AWS’ custom chip strategy is showing results, and cutting into Nvidia’s AI dominance

[Daniel Nenni]

Key Points

• - Amazon Web Services is set to announce an update to its Graviton4 chip that includes 600 gigabits per second of network bandwidth.
• - Amazon is looking to reduce AI training costs and provide an alternative to Nvidia’s expensive graphics processing units, or GPUs.
• - Anthropic’s Claude Opus 4 AI model launched on AWS’ Trainium2 GPUs and Project Rainier is powered by over half a million of the chips.

AWS announces new CPU chip: Here’s what to know

Amazon Web Services is set to announce an update to its Graviton4 chip that includes 600 gigabits per second of network bandwidth, what the company calls the highest offering in the public cloud.

Ali Saidi, a distinguished engineer at AWS, likened the speed to a machine reading 100 music CDs a second.

Graviton4, a central processing unit, or CPU, is one of many chip products that come from Amazon’s Annapurna Labs in Austin, Texas. The chip is a win for the company’s custom strategy and putting it up against traditional semiconductor players like Intel and AMD.

But the real battle is with Nvidia in the artificial intelligence infrastructure space.

At AWS’s re:Invent 2024 conference last December, the company announced Project Rainier – an AI supercomputer built for startup Anthropic. AWS has put $8 billion into backing Anthropic.

AWS Senior Director for Customer and Product Engineering Gadi Hutt said Amazon is looking to reduce AI training costs and provide an alternative to Nvidia’s expensive graphics processing units, or GPUs.

Anthropic’s Claude Opus 4 AI model launched on Trainium2 GPUs, according to AWS, and Project Rainier is powered by over half a million of the chips – an order that would have traditionally gone to Nvidia.


Hutt said that while Nvidia’s Blackwell is a higher-performing chip than Trainium2, the AWS chip offers better cost performance.

“Trainium3 is coming up this year, and it’s doubling the performance of Trainium2, and it’s going to save energy by an additional 50%,” he said.

The demand for these chips is already outpacing supply, according to Rami Sinno, director of engineering at AWS’ Annapurna Labs.

“Our supply is very, very large, but every single service that we build has a customer attached to it,” he said.

With Graviton4′s upgrade on the horizon and Project Rainier’s Trainium chips, Amazon is demonstrating its broader ambition to control the entire AI infrastructure stack, from networking to training to inference.

And as more major AI models like Claude 4 prove they can train successfully on non-Nvidia hardware, the question isn’t whether AWS can compete with the chip giant — it’s how much market share it can take.

The release schedule for the Graviton4 update will be provided by the end of June, according to an AWS spokesperson.

https://www.cnbc.com/2025/06/17/aws-chips-nvidia-ai.html?

 

[hist78]

Key Points

• - Amazon Web Services is set to announce an update to its Graviton4 chip that includes 600 gigabits per second of network bandwidth.
• - Amazon is looking to reduce AI training costs and provide an alternative to Nvidia’s expensive graphics processing units, or GPUs.
• - Anthropic’s Claude Opus 4 AI model launched on AWS’ Trainium2 GPUs and Project Rainier is powered by over half a million of the chips.

AWS announces new CPU chip: Here’s what to know

Amazon Web Services is set to announce an update to its Graviton4 chip that includes 600 gigabits per second of network bandwidth, what the company calls the highest offering in the public cloud.

Ali Saidi, a distinguished engineer at AWS, likened the speed to a machine reading 100 music CDs a second.

Graviton4, a central processing unit, or CPU, is one of many chip products that come from Amazon’s Annapurna Labs in Austin, Texas. The chip is a win for the company’s custom strategy and putting it up against traditional semiconductor players like Intel and AMD.

But the real battle is with Nvidia in the artificial intelligence infrastructure space.

At AWS’s re:Invent 2024 conference last December, the company announced Project Rainier – an AI supercomputer built for startup Anthropic. AWS has put $8 billion into backing Anthropic.

AWS Senior Director for Customer and Product Engineering Gadi Hutt said Amazon is looking to reduce AI training costs and provide an alternative to Nvidia’s expensive graphics processing units, or GPUs.

Anthropic’s Claude Opus 4 AI model launched on Trainium2 GPUs, according to AWS, and Project Rainier is powered by over half a million of the chips – an order that would have traditionally gone to Nvidia.


Hutt said that while Nvidia’s Blackwell is a higher-performing chip than Trainium2, the AWS chip offers better cost performance.

“Trainium3 is coming up this year, and it’s doubling the performance of Trainium2, and it’s going to save energy by an additional 50%,” he said.

The demand for these chips is already outpacing supply, according to Rami Sinno, director of engineering at AWS’ Annapurna Labs.

“Our supply is very, very large, but every single service that we build has a customer attached to it,” he said.

With Graviton4′s upgrade on the horizon and Project Rainier’s Trainium chips, Amazon is demonstrating its broader ambition to control the entire AI infrastructure stack, from networking to training to inference.

And as more major AI models like Claude 4 prove they can train successfully on non-Nvidia hardware, the question isn’t whether AWS can compete with the chip giant — it’s how much market share it can take.

