There’s no doubt that enterprise leaders are excited about generative AI (genAI) and its transformative applications.

But the technology is still in its fledgling stages in the enterprise, with many IT decision makers unsure of how or where to begin and grappling with questions around customization, data privacy, cost, complexity and numerous other factors.

“A lot of enterprises are in experimentation mode, looking at what are the net results,” said Carol Wilder, VP for Dell’s Infrastructure Solution Group (ISG) cross portfolio software and solutions.

To help usher enterprises along in their genAI journeys, Dell is today expanding its genAI portfolio from initial-stage inferencing to model customization, tuning and deployment. The company is also announcing its development around what it calls a “milestone” new open modern data lakehouse tool.

“Your data is more valuable with customization,” said Wilder. “You can have a faster business outcome, a more specific business outcome.”

Generative AI is undoubtedly one of the world’s fastest-growing technologies: Its market size is expected to reach nearly $77 billion by 2030, representing a compound annual growth rate (CAGR) of 31.5% from 2023.

Goldman Sachs predicts that the technology could drive a 7% (or $7 trillion) increase in global GDP. Furthermore, Gartner puts the overall AI software market at nearly $135 billion by 2025, reflecting a 31% acceleration in market growth.

The Dell Technologies Generative AI Pulse Survey — also out today — reflects these projections, finding that 78% of IT decision makers are largely excited in genAI’s potential and 76% expect it will be significant, if not transformative, for their businesses.

Additionally, 76% of IT decision makers are increasing their budgets for AI initiatives and anticipate the technology to provide productivity gains, streamline processes and help achieve cost savings.

“Most IT leaders expect to have some meaningful results within 12 months,” said Wilder. “However, there's a lot of hesitancy.”

The survey identified several reasons for that trepidation: security concerns (data and IP risks), technical complexity, data governance (regulations and compliance), cost of implementation and ethical and responsible implementation.

For these reasons, just 16% of enterprises prefer public models in the cloud; the majority of IT decision maker respondents (82%) prefer an on-premises or hybrid model for genAI tools.

IT decision makers also report issues with customization: 37% are considering building an enterprise-specific large language model (LLM) because they need to go beyond pre-trained models for their corporate data requirements.

“They're already finding that their pre-trained models are not sufficient for their success, and they're having to customize those models,” said Wilder.

The expanded Dell Validated Design aims to help organizations leverage genAI by integrating best practices for customizing, fine-tuning, deploying and managing models.

The expanded platform includes data preparation services because “first of all, you bring AI to your data,” Wilder said.

The new Dell Implementation Services helps establish an operational genAI model for inferencing and customization. Dell Managed Services can then manage the full Nvidia-based platform, continuously optimizing infrastructure, monitoring and tracking performance and handling day-to-day activities.

Enhanced Education Services also help reduce the data learning curve to help customers gain critical genAI skills, Wilder explained.

“Now we're delivering guidance and examples of how [enterprises] can transfer learning model tuning and prompt engineering to take available models and tune them to exactly what they’re looking at for a business outcome,” she said.

The company first announced its collaboration with Nvidia in May, then rolled out the first phase of Dell Validated Design inferencing tools in July. The new features announced today will enable enterprises to now actually deploy and take advantage of genAI, Wilder said.

“In inferencing there's a pretty low bar in that we're running and just inferring what is needed,” she said. “When you get to model customization and tuning, it's really taking a model, understanding how that model works and tuning it to the actual use case.”

The new capabilities allow customers to use performance data that Dell tested on Nvidia as a “recipe for model customization.”

With this step up, Wilder added, higher levels of GPUs are also required, so “hardware requirements step up from inferencing.”

To further support genAI applications and broader analytics scenarios, Dell announced today that it is building out an open modern data lakehouse.

Greg Findlen, SVP for ISG data management called it the “first phase of a multi-phase journey.”

He pointed out that AI, machine learning (ML) and big data analytics “require much faster, much more open capabilities on the data layer side.”

The lakehouse tool integrating Starburst data analytics software is expected to be globally available in the first half of 2024. It will be “multicloud by design” to store, process, manage and secure data across on-premises, cloud and edge environments.

The tool will help reduce data movement and gravity to help discover, query and process data where it is, including at the edge, Findlen explained. It will leverage the latest high performance engines to query and process data for different workloads.

Open scale-out architecture will help to decouple compute and storage, and a unified data stack will handle different workloads (business intelligence, AL, ML, genAI) and structured, semi and unstructured data.

Furthermore, the integrated turnkey tool will provide data security and governance including access control, masking, encryption and data localization, Findlen said.

“Customers are having to deal with stitching pieces of this together by themselves,” he said.

Also, “many of our customers feel like they're on a treadmill where they need to consolidate all of their data in one place before their data scientists can start to use it.”

The platform will allow enterprises to leverage data where it exists, identify what data to consolidate and accelerate the pace at which data scientists can access data.

Wilder, for her part, pointed out that the data lakehouse will allow enterprises to work more efficiently with data, making it elemental to genAI. “Everyone is in love with LLMs, but the data story is really foundational,” she said.

Per Dell’s survey, 44% of organizations are at the early-to-mid stage in their genAI journey and 65% of IT leaders whose organizations have moved beyond pilot stages expect near-immediate value (within 12 months).

To help accelerate this process, Dell will work with customers through Dell Validated Design from day zero “all throughout operate and scale,” Wilder said.

On the initial step — day zero — Dell will consult with customers to establish strategy and vision and identify use cases and models. This step includes value stream mapping and an accelerator workshop.

Wilder pointed out that “use cases can be extremely broad” and that enterprises must identify their objectives and what they want to accomplish with genAI. It is also important to analyze their market segment.

Then, on “day 1,” Dell will perform data management and engineering to help customers discover and validate data for models.

“We'll work with customers on realistically what is the right way, the best way, to do data preparation services,” Wilder explained. “A modern data strategy is required to bring AI to the data.”

Then comes “day 2” services, in which use cases are established and models are deployed and tested. “Day 2-plus” management and education services then help enterprises operate and scale genAI.

Potential use cases for the expanded Dell Validated Design offering include marketing and sales content generation; chatbots and virtual assistants; and software development.

Customers will be able to deploy genAI models with infrastructure including the Dell PowerEdge XE9680 or the Dell PowerEdge XE8640, with a choice of Nvidia Tensor Core GPUs and Nvidia AI Enterprise software, which offers frameworks, pretrained models and development tools. Storage options will include Dell PowerScale and Dell ObjectScale.

Dell Validated Design model customization will be available globally through Dell Apex beginning in late October, while Professional Services for Generative AI will be rolled out in select countries starting in late October.