AI: Nvidia rivals jostle for ground in chip wars – 06/17/2023 – Tech

In the three weeks since Nvidia shocked the tech world with its prediction of an unprecedented jump in sales, Wall Street has been looking for other chip companies that could benefit from the latest success of artificial intelligence. But as the search continued, the gulf that opened up between Nvidia and the rest of the chip industry only widened.

In one of the most anticipated attempts to catch up with Nvidia, rival AMD showed off a new AI chip this week called the MI300X. The chip includes a GPU – an acronym for Graphics Processing Unit, a product originally designed for video games that is at the heart of Nvidia’s success–, as well as a CPU. [Unidade Central de Processamento] more general purpose and integrated memory to provide data to both processors.

The design reflects attempts by chip makers to bundle different technologies together in search of the most efficient way to process the massive amounts of data needed to train and apply the large models used in generative AI.

AMD claimed impressive performance for its new chip, which it said would outperform Nvidia’s flagship H100 by several measurements. But it failed to pinpoint any potential customers who would be considering the chip, and only highlighted the product’s ability to handle AI inference – applying pre-trained AI models – rather than the more intensive training work, which explains the problem. increase in Nvidia sales. It also said it won’t start ramping up production of the new chip until the last quarter of this year.

When AMD’s new chip becomes available in the first half of 2024, Nvidia’s H100 will have been on the market for 18 months, giving it a big head start, said Stacy Rasgon, an analyst at Bernstein. AMD is “far behind. They can have the bagasse [do mercado de IA] —although perhaps even that is enough” to justify Wall Street’s recent enthusiasm for the company’s stock, he said.

“Nvidia is free and clear this round” of the chip wars that have erupted over AI, added Patrick Moorhead, an analyst at Moor Insights & Strategy.

Wall Street has highlighted several chip companies that could get a boost from generative AI. The combined stock market value of AMD, Broadcom and Marvell jumped $99 billion, or 20%, in the two days following Nvidia’s impressive sales forecast last month. But its AI-related sales are not expected to come from the market dominated by Nvidia.

Broadcom, for example, should benefit from growing demand for its data communications products, as well as its work with Google on the development of an internal data center chip known as a TPU. [Unidade de Processamento Tensor]. Earlier this month, Broadcom predicted that AI-related businesses would make up around 25% of its revenue in 2024, up from just 10% last year.

However, the processors used to train and apply large AI models are seeing the biggest increase in demand and generating the most excitement in the stock market. While AMD’s new chip disappointed Wall Street, Nvidia’s market value rose again to more than $1 trillion, a level it first reached two weeks ago.

“There is no doubt that AI will be the main driver of silicon consumption for the foreseeable future,” with data centers being the main focus of investment, said AMD Chief Executive Lisa Su. She predicted that the market for AI accelerators — GPUs and other specialized chips designed to speed up the data-intensive processing needed to train or run — would soar from $30 billion this year to more than $150 billion by 2027.

As they struggle to compete with Nvidia in the most advanced AI chips, companies like AMD and Intel are counting on an evolution in the generative AI market to increase demand for other types of processors. Large-scale models such as OpenAI’s GPT-4 dominated the early stages of the technology, but a recent explosion in the use of smaller, more specialized models could lead to increased sales of less powerful chips, they say.

Many customers who want to train models with their corporate data also want to keep their information close to home rather than risk handing it over to companies that provide large AI models, said Kavitha Prasad, vice president at Intel. Coupled with all the computing work needed to prepare the data to feed the training, this will create a lot of work for the CPUs and AI accelerators made by Intel, she said.

However, the rapidly changing demands on data centers caused by the rise of services like ChatGPT have left chip makers struggling to predict how their markets will evolve. CPU sales could even decline in the coming years as data center customers invest in AI accelerators, Rasgon said.

Rivals hoping to get a bite out of Nvidia’s growing AI business face an equally big challenge when it comes to software. The widespread use of Nvidia’s chips in AI and other applications owes much to the ease with which its GPUs, originally designed for video games, can be programmed for other tasks using its Cuda software.

In an attempt to attract more developers to its AI chips, AMD this week highlighted its efforts to work with PyTorch, a widely used AI framework. However, there is still a long way to go to match the many software libraries and applications that have already been developed for Cuda, Rasgon said. “It will take a decade” for rivals to match Nvidia’s software – a period during which the company will continue to move quickly to increase its lead, he said.

“Nobody wants an industry where one player dominates,” Moorhead said. But for now, the booming market for chips capable of handling generative AI belongs to Nvidia.