combined $99bn or 20% boost in their stock market value. However, the AI sales they anticipate aren’t expected to rival Nvidia’s market share.
Broadcom, for instance, is likely to capitalize on increasing demand for its data communication products and partnership with Google on an in-house data center chip, the TPU.
Still, while these companies are hopeful of a boost from generative AI, none seem poised to shake Nvidia’s dominance. In fact, as AMD’s new chip underwhelmed investors, the company’s stock market value soared past the $1tn mark again.
The processors used for training and applying large AI models are witnessing an explosive demand surge and generating the most stock market excitement.
As other companies struggle to match Nvidia’s advanced AI chips, they hope that the generative AI market’s evolution will bolster demand for other processor types.
AMD CEO Lisa Su believes the AI accelerator market – GPUs and other specialized chips designed to accelerate data-intensive processing – will catapult from $30bn this year to over $150bn in 2027.
Despite this, the rapid shifts in data center demands, largely driven by services like ChatGPT, have left chip manufacturers struggling to predict their markets’ evolution. As per Rasgon, CPU sales could even dip in the coming years as customers prioritize investment in AI accelerators.