Local AI plus outsourcing could soon beat frontier labs economically
Original: Outsourcing plus local AI will soon become more economical vs. frontier labs
Why This Matters
Highlights potential economic disruption to frontier AI labs from cost-competitive alternatives
Analysis suggests combining offshore engineers with local AI models like DeepSeek may become more cost-effective than using frontier closed-source LLMs. Study shows 30x price difference between DeepSeek ($0.094) and frontier models ($2.80) per million tokens, while token consumption rises across the industry.
A new analysis examines when hiring engineers in lower-cost countries with local AI access becomes more economical than using frontier LLMs. The study uses DeepSeek as a proxy for local AI costs, showing dramatic price differences: DeepSeek costs $0.094 per million blended tokens versus $2.80-$2.82 for OpenAI and Anthropic. Frontier lab pricing has increased significantly - GPT-5.5 costs 3x more than GPT-5 eight months ago, while Gemini 3.5 Flash tripled pricing over its predecessor. The analysis assumes 50k output tokens per 1M input tokens and factors in caching rates. While current closed-source models remain more capable, the question is whether the capability gap justifies the 30x price difference when combined with human engineers for coding tasks.