In early July, memory chip giants like Samsung saw their stock prices plunge despite surging profits, raising concerns about AI's market leadership. Samsung Electronics is expected to report a 19-fold surge in second-quarter operating profit, and TrendForce forecasts that DRAM and NAND contract prices will continue rising in the third quarter. Yet Samsung's stock plunged 6.9% on July 7, with memory giants like Micron Technology and SanDisk following suit. The core of AI demand has shifted from hardware procurement to task penetration, with token consumption and electricity usage continuing to skyrocket. A 2026 report from the International Energy Agency (IEA) shows that electricity consumption per AI task has dropped by at least 90% annually in recent years, yet electricity usage by AI-dedicated data centers grew by 50% in 2025. Google disclosed that its monthly token processing volume has grown approximately 330-fold over two years. Frontier labs including DeepSeek, Zhipu AI, OpenAI, and Anthropic are collectively moving to develop custom inference chips. In capital markets, MiniMax's crash due to a valuation model shift sounded an industry alarm. Zhipu AI's CEO issued an internal letter, pivoting the company narrative toward long-horizon tasks and AGI vision.