GPUs encounter bad news again, Amazon AI chips win big orders

Latest update time：2024-10-16

Reads：

????If you hope to meet more often, please mark the star ?????? and add it to your collection~

Source: Content Translated from WSJ , thank you.

Amazon and startup Databricks have reached a five-year agreement that could lower costs for businesses looking to build their own artificial intelligence capabilities.

Databricks will use Amazon’s Trainium AI chips to power a service that helps companies customize AI models or build their own. Amazon said its in-house chips will cost customers less than rivals that dominate the AI chip market, such as Nvidia’s graphics processing units (GPUs).

Amazon and Databricks declined to disclose the financial terms of the agreement.

The deal comes as Databricks, Amazon and other enterprise technology companies such as Microsoft, Salesforce and Databricks rival Snowflake are aggressively courting AI business. At the same time, enterprise technology executives say now is the time to prove that AI investments are paying off.

Databricks spent about $1.3 billion last year to acquire AI startup MosaicML and is expanding the acquired company’s services to get a piece of enterprise AI deals. Naveen Rao, Databricks vice president of generative AI, said the partnership with Amazon will ultimately make enterprise AI faster and cheaper because it can pass on the cost savings of using Amazon’s AI chips.

Early AI successes relied on using a company’s private data to customize the AI. For example, building a custom customer service chatbot could help reduce employee costs.

For Amazon, this means continuing to position itself as a neutral provider of AI technology, offering businesses the ability to use and customize a variety of AI models from a wide range of vendors on its platform.

Databricks makes money by renting out analytics, artificial intelligence and other cloud-based software that helps companies build enterprise technology tools using AI data. The San Francisco-based company said last September that it was valued at $43 billion.

The two companies currently have a partnership that allows customers to run Databricks data services on Amazon Web Services, Amazon's cloud computing platform. Databricks also rents Nvidia GPUs through AWS and will use more GPUs under the deal. Rao said that customers using AWS have generated more than $1 billion in revenue for Databricks, and AWS is the fastest growing cloud partner for the data company.

WW Grainger Chief Technology Officer Jonny LeRoy said the industrial supplier is using artificial intelligence to help customers understand its products. The Illinois-based company is using AI models and Databricks’ search-enhanced generation system to build its customer service tools, and plans to use Amazon’s chips in the background, LeRoy said.

Some technology analysts and business leaders say Amazon, which is not widely seen as a leader in AI innovation, needs to prove it can compete with Microsoft and Google. Part of Amazon's AI reboot involves its AI chips, Trainium and Inferentia, which are designed specifically for building and using AI models. The custom chips are more efficient than Nvidia's more general-purpose GPUs because they are designed for just one thing.

Amazon's pitch for its custom AI chips is lower costs. Dave Brown, vice president of AWS computing and network services, said customers can pay about 40% less than with other hardware.

“If customers can’t save any money and their existing solution is working well for them, then they’re not going to switch solutions,” Brown said. “So it’s really important to realize those cost savings.”

Brown declined to say how many Amazon customers use its custom chips instead of Nvidia’s GPUs.

Greg Rokita, vice president of technology at car shopping site Edmunds.com, said the site is using Databricks to build an AI tool to help customers determine what deals are available when buying electric vehicles. Rokita said the lower cost of building AI systems is a big advantage, especially because building your own AI is more popular than renting a vendor's private model.

NinjaTech AI, a startup that develops AI agents to perform tasks, said it spends about $250,000 a month on computing using Trainium chips, compared with $750,000 to $1.2 million using Nvidia’s GPUs. Other customers using Amazon’s custom AI chips include Anthropic, Airbnb, Pinterest and Snap.

But Amazon isn’t the only alternative to Nvidia. Nvidia’s old rival AMD has its own line of GPUs, Google makes its own in-house chips called Tensor Processing Units, and startups like Groq and Cerebras have developed their own single-purpose AI chips.

“For enterprises, the underlying technology is not important, it’s the value that the technology brings that’s important,” said Chirag Dekate, an analyst at Gartner, a market research and information technology consulting company. “If it’s Trainium, that’s great. If it’s a CPU, that’s great. If it’s a GPU, that’s great. But for the end user, it doesn’t matter.”

Reference Links

https://www.wsj.com/articles/amazon-databricks-strike-five-year-deal-around-ai-chips-a2a135ab?mod=tech_lead_pos2

END

????Semiconductor boutique public account recommendation????

▲Click on the business card above to follow

Focus on more original content in the semiconductor field

▲Click on the business card above to follow

Focus on the trends and developments of the global semiconductor industry

*Disclaimer: This article is originally written by the author. The content of the article is the author's personal opinion. Semiconductor Industry Observer reprints it only to convey a different point of view. It does not mean that Semiconductor Industry Observer agrees or supports this point of view. If you have any objections, please contact Semiconductor Industry Observer.

Today is the 3917th content shared by "Semiconductor Industry Observer" for you, welcome to follow.