Back to feeds
Aravind Saji
•
IIITKottayam • 1d
DeepSeek just shook the AI world—and NVIDIA felt it. A single company wiped 17% off NVIDIA’s market cap—that’s $600B (~₹48,00,000 Cr)gone overnight. For perspective, that’s: - More than the combined value of TCS, Reliance, and HDFC Bank - Bigger than the GDP of Pakistan, Sri Lanka, and Nepal combined - More than two decades of VC funding in India Why? AI training is insanely expensive—OpenAI and others burn $100M+ a year on NVIDIA’s $35K GPUs. DeepSeek rewrote the rules: ✅ Slashed training costs from $100M → $5M ✅ Used ‘crippled’ NVIDIA GPUs (cheaper due to artificial limits) ✅ Created ‘expert clusters’ instead of training an entire model at once ✅ 95% cheaper API costs—runs on gaming GPUs, not data centers The impact? 🚀 AI training isn’t just for billion-dollar giants anymore 🚀 Smaller players can now build cutting-edge AI models 🚀 NVIDIA & AI giants might lose their pricing power And the wildest part? DeepSeek made their research public. Now, anyone can replicate it. This isn’t just about AI—it’s a masterclass in innovation, cost-cutting, and first-mover advantage. If AI training costs drop 20x overnight, what happens next? 👇 #AI #DeepSeek #Innovation #StartupLessons
Replies (5)
More like this
Recommendations from Medial
Download the medial app to read full posts, comements and news.