AWS re:Invent 2024 | 7 takeaways after drinking from the firehose
Brought to you by Hitachi Vantara
AWS re:Invent 2024 is in the books with a barrage of news items and product launches, a few steps toward integrated suites of cloud services, Matt Garman’s debut as CEO, a bevy of customer takeaways and a lot of leftover questions.
Here’s a look at my big takeaways from re:Invent 2024.
Amazon SageMaker and Bedrock better together? If you have been following AWS closely for the last year, you’d get the impression that Amazon Bedrock was the lead genAI dog. What’s not to love about model choices, serverless infrastructure and the ability to experiment with large language models quickly?
Even though Amazon SageMaker was the workhorse for data science, machine learning and AI, it took a backseat in buzz. Amazon SageMaker and Bedrock were seen as separate products. AWS re:Invent 2024 changed that balance somewhat. The underlying message: Amazon SageMaker and Amazon Bedrock are better together. These two will increasingly co-mingle in the future.
AWS’ move to unify data,?analytics and AI under SageMaker was well received and Amazon SageMaker Lakehouse makes a ton of sense too. Bedrock rides shotgun as a curated entry to generative AI. Going forward, there will be clear connections and handoffs between the two. It’s also worth noting that many of AWS’ moves to make data management easier on the compute and storage layers all roll up into making SageMaker and Bedrock better.
In an analyst Q&A, AWS CEO Matt Garman said SageMaker and Bedrock play both ends of the AI spectrum for customers.
He said: “Customers are going to use a wide range of models, and they'll fine-tune some of the models in Bedrock, and they will build their own. The answer is evolving. You'll find it easier to build a model from scratch with your own proprietary set of data. There’s going to be a lot of customers who are continuing to do that on SageMaker. We're seeing no amount of slowing down of customers doing that.”
Commoditizing LLMs and pricey GPUs. One of the big news items at AWS re:Invent was the company’s launch of its Nova family of models. The company is replacing Titan with next-generation Nova models. When paired with the latest Trainium launch, it’s clear that AWS is doing what it always does—commoditize and provide choices.
For folks who think in sports terms and zero-sum outcomes, Nova could be seen to conflict with Anthropic and AWS’ third-party model strategy. However, that narrative is a bit of a nothingburger given AWS’ history of offering choices even if some of its efforts compete. AWS actually expanded its model choices with Luma AI, poolside and its Nova efforts.
On the GPU side, Garman noted that AWS is a big supporter of AMD and Nvidia. Trainium isn’t competition but offers a way for enterprises to balance price and performance. My take is that Trainium will be very interesting should large language models plateau a bit. At some stage, good enough genAI is going to be plenty.
Amazon Q Business gets a story. AWS launched Q Business a year ago and the effort has come a long way. With the addition of ISV partnerships, Q Index and the QuickSight integration, Q Business has more of a narrative.
To say, Q Business’ storyline has rallied is an understatement. Just a few months ago, AWS had trouble telling the story. Q Business will ultimately evolve into an engine to automate workflows and there are genAI centers of excellence being built to leverage Q Business.
Hunting dinosaurs--Net, VMware, mainframes. Q Developer, which started as CodeWhisperer, has had a much easier narrative. Q Developer returns are clear cut—it makes the software developer lifecycle more efficient. After all, developers are high-ticket labor. Q Developer at re:Invent rolled out features that can easily convert .Net conversions and VMware transformations.
The VMware effort is interesting given how many enterprises are looking to migrate to Nutanix and other offerings. It’s worth noting that enterprises can take years to migrate off of VMware so anything that speeds up that process will see strong demand.
Innovation vs. too many services. A common theme with AWS is that it has a service for everything and some of them compete and cause confusion. All you have to do is look at the AWS news blog on everything announced and you can find more than a few overlaps. On the bright side, AWS’ practice to let product teams run independently spurs innovation.
But there is an argument that a little coordination would go a long way. Garman in a Q&A addressed the issue. "If you have 50 teams that all have to be in lockstep to deliver something you're going to move slow," said Garman. "Our strategy has been to let those teams invent and move fast. I appreciate that approach introduces some complexity for customers and we're moving to cover that."
