AI costs spike as subscriptions hit pricing wall — firms turn towards Chinese LLMs, open-source models to extend budget

As artificial intelligence (AI) models become more widely adopted, enterprises are facing escalating computational costs. Recent analysis reveals that while subscription prices appear reasonable, actu... Read original →

Soaring AI inference costs are triggering a deep-stack recalibration across semiconductors and software. While 3nm nodes and EUV tools were optimized for high-throughput AI workloads, enterprise migration to lightweight open-source models like DeepSeek undermines sustained demand for premium GPUs and HBM DRAM—pressuring NVIDIA, SK hynix, and Samsung’s capex assumptions. On compliance, adopting Chinese LLMs introduces data sovereignty and export control risks, especially when U.S. entity-listed vendors are involved, inflating supply chain redundancy costs. Strategically, OpenAI and Anthropic may be forced into tiered token pricing, while Meta and Amazon will leverage Llama and Bedrock to lock in cloud ecosystems. Within 18 months, AI agents with dynamic model routing will become standard, collapsing today’s subscription illusion into a hybrid, on-demand MaaS reality.