How to lower your AI costs: 3 ways the NVIDIA-powered Dell Pro Max with GB10 and GB300 offer an alternative to token-maxxing

The Dell Pro Max with GB10 and GB300 are AI accelerators built to bring high-performing, local AI compute directly to an employee’s desk. Powered by NVIDIA Grace Blackwell superchips, the two devices give enterprises a way to run demanding AI and agentic workloads on-premises, helping curb the runaway token costs typically associated with cloud-based APIs.

TL;DR

• The cost of AI tokens are rising fast, with projections showing they could exceed average software developer salaries within two years
• Local AI capabilities help curtail runaway token consumption and reduce costs for enterprises, while scalability allows enterprises to continually keep pace with growing workloads
• Deskside Agentic AI with products such as the Dell Pro Max with GB10 and GB300 bring high-performance, local AI compute to the desk, giving enterprises a cost-effective alternative to cloud-based token pricing

Rising AI costs have become a recurring flashpoint in the technology industry over the past 18 months. The rise of ‘token-maxing’ combined with growing infrastructure-related costs mean some enterprises now face mounting bills.

Notably, the enterprise focus is shifting toward agentic AI adoption. However, this push is causing some financial headaches. Agents are more expensive than traditional chatbots, mainly because a single prompt requires more multi-step, LLM-powered reasoning capabilities.

Research shows agents typically consume between four and fifteen times more tokens compared to traditional chatbot interactions.

The scale of this price disparity was laid bare by a recent Gartner® study, which mentions, “By 2028, AI coding costs will overtake the average developer’s salary due to rising large language models (LLM) token consumption and the shift to consumption-based licensing models.”

With agentic AI adoption expected to continue at pace, there’s no sign these concerns will dissipate either. Separate research from Gartner projects, “In 2026, global spending on inference ($23.3 billion) will surpass that of training ($19 billion). Fifty-five percent of AI-optimized IaaS spending is forecast to support inference in 2026 and is set to reach 59% in 2027.”

With these considerations in mind, it is vital that organizations embarking on (or scaling) agentic AI adoption projects take these factors into account. For IT leaders, the solution to unchecked token consumption could lie in new, dedicated hardware.

Deskside AI

AI accelerators aren’t just a new industry trend. They offer a solution to the confluence of issues facing enterprises innovating in AI, particularly around costs.

Devices such as the Dell Pro Max GB10 and GB300 provide enterprises with high-performance local AI capabilities. Powered by NVIDIA Grace Blackwell chips, these devices are aimed primarily at AI developers, researchers, and data scientists.

The GB10, for example, is capable of handling 30-200 billion parameter AI models and can run up to eight agents concurrently. The larger GB300, meanwhile, offers data center-class capabilities.

This is capable of handling 120 billion-to-1 trillion parameter models alongside up to 150 concurrent agents.

These devices offer enterprises both the ability to deliver high-performance AI capabilities to an employee’s desk and help keep a tight grasp on potential overspend.

Here’s how the GB10 and GB300 can help manage token consumption.

Local orchestration

The Dell Pro Max with GB10 is designed with token consumption in mind, according to Dell. Because the GB10 allows users to run agents locally, this delivers marked cost efficiency improvements across two fronts:

• Running clusters of agents locally reduces infrastructure strain by reducing the need for a dedicated GPU cluster.
• Local orchestration also offers lower costs compared to cloud-based API tokens, which operate on a pay-per-token basis.

Agent fleet orchestration

As with the Dell Pro Max with GB10, the Dell Pro Max with GB300 also provides users with the same capabilities, albeit on a larger scale. The ability to run up to 150 agents concurrently in a local capacity drastically reduces costs compared to cloud-based API options.

Signal65 benchmarking found this agentic AI solution delivered up to 87% cost savings in software development-related use cases, working with 60 agents concurrently. Given Gartner’s projected cost increases in this domain, this represents a huge financial benefit for enterprises.

In its benchmarking, Signal65 noted: “Over a two-year period, modeled savings ranged from thousands of dollars for individual AI knowledge worker deployments to multimillion-dollar reductions for enterprise-scale autonomous AI environments.

“The results indicate that organizations can realize meaningful economic advantages from on-premises AI infrastructure across a broad range of agentic AI deployment models.”

The performance capacity afforded by the Dell Pro Max with GB300 also provides advantages. Boasting 748GB of coherent memory and up to 16TB of local storage, this means multi-billion parameter models, traditionally reserved for shared infrastructure, are now placed directly in the hands of frontline workers.

Scalability within the Dell ecosystem

With the Dell Pro Max range, scalability also provides notable advantages for users. The GB10, for example, can be used either as a standalone system or connected to another Dell Pro Max with GB10 to double performance and support models with up to 400 billion parameters.

Similarly, both workstations are interoperable with Dell Pro Precision 9 Series desktop workstations. This means users can scale capacity workloads to the desktops to reduce strain on their workstations. This improves workstation performance and security—enabling users to retain sensitive workloads locally before scaling to the cloud.

If you think Dell's deskside agentic AI workstations are the right fit for your business, find out more on the Dell website. US readers click here.

Gartner Press Release, Gartner Predicts AI Coding Costs Will Surpass Average Developer’s Salary by 2028 as Token Consumption Surges, June 24, 2026

Gartner Press Release, Gartner Forecasts Worldwide AI-Optimized IaaS Spending to Grow 96% Through 2026, August 10, 2026

GARTNER is a trademark of Gartner, Inc. and/or its affiliates.