About 11 years ago, with our very talented Annapurna team and informed by the unusual scale and insight we had in operating the largest cloud infrastructure, we decided to design and build our own CPU chip.
This CPU chip is now known as Graviton, and is well-loved by our AWS customers. About 98% of our top 1,000 EC2 customers use Graviton expansively. Over 120,000 customers are using Graviton. Meta just committed to tens of millions of Graviton cores for their agentic AI efforts. Uber and Snowflake are deploying Graviton as well.
Today, we released Graviton5 into General Availability.
The reason customers are so excited about Graviton is that it offers about 30-40% better price-performance than comparable instances. This is a big deal at any scale.
But, when you layer on top of how much CPU customers normally use with the fact that AI’s growth is driving explosive CPU expansion given that post-training, reinforcement learning, and agentic actions use CPU, Graviton becomes even more compelling.
Excited for how Graviton5 will help customers. https://t.co/I5IjZJ2OCh
Almost every customer conversation right now comes back to the same thing: they want AI that can take on more complex tasks autonomously, not just answer questions.
@AnthropicAI's Claude Fable 5, its first 5th-generation model, is built for that and is now available on Amazon Bedrock and Claude Platform on @awscloud: https://t.co/k3GCSOF6Uj
We just shipped stablecoins on @CashApp.
Everything runs from your existing USD balance - no separate wallet, no managing multiple chains, no extra setup, and importantly no fees.
Send and receive USDC on SOL, ETH, POL, and ARB.
The most seamless integration in the world imo.
i like how @kevinxu just sent everyone into poverty with this turd of a penny stock and sure enough he'll post soon telling everyone how he's worth $11m just bc. Truly one of the worst guys on this entire app.
Apple Intelligence is about to make some of our accessibility features even more powerful: from stronger descriptions in VoiceOver and Magnifier to a more intuitive Voice Control with natural language and an even more adaptive Accessibility Reader! #GAAD https://t.co/v5H0YhbUZu
The future of AI won’t just be shaped by bigger models. It’ll be shaped by the researchers building better systems underneath them.
Build on Trainium is @awscloud's $110M investment helping universities push AI research forward on Trainium, Neuron, and large-scale AWS infrastructure.
Excited to see what this next generation builds. https://t.co/jU0nP0AjYv
We were at Annapurna Labs when AWS engineers unboxed the very first Trainium3 chip. What followed was a rollercoaster of real-time problem-solving to prove the chip could power next-generation AI.
Watch the full bring-up video: https://t.co/huwzIRMZxi
We're the first cloud provider to offer @AnthropicAI 's native Claude Platform directly through customer accounts. Teams get access to the full platform, from Managed Agents to code execution, using the @awscloud credentials, billing, and audit controls they already rely on.
No new vendor accounts, no new systems. https://t.co/I9w1ypzhXH
WhatsApp Pay is live in India, Brazil, and Singapore only — sub-scale even in those markets. The competency gap isn't technology; it's regulatory navigation and bank partnership. WeChat Pay has 1.3B active users at scale. Stripe + WhatsApp would be a hyperscale fintech overnight in 50+ markets.
It's existential for @Meta to expand @WhatsApp via AI and convert 3B users into a topline alongside ads. The infra spend is there; the deployment layer isn't. They're lagging on execution and the market is pricing the gap, not the capex.
WhatsApp Business is currently a thin CRM tool — businesses send template messages and pay per conversation. The right product is a fully Llama-powered agent deployed for every SMB: handles customer service, scheduling, sales, returns, recommendations, all in natural conversation, with handoff to humans when needed.