Move over, ChatGPT—there’s a new AI king in town. And it’s not American.
DeepSeek AI, a Chinese startup, has shaken Silicon Valley to its core. In just weeks, it crushed ChatGPT, becoming the #1 free app on the App Store. The shockwaves were so strong that NVIDIA lost a staggering $400 billion in market value. That’s not just a win—it’s a full-blown AI coup.
Now, Meta is scrambling. Mark Zuckerberg has assembled four "war rooms"—yes, war rooms—to dissect DeepSeek’s secret sauce. How did this Chinese AI, backed by High-Flyer Capital, outperform the biggest AI giants while spending less than $6 million on training? (For comparison, OpenAI burned through $100 million to train ChatGPT.)
Inside Meta’s War Rooms
- Two teams are desperately trying to figure out how DeepSeek built its R1 model so cheaply and efficiently.
- One team is digging into the data DeepSeek used—where did they get it, and why is it so good?
- The last team is rethinking Meta’s AI from the ground up to compete
And it gets worse for Meta. Their own AI chief, Mathew Oldham, is warning that DeepSeek’s newest model could even outperform Meta’s next big release, Llama 4. That’s right—before Llama 4 even launches, it’s already being written off.
Even Sam Altman, OpenAI’s CEO, has admitted DeepSeek is “impressive.” And NVIDIA, despite losing billions, gave DeepSeek props—while subtly reminding the world that their GPUs powered DeepSeek’s rise.
So, is Meta too late? Will the future of AI belong to China? The AI war is just beginning. And Silicon Valley is scared.
