Presented by Nvidia
Nvidia made a huge splash at its annual GTC 2024 conference with major news and announcements about its next-generation AI technologies. VentureBeat was there to cover it in full force, with backstage access to some of the key companies and industry leaders at the event. Throughout the course of four days, the VB editorial team streamed 21 interviews live from the show floor, all made possible through Nvidia’s partnership with VentureBeat Lab (VB Lab).
But in case you missed it, here are some highlights from the GTC 2024 interview series (you can catch up on all of VB’s video coverage here).
The future of AI factories
In a discussion with VentureBeat senior writer Sharon Goldman, representatives from Nvidia, Deloitte, and Sustainable Metal Cloud talked about the evolution of GPU-based data centers — what Nvidia CEO Jensen Huang likes to call AI factories. Rohit Tandon, general manager of AI at Deloitte, echoed Huang’s comments about how GPU power isn’t just about the chip, but the entire infrastructure.
“It’s like you have the best engine, but you have to build the car around it and put the driver in the driver’s seat, somebody who knows how to drive that car and the ways it can be pushed. If you add it together, that’s the vision of the gen AI, AI factory,” he explained. “You have the infrastructure pieces, all the enablement, networking, software, optimization, and you have the capability and skills to be able to drive that machine the way it’s supposed to be driven.”
More cost-effective AI models
VentureBeat lead writer Dean Takahashi sat down with Dan Diasio, the Global AI consulting leader from Ernst and Young (EY), about how the accounting firm is helping to build AI-powered businesses. EY has come up with a framework for its clients called MIN, the Minimum Intelligence Necessary to power an AI system. He said not every company needs to use costly frontier AI models, which he likened to buying a multi-million-dollar Formula 1 car.
“For many organizations, they don’t need an F1 car to go on a family vacation or get to work every day,” said Diasio, continuing his metaphor. “We find that organizations are finding many smaller models that they might be able to fine-tune with their expertise to get to the same sorts of results as a frontier model.”
The building blocks of gen AI
Microsoft has had an ongoing partnership with OpenAI since 2019, creating multiple generations of supercomputers that helped prepare the company for the massive launch of ChatGPT. That’s why general manager of Azure generative AI and HPC Nidhi Chappell can safely say that Microsoft was an instrumental part in creating the gen AI market. The company also has a deep partnership with Nvidia through the Azure OpenAI service, which acts as “building blocks” that clients can use to deploy AI at any scale.
“Every time I’m in a customer meeting, I’m just so impressed by how far they’ve taken their innovation based on these building blocks,” said Chappell. “They’re using language models to decide what car parts to build. They’re improving the efficiency of their employees. They’re improving chatbot interactions. The domains in which this has been applied are just phenomenal.”
Taking AI into production
It’s a good time for Run:ai, the creator of a middleware solution that allows organizations to efficiently access compute power when building their AI models. Companies like Adobe built their AI infrastructure with Run:ai, which helped them deploy models to production at a faster rate. That’s one common trend that Run:ai CEO and co-founder Omri Geller noticed from his meetings during the show, with more and more companies focusing on building out that infrastructure layer.
“They’re thinking about how to build their AI infrastructure right, so that they will be able to deploy models in production quickly and efficiently,” said Geller. “That’s the main focus for all the conversations that I’ve been in, whether it’s with end customers or partners of ours who, together with Run:ai, deploy those production clusters for organizations.”
Simplifying AI adoption
Dell Technologies knows how overwhelming it can be for enterprises looking to build with AI, so it sought to create an “easy button” that can help them no matter where they are in their AI journey. This resulted in the GTC announcement of the Dell AI Factory with Nvidia, the industry’s first end-to-end AI enterprise solution meant to simplify and accelerate AI adoption. With the Dell AI Factory, Dell SVP of product marketing, Varun Chhabra said that organizations can create a custom set of solutions based on a company’s specific needs.
“The excitement continues to be off the charts. This year is the year of enterprise AI adoption,” said Chhabra. “We’ve seen a lot of adoption from CSPs, cloud providers, GPU as-a-service providers. But where we think this is going this year is going to be enterprise adoption. There’s a lot of POCs that have been happening within enterprises.”
Using AI on the far edge
Thanks to its over $2 billion of investments in AI and a multifaceted partnership with Nvidia, Lenovo is now the third-largest AI infrastructure company in the world. While its vast portfolio spans workstations, autonomous driving solutions, data centers and more, one area that the company is excited about is getting AI to where data is created via its ThinkEdge product line. Some real-world examples include automatic self-checkout stations at Kroger stores and working with the city of Barcelona to install thousands of kiosks that help detect car crashes.
“These are use cases not just in a factory, but in retail, all the way to working with the largest fast-food chains in the world to automate their drive-throughs, keep the fries warm, and food safety,” said Kirk Skaugen, EVP and president of infrastructure solutions group at Lenovo. “It’s been a crazy year at the edge.”
Catch up with all the VB interviews at GTC 24 right here.
VB Lab Insights content is created in collaboration with a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact
