Поиск по этому блогу

Search1

123

понедельник, 8 декабря 2025 г.

Trainium. Graviton. Madness. AWS.

There was something unique about AWS re:Invent 2025.‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  ‌  
  -  
AIM-logo-black
AIM-logo-black

Trainium. Graviton. Madness. AWS.

THE BELAMY

Weekly Newsletter of AIM

Monday, Dec 08, 2025 | By Mohit Pandey

Now, subscribe to our Digital & Print Editions >


There was something unique about AWS re:Invent 2025. 

It arrived at a time when OpenAI, Google, Microsoft and Anthropic were all sprinting towards model supremacy. Yet, AWS remained unfazed and focused on the unglamorous but essential part of AI: the infrastructure, the chips, the agent platforms and how enterprises will actually run these systems when hype gives way to scale. 

Enterprise AI is finally becoming real, and AWS wants to be the place where it lives and breathes.

CEO Matt Garman's keynote set the tone. He talked about giving customers speed, power, and stable foundations for training, inference and massive agent workflows

It also felt like the strongest signal yet that AWS is not trying to win the model race. Instead, it is building the rails on which everyone else will run.

Weekly-Infographics-14
Weekly-Infographics-14

Garman reminded the crowd that the NVIDIA partnership is older than the generative AI wave and that the two companies have collaborated on systems together for more than 15 years. "Nothing's too small for us to really work together to make sure that we have the most reliable performance," he said. 

He added that NVIDIA trains its largest models on AWS and called it "a testament to working together." The message was clear. AWS sees itself as the most stable place to build and train frontier systems, whether the customer is OpenAI, Humain, or anyone else pushing boundaries.

 

AIM Network Deep Dive >>

Adding another India-focused update, Amazon stated that it will invest $12.7 billion in local cloud and AI infrastructure and bring AI tools to 15 million small businesses while supporting AI education for four million students by 2030. The company added that the goal is to expand access, productivity and digital inclusion in line with India's AI Mission. 

video_preview_d80cd7f1ceea4583f372d0b494533dda.jpg
video_preview_d80cd7f1ceea4583f372d0b494533dda.jpg

AI Factories , Trainium3, Bedrock & Nova

Garman also moved into one of the biggest shifts in AWS strategy. AI factories are the cloud giant's attempt to bring hyperscale training inside customer walls

Many governments and enterprises have racks of data centre capacity but not the talent to stitch together giant clusters. "Why can't we help more customers—the ones who really need this large-scale infrastructure, see what our expertise, our services, are understanding?" he said. 

These AI factories allow AWS to drop its full stack inside a customer environment while meeting sovereignty rules, essentially giving the feeling of owned infrastructure without the pain of building it.

Then came silicon. AWS has quietly shipped more than one million Trainium chips. Moreover, at re:Invent, it unveiled the Trainium3 UltraServers and previewed Trainium4

According to Garman, Trainium4 delivers "over 6x the FP, 4x performance, 4x more memory family and 2x more high bandwidth memory capacity.

He also revealed that it doubles power efficiency. The point was not just raw compute. Garman said inference loads now resemble training. "There's not going to be an experienced application of a system built that doesn't rely on inference." 

He wants Trainium to become the base for training, low-latency inference and giant agent systems that never sleep.

Alongside all this came AWS's biggest CPU upgrade yet. Graviton5 will power the new M9g EC2 instances

AWS said it delivers up to 25% higher performance than the previous generation while improving energy efficiency. It uses 192 cores, a 5x larger L3 cache, faster memory speeds and 3nm technology. 

Bedrock continued its march toward becoming the world's largest neutral model hub. It now serves more than one lakh customers, and will add 18 new open-weight models across Google, MiniMax, Mistral, NVIDIA and OpenAI gpt-oss. 

Garman said customers are running many models at once. "This mix and match is going to be normal." AWS also refreshed its Nova line

  • Nova Light cuts costs on reasoning tasks. 
  • Nova Pro handles heavier reasoning across video and documents. 
  • Nova Sonic adds multilingual speech-to-speech. 

The unified Nova multimodal model accepts text, images, video and speech inputs for teams that want a single system that can "output different forms of text and imagery" without maintaining multiple workflows.

 

[AIM Exclusive at AWS re:Invent 2025]

AWS positions India as a priority market for regulated, multi-model and sovereign deployments across IT, BFSI, GCCs and mobility. AIM Network was reporting live from Las Vegas with exclusive insights from industry experts.

video_preview_7dfef47d943aff88548e30157db9ee8a.jpg
video_preview_7dfef47d943aff88548e30157db9ee8a.jpg

Russia is Becoming a Centre of Technology Gravity

The 10th anniversary of Russia's flagship AI conference, AI Journey, was never going to be a modest affair. In Moscow, the country's largest bank, Sber, now a fully-fledged technology group, used the event to parade an ecosystem that many in the West had assumed would be impossible under sanctions: LLMs trained entirely on domestic data, industrial-grade robotics and a new generation of "intelligent" devices built almost entirely on a Russian stack. And that's only scratching the surface. Click here to find out.

 

No More Babysitting

All this momentum set the scene for one of AWS' boldest announcements. 

Nova Forge will let companies build their own frontier model using Nova checkpoints and their own data. Many enterprises now want customisation that goes far beyond surface-level tweaks. "Why not make that possible? Why can't that be true?" Garman said.

Nova Forge lets customers blend internal data with Amazon curated sets and deploy the model on Bedrock with full guardrails. It is meant for companies that want models that behave like internal experts instead of generic assistants.

Then come the AI agents, which Garman said the industry is entering a time "where there were literally billions of agents working together." AWS wants to provide the full stack for building and governing them. 

Garman said teams were still treating agents like basic assistants when they should be self-directed workers who operate across full workflows. "I don't have to overwork, I don't have to babysit," he said.

Bedrock Agent Core now provides memory, identity, gateway, policy and evaluation layers. Policy lets customers write rules in simple language. 

AWS unveiled Kiro Powers and expanded its Frontier Agents for engineering and operations. In an exclusive conversation with AIM, Amit Patel, who leads engineering for Kiro, said 2025 has already been full of discovery and rapid change

Joking about why teams need these systems, he said, "These things happen at 2 o'clock in the morning." He explained that the DevOps agent is built to find incidents, analyse causes and fix issues before humans are paged. He also expects strong adoption in India. 

"It's going to be very interesting for India because we have such a big tech community," he said.

AWS walked into re:Invent wanting to show it can move as fast as any model lab while solving harder problems than any of them. AI scale will not be won by the biggest model. It will be won by the platform that helps everyone else build theirs.


AIM, in collaboration with Snowflake, is excited to present an inspiring and future-focused webinar, 'AI Leadership & Innovation: Are You Ready for the Next Tech Wave?'

Snowflake-Zoom-Creative-copy-scaled
Snowflake-Zoom-Creative-copy-scaled

The thought-provoking conversation will feature two trailblazing technology leaders, Sowmya V Kumaran, director of engineering and AI infrastructure management at Cisco, and Kanika Kapoor, senior VP of data management and analytics leader at NatWest.

Don't miss this opportunity to learn from industry pioneers and position yourself for the next wave of innovation. Register now.


Now, subscribe to our Digital & Print Editions >

For Brand collaborations, reply to this email or write to info@aim.media

You received this email because you signed up to the updates from AIM. Click here to unsubscribe if you do not want to receive emails from us.

  -  

Комментариев нет:

Отправить комментарий

Примечание. Отправлять комментарии могут только участники этого блога.