The release schedule for the Graviton4 update will be provided by the end of June, according to an AWS spokesperson.

https://www.cnbc.com/2025/06/17/aws-chips-nvidia-ai.html?

According to this CNBC report, over the past two years, 50% of the new AWS CPU capacity has been running on Amazon’s own Graviton 4 processors. This presents a serious challenge for both Intel and AMD in growing their future server processor businesses. It could be especially problematic for Intel as it looks for ways to prevent declining revenue.

 

[Daniel Nenni]

According to this CNBC report, over the past two years, 50% of new CPU capacity has been running on Amazon’s own Graviton 4 processors. This presents a serious challenge for both Intel and AMD in growing their future server processor businesses. It could be especially problematic for Intel as it looks for ways to prevent declining revenue.


View attachment 3297

Google and the other hyper scalars are doing the same but it is Nvidia they are going after since the cost of GPUs are skyrocketing. The Intel/AMD competition is keeping costs down but Nvidia is a monopoly.

 

[hist78]

Google and the other hyper scalars are doing the same but it is Nvidia they are going after since the cost of GPUs are skyrocketing. The Intel/AMD competition is keeping costs down but Nvidia is a monopoly.

Amazon’s Graviton processors are designed for servers and cloud computing, not specifically for AI. They compete directly with Intel’s and AMD’s server processorsrs.

 

[Xebec]

Google and the other hyper scalars are doing the same but it is Nvidia they are going after since the cost of GPUs are skyrocketing. The Intel/AMD competition is keeping costs down but Nvidia is a monopoly.

This seems analogous to oil supply.

Price it too low and you miss $$. Price it too high and people develop alternatives.

It will be interesting to see if Nvidia can differentiate it's market segments enough to keep the gravy train going long term.

 

[hist78]

This seems analogous to oil supply.

Price it too low and you miss $$. Price it too high and people develop alternatives.

It will be interesting to see if Nvidia can differentiate it's market segments enough to keep the gravy train going long term.

This is also a major challenge for both Intel and AMD, especially Intel. The company is facing onslaughts from all directions, and in many cases, it doesn’t even have a competitive product.

 

[kingmouf]

All is good and fine about creating alternatives to nVidia, but from our perspective (I work in a research institute), the real monopoly fear is not nVidia, it is AWS (or any other cloud provider) specific products. With nVidia, we can scale depending our varying economic conditions (stemming from different funding processes, each with different strings attached), from our workstations, our servers (with nVidia we can easily choose ANY server manufacturer, even make our systems from different components) in our data center to any cloud if need be. To use an AWS specific technology is fearsome. We can only use this cloud service. And in the long run, it is MUCH more expensive. MUCH MUCH more expensive, especially when we are in new projects that we stand on shaking ground and we cannot predict a budget. For experimenting with different technologies, AWS is very good. But when we establish what we need or we have already setup our workflow, then we cannot simply afford to go AWS, especially for long runs. We faced the same thing with FPGAs and F1 instances, although these were kinda "sponsored" and extremely competitively priced. In the end, when we acquired a few Xilinx data center cards, everything was cheaper.

 

[kevin01]

This seems analogous to oil supply.

Price it too low and you miss $$. Price it too high and people develop alternatives.

It will be interesting to see if Nvidia can differentiate it's market segments enough to keep the gravy train going long term.

I was thinking about this the other day:
In the era of 1980-2010, I would argue that semiconductor companies had a bigger impact than software companies, yet the semiconductor king of that era, $intc, was not more valuable than the software king, $msft. Even network gear provider $csco was the most valuable company in the world for a few days, but $intc never was.

In my opinion, the reason $intc never became the most valuable company is that the business of providing computing devices is not a monopoly. There were always at least three companies providing alternatives.

Why is $nvda, today's dominant computing provider, the most valuable company? I think it has to do with the sudden, significant, rise in AI computing demand over the last two to three years. $csco once experienced a similar surge, but $intc never did.

Can $nvda continue to command high margins coming with this sudden "monopoly"? I bet not.

 

[jmlobert]

It's only a direct loss to Intel because the chips are made by TSMC. If manufacturing was brought to Intel foundry, it would be a tolerable partial loss. The interesting part here is that Amazon (Annapurna Labs?) can design chips that can compete with Intel's, AMD's and NVIDIA's...

 

[siliconbruh999]

I was thinking about this the other day:
In the era of 1980-2010, I would argue that semiconductor companies had a bigger impact than software companies, yet the semiconductor king of that era, $intc, was not more valuable than the software king, $msft. Even network gear provider $csco was the most valuable company in the world for a few days, but $intc never was.

In my opinion, the reason $intc never became the most valuable company is that the business of providing computing devices is not a monopoly. There were always at least three companies providing alternatives.

Why is $nvda, today's dominant computing provider, the most valuable company? I think it has to do with the sudden, significant, rise in AI computing demand over the last two to three years. $csco once experienced a similar surge, but $intc never did.

Can $nvda continue to command high margins coming with this sudden "monopoly"? I bet not.

Nvidia can or cannot but TSMC can