AWS pulled out all the stops again at re:Invent 2024. The annual event in Las Vegas revolved entirely around artificial intelligence (AI), and the cloud giant made that clear.
AWS aims to show at re:Invent 2024 not only that it is ready for the AI future, but that it is shaping that future itself. The conference brings together more than 60,000 visitors, customers and partners daily, all of whom will be treated to hundreds of technical sessions as well as innovative announcements.
New tools, infrastructure and a future vision of genAI that should help companies turn the technology into concrete results… AWS clearly wants to be the number one cloud company.
From three to four building blocks
Matt Garman’s keynote, on his first re:Invent as CEO, offered a grand look at the future of the cloud. “There has never been a more exciting time for businesses going through this transformation, thanks to all the technology that players like AWS are bringing to market,” Garman stated. More than 60,000 attendees followed the event live, while another 400,000 joined online to watch the announcements. AWS painted itself as an essential partner for companies looking to keep innovating.
AWS presented its vision for a future in which AI is at the core of its cloud infrastructure. This vision is built around four building blocks: compute, storage, databases and a new pillar of inference, specifically focused on implementing AI models.
“My view is that generative AI inference will become a crucial building block for any application. In fact, I think generative AI has the potential to transform every industry, every business, every workflow and every user experience. Just look at what is already happening,” Garman clarified.
Generative AI inference is becoming a crucial building block for any application.
Matt Garman, CEO AWS
Nevertheless, the three classic building blocks remain essential. Compute, storage and databases form the foundation of the AWS cloud infrastructure. Inference ensures that AI models provide correct answers within a context. “Inference is our logical next step to make AI not only scalable, but also usable in day-to-day operations,” Garman said.
Everything in the service of AI
AWS made it clear during re:Invent that it wants AI to become the core of its services. One of the most notable announcements were Amazon’ s Nova tools. Those are four new text models and photo and video generation models. In addition, Amazon Q Developer and Q Business AI agents also received updates. “With Q, we don’t just provide help, we accelerate innovation,” Garman emphasized.
read also
Amazon Bedrock gets new features
Amazon Bedrock also received a major update. Bedrock provides access to genAI models, including those from AWS itself and from third-party vendors such as Anthropic. New features such as model distillation make it easier to build (or have built) smaller, more efficient AI models, while reasoning checks improve the reliability of AI output.
Multi-agent collaboration would be a powerful feature to automate complicated workflows. Moody’s illustrated this by showing how it made formatting financial reports from in an hour instead of a week.
Compute: from Graviton to genAI
Compute remains one of the most important parts of the AWS strategy. The company claimed that the previously launched Graviton 4 processor is 40 percent faster and performs 30 percent better than Graviton 3. Garman was referring to Pinterest, which realized a 47 percent savings in computing costs thanks to the switch to Graviton. “Ninety of the top 1,000 EC2 customers are already running on Graviton,” he added.
For companies training AI models, AWS is introducing the Trainium 2 Ultraservers. These are training chips. With 83 petaflops, these servers are said to offer unprecedented efficiency. Garman also announced Trainium 3, the next generation of AI training chips that will be available in 2025 and should be 40 percent more energy efficient.
Faster storage, smarter databases
Storage and databases also got plenty of attention. Amazon S3 Tables promise up to three times faster query performance and up to 10 times higher transactions per second for Apache Iceberg, perfect for large-scale data analytics. S3 Tables is only the second extension of AWS’ S3 bucket format, following the introduction of S3 Express last year.
For companies bogged down by unstructured data, AWS introduced S3 Metadata that automatically updates metadata to make managing datasets and workflows more efficient.
The innovation for databases was Amazon Aurora DSQL, an SQL database that promises unlimited scalability and requires no infrastructure management. DynamoDB now supports multi-region. That speeds up the process of managing databases in different regions without getting into trouble with commits.
Efficiency and sustainability in data centers
AWS also came up with improvements in its infrastructure. Innovations in cooling and power management have increased data center efficiency by 12 percent. These improvements reduce outages and increase the capacity to run heavy workloads, such as genAI.
read also
Re:Invent 2024: AWS focuses entirely on AI
Bombastic atmosphere
Re:Invent continues to be an experience that completely takes over Las Vegas, or at least the hotels in which they take place. Gigantic halls filled with thousands of visitors, networking events in luxurious hotels and a jam-packed program of keynotes, sessions and workshops.
The event not only provided an inside look at different companies and their technologies, but more importantly, it was inspiring. Visitors came not only to hear the latest innovations, but also to get insights from big names in the industry (Apple’s Benoît Dupin made an appearance during a keynote!) and to learn from partners and customers in the Expo.
AWS is showing itself at re:Invent as a technology giant that continues to reinvent itself. From compute to storage, databases and inference, the announcements clearly show AWS’ ambition to integrate AI into all their applications.
The focus comes as no surprise. Every technology company today shouts that it is an AI company, and AWS is no exception. However, the cloud giant has some catching up to do: AWS jumped on the AI train later than major competitor Microsoft Azure, and needs to make it clear that its own cloud offerings are also ideal for those who want to snack on the AI pie.