Call Center Doctors Tested DeepSeek GPU Rental Costs
The firm examined if DeepSeek could lower coding expenses compared to Claude Code subscriptions.
Updated on Oct. 1, 2026 in Job Search
Live Poll
Is now a good time for your business to switch from proprietary AI to open models?
The Call Center Doctors rented four Nvidia H200 GPUs to determine if DeepSeek models could reduce software development costs. The test failed to fully implement code-writing agents due to security vulnerabilities identified during the trial.
Why it matters
The firm conducted this experiment to verify claims that DeepSeek is 80 times more affordable than Claude. If successful, such a shift could significantly alter operational budgets for firms relying heavily on AI-assisted coding.
The firm paid $13,200 for a monthly GPU rental, while testing identified a daily server token capacity of 20 billion. In September, the firm recorded 2.03 million model calls and 5,610 merged code changes.
The players
Call Center Doctors
This professional consultancy firm specializes in optimizing operational workflows and software integration.
Nvidia
This technology company designs the H200 graphics processing units used for high-performance artificial intelligence computing.
The details
The consultancy configured the server to run DeepSeek V4.1 Flash to support Claude Code agents, but the setup faced sandbox escape risks. Consequently, DeepSeek functioned only as a read-only reviewer during the tests before the rental provider reclaimed the server hardware.
Timeline
September 1-27: The consultancy tracked its Claude Code subscription usage and token volume.
September 27: The consultancy rented a server equipped with four H200 GPUs.
Market Landscape
This move reflects a growing trend among firms attempting to move from subscription-based API models to self-hosted infrastructure to optimize costs. By testing open-weights models, the firm is positioning itself to bypass high per-token pricing charged by major AI developers.
Small businesses and developers may see future price volatility in AI tools as firms experiment with cheaper hosting alternatives. Companies that successfully implement self-hosted AI models could eventually offer lower service prices to their own clients.
The takeaway
Companies should prioritize robust security sandboxing when integrating external AI models into proprietary codebases to avoid data risks. Relying on cost-efficiency alone is insufficient if the infrastructure fails to meet safety requirements for production environments.
Further reading
Explore more developments in Job Search regarding how AI is reshaping the technical workforce.
Source note: This article includes information reported by Tom's Hardware.
Live Poll
Is now a good time for your business to switch from proprietary AI to open models?
