Amazon to back OpenAI with $50B for memory AI

The multiyear deal includes exclusive AWS cloud distribution, expanded compute capacity and development of long-term memory AI systems for enterprises.

 


Una inversión de 50.000 millones de dólares y la promesa de una IA con memoria a largo plazo.

OpenAI y Amazon apuestan por transformar cómo las organizaciones construyen y usan inteligencia artificial.


At the heart of thousands of companies lies a common frustration: having to explain everything to technology all over again every time a task begins. Today, that dynamic could start to change.
OpenAI and Amazon have announced a multiyear strategic partnership to accelerate innovation in artificial intelligence for enterprises, startups, and consumers worldwide, with a total committed investment of $50 billion.

The agreement includes an initial investment of $15 billion from Amazon, followed by an additional $35 billion over the coming months, contingent upon certain conditions being met.
The partnership also involves the development of long-term memory AI capabilities, expected to be available in the coming months.

Memory AI.

Both companies are working on a system called the “Stateful Runtime Environment,” powered by OpenAI models and accessible through Amazon Bedrock.
This environment will allow AI to remember previous projects, learn how each company operates, and resume conversations or tasks exactly where they left off, without needing to reset context.
It will also be able to move seamlessly between different tools and databases as part of a continuous workflow.

These environments will be trained to operate optimally on the Amazon Web Services (AWS) infrastructure and integrated with Amazon Bedrock AgentCore and other company services, ensuring AI applications and agents function cohesively with the rest of an organization’s cloud-based systems.

Infrastructure and exclusivity.

As part of the agreement, AWS will become the exclusive cloud distribution provider for OpenAI Frontier, OpenAI’s most advanced enterprise platform.
Frontier is designed for organizations to build and manage teams of AI agents operating on real systems, with governance and security embedded, without needing to manage their own infrastructure. This aims to facilitate the integration of advanced AI into existing processes quickly and at a global scale.

The partnership also expands the existing $38 billion agreement between the two companies, adding an additional $100 billion over the next eight years.
Within this framework, OpenAI will consume approximately 2 gigawatts of Trainium chip capacity, Amazon’s AI-designed processors, to support high-demand products, including the Stateful Runtime Environment and Frontier.

Chips and expansion.

The commitment includes current Trainium3 chips and the next-generation Trainium4, expected in 2027, offering significant improvements in processing speed and memory capacity to support increasingly complex AI systems.
This ensures OpenAI has long-term compute capacity while reducing operational costs, allowing companies to access AI on-demand without managing their own infrastructure.

The collaboration also includes the development of custom AI models for Amazon’s internal teams, which can be integrated into products and agents that interact directly with users.
These models will complement those Amazon already provides to developers, such as the Nova family, expanding the available ecosystem of tools.

Regarding the partnership, Sam Altman, cofounder and CEO of OpenAI, noted:

“OpenAI and Amazon share the conviction that AI should be realized in practical ways and truly useful for people. Combining OpenAI’s models with Amazon’s infrastructure and global reach allows us to deliver powerful AI to enterprises and users at real scale.”

Andy Jassy, president and CEO of Amazon, stated:

“We have numerous developers and companies interested in running services powered by OpenAI models on AWS, and our unique collaboration to deliver stateful runtime environments will transform what’s possible for those building AI applications and agents. We remain impressed with what OpenAI is building, and we are excited not only by their decision to leverage our custom AI silicon (Trainium) but also by the opportunity to invest in the company and strengthen this long-term partnership.”

Share:
Hosting Web
Most Read