In an unmarked office constructing in Austin, Texas, two small rooms contain a handful of Amazon employees designing two forms of microchips for training and accelerating generative AI. These custom chips, Inferentia and Trainium, offer AWS customers an alternative choice to training their large language models on Nvidia GPUs, which have been getting difficult and expensive to acquire.
“Your entire world would really like more chips for doing generative AI, whether that is GPUs or whether that is Amazon’s own chips that we’re designing,” Amazon Web Services CEO Adam Selipsky told CNBC in an interview in June. “I believe that we’re in a greater position than anybody else on Earth to provide the capability that our customers collectively are going to want.”
Yet others have acted faster, and invested more, to capture business from the generative AI boom. When OpenAI launched ChatGPT in November, Microsoft gained widespread attention for hosting the viral chatbot, and investing a reported $13 billion in OpenAI. It was quick so as to add the generative AI models to its own products, incorporating them into Bing in February.
That very same month, Google launched its own large language model, Bard, followed by a $300 million investment in OpenAI rival Anthropic.
It wasn’t until April that Amazon announced its circle of relatives of huge language models, called Titan, together with a service called Bedrock to assist developers enhance software using generative AI.
“Amazon isn’t used to chasing markets. Amazon is used to creating markets. And I believe for the primary time in a protracted time, they’re finding themselves on the back foot they usually are working to play catch up,” said Chirag Dekate, VP analyst at Gartner.
Meta also recently released its own LLM, Llama 2. The open-source ChatGPT rival is now available for people to check on Microsoft’s Azure public cloud.
Chips as ‘true differentiation’
In the long term, Dekate said, Amazon’s custom silicon could give it an edge in generative AI.
“I believe the true differentiation is the technical capabilities that they are bringing to bear,” he said. “Because guess what? Microsoft doesn’t have Trainium or Inferentia,” he said.
AWS quietly began production of custom silicon back in 2013 with a bit of specialised hardware called Nitro. It’s now the highest-volume AWS chip. Amazon told CNBC there may be no less than one in every AWS server, with a complete of greater than 20 million in use.
AWS began production of custom silicon back in 2013 with this piece of specialised hardware called Nitro. Amazon told CNBC in August that Nitro is now the best volume AWS chip, with no less than one in every AWS server and a complete of greater than 20 million in use.
Courtesy Amazon
In 2015, Amazon bought Israeli chip startup Annapurna Labs. Then in 2018, Amazon launched its Arm-based server chip, Graviton, a rival to x86 CPUs from giants like AMD and Intel.
“Probably high single-digit to perhaps 10% of total server sales are Arm, and a great chunk of those are going to be Amazon. So on the CPU side, they’ve done quite well,” said Stacy Rasgon, senior analyst at Bernstein Research.
Also in 2018, Amazon launched its AI-focused chips. That got here two years after Google announced its first Tensor Processor Unit, or TPU. Microsoft has yet to announce the Athena AI chip it has been working on, reportedly in partnership with AMD.
CNBC got a behind-the-scenes tour of Amazon’s chip lab in Austin, Texas, where Trainium and Inferentia are developed and tested. VP of product Matt Wood explained what each chips are for.
“Machine learning breaks down into these two different stages. So that you train the machine learning models and then you definitely run inference against those trained models,” Wood said. “Trainium provides about 50% improvement by way of price performance relative to another way of coaching machine learning models on AWS.”
Trainium first got here in the marketplace in 2021, following the 2019 release of Inferentia, which is now on its second generation.
Inferentia allows customers “to deliver very, very low-cost, high-throughput, low-latency, machine learning inference, which is all of the predictions of if you type in a prompt into your generative AI model, that is where all that gets processed to provide you the response, ” Wood said.
For now, nevertheless, Nvidia’s GPUs are still king in the case of training models. In July, AWS launched recent AI acceleration hardware powered by Nvidia H100s.
“Nvidia chips have a large software ecosystem that is been built up around them during the last like 15 years that no one else has,” Rasgon said. “The massive winner from AI immediately is Nvidia.”
Amazon’s custom chips, from left to right, Inferentia, Trainium and Graviton are shown at Amazon’s Seattle headquarters on July 13, 2023.
Joseph Huerta
Leveraging cloud dominance
AWS’ cloud dominance, nevertheless, is an enormous differentiator for Amazon.
