Companies are moving AI computation away from cloud services and onto their own hardware, driven by cost savings and data privacy concerns. Instead of paying per token to cloud providers, tech leaders are deploying open-weight models on local machines, from high-end workstations to laptops, fundamentally changing how businesses think about AI infrastructure.

Why Are Companies Abandoning Token-Based Pricing?

The financial case for local AI is compelling. Alex Lisle, CTO at cybersecurity platform Reality Defender, experimented with running an "adversarial loop" on cloud-based models, where different AI systems critique each other's work to improve code quality. The approach worked in theory but became prohibitively expensive in practice. "That was costing millions of tokens," Lisle explained, noting that models would cycle through eight or nine rounds of feedback before reaching a conclusion.

When Lisle moved the same workflow to four local DGX Spark machines, each equipped with an Nvidia GB10 Grace Blackwell Superchip, the economics shifted dramatically. "The only reason that adversarial loops work for me is the fact that I can do this without paying for tokens," he said. This shift reflects a broader trend: AT&T's chief data and AI officer told the Wall Street Journal in August that a quarter of the company's AI usage now comes from open models, avoiding the per-token costs of proprietary cloud services.

What Security and Privacy Benefits Come With Local Models?

Beyond cost, data privacy is reshaping enterprise AI strategy. Parker Johnston, agentic field CTO at solutions integrator Insight, recently tested a ZGX Fury GB300 workstation from HP, equipped with an Nvidia Blackwell Ultra GPU and 748 gigabytes of coherent memory. He ran four large language models (LLMs) simultaneously and more than 30 AI agents overnight, burning 11 million tokens in 90 minutes.

For industries handling sensitive information, local models eliminate a critical vulnerability. "Maybe it's a bank-loan application, and I can go onto this device, process the information I need, and immediately remove the files," Johnston explained. "I no longer have to worry about where they're being kept, what access folks might have, who configured the cloud environment, what's going on on the other side of it. It's all for me to now wholly own across that entire landscape".

"I'm fine-tuning them for my particular use cases, I can reason about them because I see the reasoning chains, the data never leaves my environment, and then the fixed cost. Those are all very powerful indicators of where the market is going," said Alex Lisle.

Alex Lisle, CTO at Reality Defender

Open-weight models offer transparency that proprietary cloud services cannot match. Unlike closed models from companies like Google, Anthropic, and OpenAI, open models provide access to training data sources, the code used to run them, and the parameters that drive decision-making. A paper from MIT and Georgia Tech researchers published in November 2025 found that closed models account for 80 percent of model usage, but the trend toward open models is accelerating among enterprises seeking greater control.

How to Deploy Local AI Models Across Your Organization

• Choose the Right Hardware: Tools like Ollama enable open-weight models to run on consumer-grade devices such as MacBook Pro laptops, while modern PCs increasingly include specialized GPUs designed for on-device AI tasks. High-performance workstations are available for organizations requiring simultaneous model execution and complex workflows.
• Implement Network Security Controls: While local models reduce exposure of sensitive data to third-party cloud providers, security professionals must still monitor access to system resources, protect encryption keys and credentials, and ensure models run behind corporate firewalls with appropriate access controls.
• Leverage Overnight Processing Capacity: Local hardware enables organizations to run AI agents and tasks during off-hours without incurring additional token costs, allowing development teams to perform vulnerability management, dependency tracking, and other labor-intensive tasks that would be prohibitively expensive on cloud-based, per-token pricing models.

Neil Ward-Dutton, research VP of agentic automation and AI technologies at tech-intelligence firm IDC, envisions a future where enterprises deploy models across a distributed landscape. "Do I see a future where enterprises will have AI models running in lots of different places, including on laptops, company phones, local private clouds, on-premise data centers, and so on? Absolutely," Ward-Dutton noted.

The shift toward local AI represents a fundamental change in how organizations approach artificial intelligence infrastructure. Rather than consolidating all AI workloads in cloud environments controlled by third parties, enterprises are distributing models across devices they own and operate, reclaiming control over costs, data, and computational resources. As open-weight models mature and hardware becomes more capable, this decentralized approach is likely to accelerate.