It isn’t an exaggeration to say the tech sector’s recent rally might depend on Nvidia beating expectations for its earnings report this week. However, stockholders in the chip maker should be looking beyond the immediate horizon and particularly at potential competitive threats.
Nvidia’s
dominance in supplying the chips used to train artificial-intelligence systems is expected to shine through again when it reports January-quarter earnings on Wednesday, with its sales expected to more than triple from the prior year to more than $20 billion.
However, Nvidia’s growth rate is expected to slow from around the July quarter as it faces tougher comparisons since the beginning of massive spending on its chips.
“There’s a possibility that you could even see quarter-on-quarter declines as you start to get to the end of this massive cycle. Now, this isn’t to suggest that an Nvidia isn’t an incredibly well positioned company, but you have also competition entering the space and competition has been late,” Daniel Newman, CEO of The Futurum Group, told Barron’s
Newman noted that data indicated Nvidia might currently have somewhere from 98% to 99% of market share in graphics-processing units, or GPUs used for AI and companies rarely hold on to such a level of dominance.
Signs of eagerness to disrupt Nvidia’s AI chip leadership are multiplying.
SoftBank Group
founder Masayoshi Son is considering the creation of a $100 billion chip venture, Bloomberg reported over the weekend that citing people with knowledge of the matter. SoftBank could provide around $30 billion and seek $70 billion from outside investors such as Middle Eastern institutions to form a company that would make AI chips and sit alongside the Japanese company’s majority stake in chip designer
Arm Holdings,
according to Bloomberg.
Arm declined to comment on the report and SoftBank didn’t immediately respond to a request for comment early on Monday. SoftBank shares rose 2.8% in Tokyo. Shares of Arm and Nvidia weren’t trading Monday due to the Presidents Day holiday.
It’s easy to talk about eating into Nvidia’s market share and less easy to actually do so, as rivals, such as
Advanced Micro Devices,
have found. Nvidia benefits from developers’ familiarity with the company’s CUDA software for building applications. However, there is clearly an appetite from major technology companies to have alternative suppliers of AI chips, and some have established their own in-house operations.
That could mean significant interest in British AI chip company Graphcore, which was reported by The Sunday Telegraph newspaper to be exploring a sale, citing industry sources. Graphcore has sold its hardware primarily to AI start-ups looking to build and train models at a lower cost but a deal to supply Microsoft fell through and the company has faced increasing losses.
Arm, SoftBank, and Microsoft-backed OpenAI have been rumored as potential buyers for Graphcore, according to The Sunday Telegraph.
Elsewhere, AI hardware start-up Groq was generating plenty of interest over the weekend after an appearance on CNN by its CEO Jonathan Ross, who said the company was setting speed records for inference—the process of generating answers or results from AI models. Analysts expect demand for AI chips that can support inference—as opposed to training—to rise as a proportion of the industry in the next few years.
Groq claims its language-processing units are better suited for powering AI language applications than GPUs of the type Nvidia makes. Ross was formerly at
Alphabet’s
Google, where he was one of the builders of its custom tensor processing units.
“Is it the end of Nvidia? No… but will having 5-to-7 custom and direct [chip making] competitors depress margins and cause a certain percentage of the demand for Nvidia fall off, yes it will,” Futurum’s Newman said.
Write to Adam Clark at [email protected]
Read the full article here
