Thoughts on AI compute supply for year-end
Been watching the AI space closely, particularly the bottlenecks in compute supply. Everyone's talking about the demand for GPUs, but the broader infrastructure – power, cooling, real estate – is becoming a significant factor. I'm looking at $BRN, which is trading around 1.07 today, and thinking about the longer-term infrastructure play.
My take is that we're heading into a period where the constraint shifts from just chip availability to actual data center capacity. I'd give it a 60% probability that by Q4 2024, the narrative around AI will be less about 'who has the best model' and more about 'who has the capacity to run their models at scale'. This will disproportionately benefit companies involved in data center build-out and specialized power solutions. It's a longer-term trade, but the trend seems clear.
ผมว่าเรื่องไฟฟ้ากับความร้อนนี่แหละตัวปัญหาใหญ่ ไม่ใช่แค่ GPU ละตอนนี้ จะเอากำลังประมวลผลเพิ่มแต่ไม่มีไฟให้ใช้ก็เท่านั้นเองนะ จะทำยังไง