The question for enterprises is no longer whether they should adopt AI. The real question now is how quickly they can rebuild their organizations around it.
That was the core message from Dell Technologies Vice Chairman and COO Jeff Clarke during his keynote at Dell Technologies World in Las Vegas on Tuesday. According to Clarke, the future belongs to the AI-native enterprise, and the opportunity to build one is already here.
Twelve Months That Changed Enterprise AI
Clarke opened by highlighting just how dramatically the AI landscape has shifted in only one year.
AI has evolved from being an assistant to becoming an operator. Model costs have dropped by roughly 80%. Token consumption surpassed 100 trillion annually in 2025. Context windows expanded beyond one million tokens, making it possible to process an entire codebase or a full year’s worth of contracts in a single pass.
At the same time, enterprise generative AI software spending tripled to $37 billion in 2025. Physical AI also moved beyond experimentation and into real-world environments including factories, warehouses, hospitals, and farms.
In previous technology eras, any one of those milestones would have dominated headlines for months. Today, they simply define the starting point for what comes next.
“What we thought was a three-year journey happened in twelve months,” Clarke said.
Three Shifts Enterprises Can No Longer Ignore
Drawing from Dell’s internal operations and customer deployments, Clarke pointed to three major patterns reshaping enterprise AI.
The first is the collapse of token costs alongside explosive growth in token usage.
Token costs have fallen roughly 80% year over year, while reasoning-related token usage has increased by 320 times. Clarke said the industry has seen similar patterns before across bandwidth, storage, and compute, but never at this pace.
He shared an example from Dell’s engineering teams, where developers consumed an entire month’s worth of tokens in just a few hours. Not because systems failed, but because the AI systems were being heavily used and delivering value.
The second shift is the uneven nature of AI productivity.
According to Clarke, roughly 5% of people inside organizations deploying AI today are generating 95% of the value. Once users learn how to work effectively with AI agents, they begin using them across nearly every task.
Clarke framed the challenge clearly:
“Who are my super users? What are they doing? And how do I get everyone to become one of them?”
The third shift is financial.
Tokens are rapidly becoming a direct operational expense. As AI agents take on more cognitive work, spending shifts away from headcount and toward token consumption. Clarke argued that organizations planning for this change now will gain a major advantage over those that delay.
All of this leads to a larger conclusion: most enterprise operating models were built for a world that no longer exists.
Traditional systems assumed output scaled with human labor hours. Agentic AI breaks that equation entirely.
“The companies that win in the next decade will be AI native,” Clarke said.
He challenged enterprises directly:
“Just because you didn’t start with AI doesn’t mean you can’t get there. Are you ready to tear down to the ground the old ways and build the new way?”
The Five Imperatives of the AI-Native Enterprise
Clarke emphasized that speed alone is not enough. Enterprises must move quickly, but they must also build correctly.
He outlined five priorities organizations need to address immediately.
1. Build an AI-Ready Data Foundation
Most enterprise data today is spread across dozens of systems, and between 80% to 90% of it remains unstructured.
For Clarke, the architectural principle is simple:
Move AI to the data instead of moving data to the AI.
2. Build Distributed AI Infrastructure
Training and inference workloads are fundamentally different, Clarke explained.
Modern reasoning models running multi-step chains are now 10, 100, or even 1,000 times more compute-intensive than the workloads enterprises were running just 18 months ago.
The AI-native enterprise must be designed to support both efficiently.
3. Secure Autonomous Systems
AI agents increasingly interact with CRMs, ERPs, financial platforms, and customer databases.
Every one of those interactions must be secure, traceable, and reversible.
“When an agent takes an action on your behalf,” Clarke said, “you need to know what it did, why it did it, and how to undo it if it got it wrong.”
In an AI-driven workforce, every action needs accountability.
4. Integrate the Enterprise Stack
AI agents need to plan tasks, use tools, execute workflows, and manage exceptions across the enterprise stack.
According to Clarke, isolated agents create expensive inefficiencies.
“A siloed agent is an expensive agent.”
5. Restructure Around Agentic AI and Token Economics
Enterprises must decide where workloads belong across edge, on-premises, and cloud environments.
Clarke warned organizations against running everything in a single environment purely for convenience.
“Be ready for a surprise, a large bill that only gets larger.”
He argued that token routing decisions will become some of the most important infrastructure choices enterprises make in the coming years.