He noted that SageMaker Studio is an example of an effort to bring those building blocks together seamlessly. "SageMaker is an elegant example of making it easier to operate with a whole set of tools on common data sets," said Garman. "We can do that because we have all these core components behind the scenes. We will keep innovating."
There’s a lot of runway for AWS to tie its building blocks and services together and customers would most likely cheer the effort.
Matt Garman’s debut and Andy Jassy’s guest spot. Lost in the re:Invent 2024 mix was that the keynote was essentially Garman’s debut as AWS CEO. Honestly, I forgot this debut storyline after a few minutes. Garman was comfortable in his keynote, fielded questions with ease and seemed like a CEO who has been leading the company for years. The transition was seamless.
领英推荐
There was also bridge to AWS’ past as Amazon CEO Andy Jassy had a big segment on day one. He outlined Amazon’s use of AI in an interesting talk. Jassy said:
“With the explosion of generative AI, there is a ton of innovation, but what we're trying to do is solve problems for you. We think of it as practical AI. The most success that we've seen from companies everywhere in the world is in cost avoidance and productivity. You see lots of companies having gains there, but you also are starting to see completely reimagined and reinvented customer experiences.”
We have a real chance with Alexa to be the leader. We are in the process right now of rearchitecting the brains of Alexa with multiple foundation models. It’s going to not only help Alexa answer your questions even better, but it's going to do what very few generative AI applications do today, which is to understand and anticipate your needs and actually take action for you. You can expect to see this in the coming months.”
Multicloud: Something to watch in re:Invent 2025? AWS is thinking about multicloud because customers want it. AWS publicly saying multicloud could be construed as news. In a session with analysts, Rich McDonough, worldwide tech lead for AWS' hybrid and multicloud program, made a little history by even acknowledging multicloud.
McDonough said AWS is focusing multicloud efforts on customers with the most cloud maturity via partners and services. He said: "We're very much of the opinion that operating across multiple clouds needs to be the result of cloud maturity and a builder mindset. Our advice to all of our customers is still to get good at one cloud first before trying to put more clouds into your environment, understand the overhead cost and the cloud tax."
At AWS re:Invent 2024, there were a few sessions on multicloud. AWS offered a series of best practices, professional services and partners to help customers adopt multicloud architectures. Today AWS Marketplace is a venue for multicloud management applications to go along with AWS services like CloudWatch and Systems Manager.
My guess: Amazon Q (Business and Developer) will be aimed squarely at multicloud deployments in the future. Think Q CloudOps and FinOps in 2025. ?
About that AWS re:Invent firehose:
From our underwriter:
NOTEBOOK
???Salesforce posted a mixed third quarter and fourth quarter outlook as revenue was up 8% from a year ago. The company saw revenue growth decelerate sequentially across multiple categories.?However, executives were bullish on Agentforce uptake and said the AI layer is transforming internal operations.
???Google Deepmind launched Gencast, an AI model that provides faster more accurate weather forecasts 15 days in advance.
?? Pure Storage landed a hyperscale cloud account with a novel twist. The company licensed its software and storage designs to a cloud provider and not hardware. The deal was outlined in the company’s earnings report.
?? Jay Chaudhry, CEO of Zscaler, said the company's integration with competing cybersecurity platforms such as CrowdStrike, generative AI upsells and new executive additions will fuel growth in future quarters.?
?It's unclear who will be the new CEO of Intel, but one thing is certain--there will be a lot of work ahead. Here's a look at some of the big questions the new Intel leader will have to resolve.
????DeWayne Griffin, Chief Digital and Information Officer at Insight Global, said artificial intelligence has a big role in human resources, staffing and recruitment, but you will need a human in the loop to assess soft skills and cultural fit.
?? See Insights blog for earnings from UiPath, HPE and more breaking news.
INSIGHTS ARCHIVE
Like getting our weekly round-up??Subscribe to Constellation Insights?for consistent news and analysis on technology trends, earnings, enterprise updates and more.
Want to work with Constellation analysts??Our Sales team would love to get in touch!