“Amazon doesn’t have to win headlines. Amazon already has a very strong cloud install base. All they should do is to work out enable their existing customers to expand into value creation motions using generative AI,” Dekate said.
When selecting between Amazon, Google, and Microsoft for generative AI, there are tens of millions of AWS customers who could also be drawn to Amazon because they’re already aware of it, running other applications and storing their data there.
“It’s a matter of velocity. How quickly can these corporations move to develop these generative AI applications is driven by starting first on the info they’ve in AWS and using compute and machine learning tools that we offer,” explained Mai-Lan Tomsen Bukovec, VP of technology at AWS.
AWS is the world’s biggest cloud computing provider, with 40% of the market share in 2022, in accordance with technology industry researcher Gartner. Although operating income has been down year-over-year for 3 quarters in a row, AWS still accounted for 70% of Amazon’s overall $7.7 billion operating profit within the second quarter. AWS’ operating margins have historically been far wider than those at Google Cloud.
AWS also has a growing portfolio of developer tools focused on generative AI.
“Let’s rewind the clock even before ChatGPT. It is not like after that happened, suddenly we hurried and got here up with a plan because you possibly can’t engineer a chip in that quick a time, let alone you possibly can’t construct a Bedrock service in a matter of two to three months,” said Swami Sivasubramanian, AWS’ VP of database, analytics and machine learning.
Bedrock gives AWS customers access to large language models made by Anthropic, Stability AI, AI21 Labs and Amazon’s own Titan.
“We do not believe that one model goes to rule the world, and we would like our customers to have the state-of-the-art models from multiple providers because they’re going to pick the appropriate tool for the appropriate job,” Sivasubramanian said.
An Amazon worker works on custom AI chips, in a jacket branded with AWS’ chip Inferentia, on the AWS chip lab in Austin, Texas, on July 25, 2023.
Katie Tarasov
One among Amazon’s newest AI offerings is AWS HealthScribe, a service unveiled in July to assist doctors draft patient visit summaries using generative AI. Amazon also has SageMaker, a machine learning hub that gives algorithms, models and more.
One other big tool is coding companion CodeWhisperer, which Amazon said has enabled developers to complete tasks 57% faster on average. Last 12 months, Microsoft also reported productivity boosts from its coding companion, GitHub Copilot.
In June, AWS announced a $100 million generative AI innovation “center.”
“We have now so many shoppers who’re saying, ‘I would like to do generative AI,’ but they do not necessarily know what which means for them within the context of their very own businesses. And so we’re going to herald solutions architects and engineers and strategists and data scientists to work with them one on one,” AWS CEO Selipsky said.
Although to this point AWS has focused largely on tools as an alternative of constructing a competitor to ChatGPT, a recently leaked internal email shows Amazon CEO Andy Jassy is directly overseeing a recent central team constructing out expansive large language models, too.
Within the second-quarter earnings call, Jassy said a “very significant amount” of AWS business is now driven by AI and greater than 20 machine learning services it offers. Some examples of shoppers include Philips, 3M, Old Mutual and HSBC.
The explosive growth in AI has include a flurry of security concerns from corporations fearful that employees are putting proprietary information into the training data utilized by public large language models.
“I can not let you know what number of Fortune 500 corporations I’ve talked to who’ve banned ChatGPT. So with our approach to generative AI and our Bedrock service, anything you do, any model you employ through Bedrock will probably be in your personal isolated virtual private cloud environment. It’ll be encrypted, it’ll have the identical AWS access controls,” Selipsky said.
For now, Amazon is simply accelerating its push into generative AI, telling CNBC that “over 100,000” customers are using machine learning on AWS today. Although that is a small percentage of AWS’s tens of millions of shoppers, analysts say that would change.
“What we aren’t seeing is enterprises saying, ‘Oh, wait a minute, Microsoft is so ahead in generative AI, let’s just exit and let’s switch our infrastructure strategies, migrate all the things to Microsoft.’ Dekate said. “If you happen to’re already an Amazon customer, chances are high you are likely going to explore Amazon ecosystems quite extensively.”
— CNBC’s Jordan Novet contributed to this report.
CORRECTION: This text has been updated to reflect Inferentia because the chip used for machine learning inference.