【NVIDIA’s next move: bring the AI factory out of the data center】🔧⚡
Group chat:
📲 加入X先生的粉丝群聊 NVIDIA’s first wave of AI—was stacking computing power into ultra-large data centers.
The next wave of money-making opportunity is to disassemble the AI factory.
Bring it into ordinary data centers and telecom offices.🔧
Now, the most high-end AI racks can reach 140 kilowatts per rack.
But a typical edge data center can only support 30 to 50 kilowatts per rack at most.
If you force it in, you’ll burn the circuits directly—cooling and structural load just can’t handle it.⚡
The solution is straightforward.
Take a 140-kilowatt computing unit and split it into four or five 30-kilowatt racks.
Then connect them into a single logical whole using high-speed optical interconnects.
Fiber cables have plenty of bandwidth, and they’re not limited by distance and power like copper cables.✅
In the industry, there are about 30 million kilowatts of edge power distributed across locations.
If you couldn’t handle AI before, this decomposition approach lets you do it.
It’s like turning idle computing capacity back into productive use—this move contains a lot of information.🤔
To be honest, the compute industry is shifting from competing by piling up hardware.
It’s moving toward competing on how to fit that hardware into ordinary data centers.
Interconnect equipment is the new “water seller.”👀
📌 AI compute is moving from centralized to distributed. The next multi-billion-dollar market is hidden inside high-speed network interconnects.
➕ Join the fan group for strategies!🔥 Every day, I’ll help you understand crypto market hot spots and institutional capital flows in the simplest way, so you can catch the next opportunity!🚀
#AI #算力 #科技 #XRP领跌加密市场跌近7% #美元创近四周最大涨幅