AI inference cost war intensifies
Key Questions
What major price cuts have occurred in AI inference models?
DeepSeek V4 Pro saw a 75% cut, while Xiaomi and Tencent implemented reductions of up to 99% and 97.5% respectively. These moves reflect intensifying competition in the inference cost war.
Which companies are raising significant funding in the AI infrastructure space?
DeepSeek is raising $7.4B at a $50B+ valuation and doubling headcount, while Together AI raised $800M at an $8.3B valuation. SambaNova raised $1B at an $11B valuation focused on inference chips.
How might falling inference costs benefit application-layer companies?
According to The AI Pyramid framework, commoditization at the model layer benefits application companies like Creatify by lowering costs. However, it also intensifies overall market competition.
DeepSeek V4 Pro 75% cut, Xiaomi up to 99%, Tencent 97.5%. DeepSeek raising $7.4B at $50B+ valuation, doubling headcount. US holds off blacklisting but 100+ firms deemed risks. GPU rental down 40%. OpenAI pondering price cuts. Capital continues flowing to infrastructure: Together AI raised $800M at $8.3B valuation, Twelve Labs raised $100M for video AI, SambaNova raised $1B at $11B valuation for inference chips. A strategic framework article (The AI Pyramid) argues the model layer is commoditizing, benefiting application-layer companies like Creatify but also intensifying competition. OpenAI dropped GPT-5.6 family (Sol, Terra, Luna), a major model release that could further drive down costs and increase competition in AI video generation.