The name Nvidia comes from the Latin word for jealousy, “invidia”. And the semiconductor giant has something to envy. Companies rabidly wolfed all of it up these are very expensive AI chips within the race to develop latest generative artificial intelligence technologies. Over the last yr, Nvidia’s stock is up 220%. revenues skyrocketeddoing this the third most beneficial company on this planet. That’s practically “printing money,” said one analyst.
Now other tech giants are introducing their very own AI chips: MTIA MetaMicrosoft’s Maia, Amazon’s Trainium and Google TPU. Not to say Intel and Advanced Micro Devices, semiconductor chip makers which are direct competitors of Nvidia. Over the course of two days this week, Facebook parent Alphabet, Meta and Intel unveiled latest AI hardware as a part of a growing challenge to Nvidia’s dominance.
But Nvidia won’t fall easily.
Nvidia’s growing competition
According to DA Davidson analyst Gil Luria, two-thirds of Nvidia’s revenue comes from its five largest customers. These include Meta, Microsoft, Google and Amazon. All 4 have began producing AI hardware themselves, which could seriously threaten Nvidia’s revenues in the longer term.
Google presents its version of TPU v5p on Tuesday, followed by Meta’s release of the following version of MTIA on Wednesday was the newest sign that buyers of Nvidia’s AI chips have gotten more self-sufficient. Well-known artificial intelligence startups are already using Google services powered by the brand new hardware as an alternative of Nvidia’s AI chips, often known as GPUs or graphics processing units. Anthropic uses Google technology to power its Claude AI chatbot, while Hugging Face and AssemblyAI use it to reinforce image generation and speech recognition capabilities.
“Nvidia was able to generate a tremendous amount of windfall profits over the last few years by having the right product at the right time,” Luria said in an interview. “But now that the market is the size it is, [there are] many companies that are able to replace Nvidia with other products – most importantly, Nvidia’s own customers.”
Nvidia’s semiconductor rivals Intel and AMD are also upping their AI chip game. Intel introduced the Gaudi 3 AI accelerator at Tuesday’s Intel Vision 2024 conference, stating that it has higher performance and efficiency in training AI models than industry alternatives – including Nvidia’s highly sought-after H100 chip. Intel said its latest chip has more AI processing power, network bandwidth and memory than its predecessor, Gaudi 2.
Intel also said its Gaudi 3 accelerator can train language learning models (LLM) that power AI chatbots 50% faster than Nvidia’s popular H100 chip for $40,000and provide 30% faster “inference” (or using data for forecasting) for one of the best LLM in comparison with Nvidia H200 chip. The company is waiting for Nvidia to release performance results newly announced Blackwell chip before he can compare it with Gaudi 3.
Intel CEO Pat Gelsinger told reporters on Tuesday that the corporate has its own “innovation roadmap” that features improving the Gaudi 3 accelerator in collaboration with customers and business partners because it competes with Nvidia products.
“Without a doubt, the Gaudi 3 significantly outperforms the H100 today,” Gelsinger said, adding that TCO, or total cost of ownership, “is a huge difference” between the 2 chipmakers.
While he declined to disclose pricing details for the Gaudi 3, Gelsinger said Intel is “very happy” that it will be “significantly lower than the prices quoted for the H100 and Blackwell.”
But the problem of Intel competing with Nvidia is not just about performance — it is also about cost, availability and access to data, said Christoph Schell, Intel’s chief industrial officer.
“We are checking the box for each of them,” Schell said, “well ahead of Nvidia.”
After an enormous rally for early 2024 Nvidia stock has fallen into correction territory on Tuesday, down 10% from its recent high as Google and Intel targeted its AI rainmaker.
But Nvidia won’t be losing its AI shine any time soon
The correction didn’t last long: on Wednesday, Nvidia’s shares were up 2%, even in someday. there was a sell-off within the broader market.
According to Bank of America analysts, Nvidia has a 75% share within the AI accelerator market. Most corporations already know tips on how to produce AI chips, said Benjamin Lee, a professor of engineering and computer science on the University of Pennsylvania. He said the important advantage of Nvidia chips is software libraries that make it easier for developers to construct AI applications without having to write down code from scratch.
For now, Intel and AMD are simply playing catch-up, Lee said.
“Nvidia has been working on this for at least the last 10 years,” Lee said, which “reflects the maturity of their toolkit” in comparison with Intel, which has just released its first AI chip in 2019.
Meanwhile, Lee said, Nvidia customers are creating AI accelerators mainly to scale back their very own costs, not to compete directly with Nvidia.
On the opposite hand, Luria noted that Google already sells services that run on its TPU, i.e tensor processing units. Amazon is investing $4 billion in Anthropic stipulates that future AI models will only run on Amazon chips. Luria said Nvidia’s software advantage won’t last without end because customers are highly motivated to seek out ways to scale back chip spending by developing their very own AI products and supporting Nvidia’s competitors.
Luria cited Microsoft for instance.
“When Microsoft was spending less than $2 billion a year on Nvidia GPUs, it might not necessarily have had to spend as much effort on its own chips,” he said. “But last year they spent $11.5 billion on GPUs. It creates a sense of urgency.”
Luria said the identical is true for Amazon, Google and Meta.
“Now that the spending is so high,” he said, “they are working much harder to replace Nvidia with their own custom chips that are made to order for [their] AI Applications.”
Credit : qz.com