Partners Expanding the AI Ecosystem
Google Cloud CEO Thomas Kurian joined the keynote remotely to discuss Gemini availability on Google Distributed Cloud running on Dell infrastructure, including air-gapped deployment options for organizations with strict security and sovereignty requirements.
Dell itself is serving as customer zero for the joint solution.
“The hardware is from Dell and the software is from Google Cloud, but we take joint responsibility for your success,” Kurian said. “We’re making AI accessible on premises for business and governments of all sizes with security at the core.”
Clarke also brought Dave Morin, co-founder of Offline Ventures and co-founder and board member of OpenClaw Foundation, onto the stage.
Morin encouraged enterprises to start experimenting immediately.
“Give people the tools to try this,” he said. “Models need context in order to provide the best possible answer. People are coordinating their work in vastly more efficient ways. Get started. Don’t be afraid.”
For Clarke, the message was straightforward:
“The enterprise has to own its AI, not rent it.”
AI at the Desk, Not Just the Data Center
One of the keynote’s most striking demonstrations focused on Dell Deskside Agentic AI.
Clarke explained that a single developer running agent workloads in the cloud could generate a daily bill of $3,400. Running those same workloads locally on a workstation brings the cost close to zero.
The example involved one developer, ten agents, and one billion tokens processed in just 24 hours.
“That’s not a demo,” Clarke said. “That is a deployment model we’re helping customers build right now.”
Dell’s updated Pro and Pro Precision portfolio now spans a wide range of local AI compute needs, from the GB10 system delivering one petaflop for individual agent prototyping to the GB300 platform reaching 20 petabytes for large-scale workloads.
The GB300 consumes more than 1,500 watts and requires a closed-loop liquid cooling system that Clarke noted is currently unmatched elsewhere.
Building the Infrastructure Behind AI-Native Enterprises
Arthur Lewis, President of Dell’s Infrastructure Solutions Group, joined Clarke on stage to explain how Dell is scaling AI infrastructure from the desktop to the data center.
Lewis emphasized that most enterprise data still lives on-premises, and much of it remains inaccessible to AI systems.
Dell’s AI Data Platform is designed to address this challenge through an architecture focused on data preparation, high-speed disaggregated inference, and storage infrastructure that allows AI agents to access information wherever it resides.
Cyber resilience was another major focus.
Lewis noted that 94% of ransomware attacks attempt to compromise backup data, with 57% succeeding.
Dell’s new PowerProtect One platform combines backup appliances and management software into a single integrated solution. According to Dell, it reduces time to first backup by 75% and cuts daily management time in half.
Lewis also highlighted the refreshed PowerStore Elite platform before turning attention to Dell AI Factory, the company’s integrated full-stack solution combining compute, networking, storage, and software into a unified system.
More than 5,000 customers are already running workloads on Dell AI Factory, with reported returns reaching up to 269% ROI in the first year.
Sandisk used Dell AI Factory with NVIDIA to reduce factory costs by 32%, lower energy consumption by 46%, and cut defect rates from 800 parts per million down to 100.
What Dell Learned From Its Own AI Transformation
Clarke also shared lessons from Dell’s own internal AI adoption.
Three years ago, Dell discovered thousands of shadow AI projects running inside the company. Clarke said the issue was not governance failure, but evidence of demand.
Dell responded by focusing on five outcome-driven use cases, creating an internal AI Office, and reducing project sprint timelines from 90 days toward just three hours.
Today, Dell’s service assistant is deployed across its global services organization, running on Dell infrastructure inside existing on-premises data centers.
According to Clarke, the investment achieved ROI in less than three months.
“This is not a pilot,” he said. “This is how our company is running today.”
The Choice Facing Enterprises
Clarke closed his keynote with three challenges for enterprise leaders.
First, organizations must honestly evaluate whether their AI budgets account for both compute costs and token consumption.
Second, they must identify their AI super users and build around them.
“The gap between your super users and everybody else is the gap between your future and your past,” Clarke said.
Finally, enterprises need to decide whether they will lead the operational transformation driven by AI or be disrupted by it.
“Both options are in front of you,” Clarke said, “but only one of them is yours to choose.”
The message throughout the keynote was unmistakable: the modern enterprise is no longer a fixed destination. It is an operating model built for constant transformation, and the work to build it starts now.
