AWS CEO Garmin pitches ‘billions of brokers’ as enterprise AI future
Amazon Internet Providers (AWS) chief government Matt Garmin projected a full agentic synthetic intelligence (AI) future for the cloud infrastructure provider’s clients throughout his keynote speech at its Re:Invent present in Las Vegas.
He advised attendees that freedom for purchasers to invent has been the constant mantra of AWS since its inception 20 years in the past, and that the true enterprise worth of synthetic intelligence can be unlocked with the elevated deployment of AI brokers.
Garmin mentioned AWS has been innovating at “all ranges of the expertise stack” throughout the previous three years, and that “billions of brokers” producing real-world worth is the provider’s imaginative and prescient.
He was joined onstage by executives from AWS clients Sony and Adobe.
John Kodera, chief digital officer at Sony, talked in regards to the Japanese worth of “Kando”, which means deep emotional engagement and awe as a enterprise purpose. He referred to Sony’s “knowledge ocean”, constructed on AWS’s machine studying platform Bedrock, which is used to construct generative synthetic intelligence (GenAI) purposes. Kodera additionally disclosed that Sony is a consumer of AWS Nova Forge, a service that allows clients to construct so-called “frontier fashions” utilizing the provider’s cluster of basis fashions, Nova, accessible on Amazon Bedrock. By a “frontier mannequin”, the business tends to imply AI techniques which can be large-scale and sometimes multi-modal, processing knowledge from photographs, video and audio in addition to textual content.
Shantanu Narayen, chair and CEO of Adobe, additionally appeared onstage as an AWS buyer, with creativity within the period of AI his theme. Adobe software program like Acrobat, Firefly and Studio use such AWS applied sciences as Amazon SageMaker and Bedrock, he mentioned.
AWS’s Garmin mentioned AI startups had been “flocking to AWS”, and cited as one AudioShake, a sound separation specialist supplier that may assist folks with listening to difficulties.
In opposition to technical debt
Included among the many product developments the provider introduced on the convention had been agentic capabilities in AWS Remodel, mentioned to allow speedy modernisation of any code or utility.
AWS Remodel can be utilized for “full-stack Home windows modernisation, to dump advanced, tedious modernisation work throughout the Home windows utility stack”.
“Now you can establish utility and database dependencies, and modernise them in an orchestrated manner by means of a centralised expertise,” the provider mentioned.
Air Canada, Experian, QAD, Teamfront, Thomson Reuters and Verisk are utilizing AWS Remodel to assist get rid of their technical debt, mentioned AWS.
The provider additionally introduced AWS AI Factories, mentioned to rework clients’ current infrastructure into high-performance AI environments.
These “factories” include devoted AWS AI infrastructure deployed in clients’ current datacentres. They’re mentioned to mix Nvidia accelerated computing platforms with AWS Trainium chips, high-speed, low-latency networking, and AWS AI providers.
Clients can use their current datacentre area, community connectivity and energy whereas AWS handles deployment and administration of the built-in infrastructure. They’ll allow clients to fulfill their knowledge sovereignty and regulatory necessities, with accelerated deployment timelines, mentioned AWS.
IDC analyst perspective
Neil Ward Dutton, an IDC analyst centered on AI, automation, knowledge and analytics, mentioned three bulletins on the primary day of the convention had been of word. “AWS AI Factories is one,” he mentioned. “These are AWS AI infrastructure and software program stacks, constructed and managed by AWS, however deployed in clients’ personal datacentres. These are, a minimum of initially, aimed on the largest clients with probably the most difficult safety or sovereignty necessities.”
The second, he mentioned, was Amazon Nova Forge. “That is the flexibility for purchasers to convey their very own coaching knowledge corpuses to part-trained Nova fashions, finishing mannequin coaching to create domain-specialised fashions that may be deployed to Bedrock,” mentioned Ward Dutton.
The third is the “coverage specification and evaluations in Bedrock AgentCore”, which is a bunch of agent-based AI instruments and merchandise introduced by AWS in July 2025. He mentioned the brand new components “specify verifiable operational insurance policies that constrain AI brokers’ entry to instruments, sources and knowledge”.
“The primary wave of enterprise GenAI exercise is primarily centered on accessing ‘AI at a distance’,” added Ward Dutton. “This solely takes you to date, and organisations at the moment are in the midst of discovering that out. What is going to get enterprises to true worth at scale just isn’t what received them so far.
“The subsequent part of enterprise AI exercise – whether it is to achieve success – has to embrace the absorption and integration of AI applied sciences into enterprise processes, workflows, techniques of engagement, knowledge and information, safety and privateness approaches, regulatory responses, and geopolitical and sovereignty considerations,” he mentioned.
