Whalesbook
AI Chip Market Trends Change Ho Rahe Hain
Artificial intelligence ka sirf Nvidia ke graphics processing units (GPUs) se juda hone ka concept ab actively toda ja raha hai, AI chip designer Cerebras Systems ke co-founder aur CEO Andrew Feldman ke anusar. World Economic Forum Davos mein Feldman ki baaton ne Nvidia ki current market leadership mein ek significant potential disruption ka ishara diya hai.
"Mujhe lagta hai humne pehle hi Nvidia ki dominance ko challenge hote dekha hai," Feldman ne kaha. "GPU hi akela machine nahi hai jiska use AI mein ho sakta hai. Un logon ka mental moat tod diya gaya hai jo sochte the ki AI ka matlab sirf Nvidia hai."
Transition Ka Evidence
Feldman ne Groq ki Nvidia dwara reportedly $20 billion ki acquisition ko is baat ka evidence bataya ki GPUs ki limitations hain. "Nvidia ne maana ki unke paas solution nahi tha aur number two player ko kharidne ke liye $20 billion kharch kar diye. Yeh dikhata hai ki GPUs kuch cheezein achhe se nahi kar paate," unhone explain kiya. Nvidia dwara Groq ki technology ko acquire karne ka yeh move ek strategic effort hai, jabki Groq ek independent entity ke roop mein kaam kar raha hai.
Google Ka Shift aur Inference Par Focus
Feldman ke anusar, ek bada industry inflection point Google ka apna Gemini AI model train karne ka faisla tha, jismein Nvidia hardware ki jagah apne Tensor Processing Units (TPUs) use kiye gaye. "Humne dekha ki ek foundation model ki training poori tarah Nvidia ke bina hui. Yeh ek bada moment tha," unhone kaha.
Cerebras CEO ne rapid inference ki market ke tezi se pivot hone par zor diya – woh critical stage jahan AI models ko real-world applications ke liye deploy kiya jaata hai. "Inference woh jagah hai jahan AI real economy se milti hai. Wohi coding, agentic work aur deep research hota hai," Feldman ne samjhaya. Cerebras Systems, jo 2016 mein bani thi, apni Wafer Scale Engine (WSE) offer karti hai, jise duniya ka sabse bada AI accelerator kaha jaata hai, aur iska aim NVIDIA jaise established giants se sidhe compete karna hai.
AI Bubble Concerns Ko Dismiss Karna
Feldman ne artificial intelligence market mein bubble ke daron par bhi baat ki, is idea ko dismiss karte hue. "Yeh bubble nahi hai. Yeh woh tareeka hai jissey ek nayi technology poori economy mein spread hoti hai... Log AI ko tez karna chahte hain. Woh intezar nahi karna chahte. Tez inference hi hai jisme users aaj sabse zyada interest rakhte hain," unhone conclude kiya.