History doesn't repeat, but it rhymes.
January 2025: DeepSeek drops. $NVDA crashes 17% in a day. Everyone screams "compute is dead, GPUs are over." Result? Stock more than doubled from that bottom.
Summer hits, Kimi K2 runs the exact same playbook. Cheap models never killed hardware demand—they exploded inference demand because more people started using it.
75% of the world still hasn't touched an LLM. When AI goes from niche to everyone running tasks daily, memory will be the first bottleneck to blow up. This cycle is going to be more insane than you think.
January 2025: DeepSeek drops. $NVDA crashes 17% in a day. Everyone screams "compute is dead, GPUs are over." Result? Stock more than doubled from that bottom.
Summer hits, Kimi K2 runs the exact same playbook. Cheap models never killed hardware demand—they exploded inference demand because more people started using it.
75% of the world still hasn't touched an LLM. When AI goes from niche to everyone running tasks daily, memory will be the first bottleneck to blow up. This cycle is going to be more insane than you think.