The AI industry chain is accelerating at both ends

To keep it brief, focus only on both ends
Upstream DDR: Changxin Storage went public, likely to replicate CATL's price slaughter in batteries, but the real industry chain is not a single-threaded 500-direction K-line on the chain; everything takes time. Memory shortages will gradually ease within 1-2 years, and supply growth will reverse and surpass demand growth.
HBM: Changxin Storage still can't do long-term storage for now; since SK Hynix, Samsung, and others dare to abandon the consumer market, it’s primarily due to long-term agreements and a technological moat.
At the very bottom, large model manufacturers are actively getting involved, consolidating efforts.
1. OpenAI and Anthropic have respectively partnered with PE and VC firms to establish joint ventures, automating business processes of these PE/VC-invested network companies, recruiting talent, and boosting performance.
2. Google is hiring AI on-site engineers on a large scale, similar to the Palantir model, accelerating B-end business deployment.
3. Google and Blackstone have formed a joint venture Cloud company, leveraging Blackstone’s investment network to attract AI cloud clients.
If B-end and C-end customers actively subscribe to large models, that’s farming; but the proactive approach of large models can be seen as hunting.
Several hype scalers are issuing bonds worldwide to increase CapEx for computing infrastructure, but at the same time, OpenAI, Anthropic, and Google are actively attracting clients, starting to verify and realize the returns from B-end customers through hunting.
Especially OpenAI and Anthropic need impressive numbers to support their future trillion-dollar market value.
To sum up in one sentence, the landlords are running out of resources too.
If B-end customer revenue and AI deployment are not ideal, the rapid growth of large-scale AI investments may reverse, and the US stock market could face a deep correction.
Invest in expectations, stay alert, stay safe.