The memory shortage could get MUCH worse in 2027. According to Dell's CEO, the memory shortage isn't going away anytime soon. In fact, he expects it to get worse in 2027 because building a new memory fabrication plant takes roughly four years. And this is exactly why I've been so bullish on the memory trade. Demand for DRAM, NAND, and HBM continues to increase as hyperscalers aggressively expand their AI infrastructure, but manufacturers cannot simply bring new capacity online overnight. Even with billions being invested into expanding production, much of that additional capacity won't be available for several years. Meanwhile, AI inference and agentic workloads are creating entirely new sources of memory demand beyond just training larger models. As these systems become more sophisticated, they require more memory to handle larger context windows, store information, and process increasingly complex workloads. And remember, it's not just AI companies competing for this supply. Traditional data centers, smartphones, PCs, and enterprise servers still need memory, which means manufacturers have to balance growing AI demand with the rest of the market. This is why I believe the current memory cycle could last much longer than previous cycles. If Michael Dell is right and shortages worsen in 2027, we could see memory pricing remain elevated, stronger margins for manufacturers, and potentially a much longer earnings growth cycle than investors are currently pricing in. I still believe memory is one of the most underrated opportunities in the entire AI infrastructure trade. If you want to see how I’m investing around this shortage and the names I think benefit most, check out my full Milk Road Pro portfolio below. link.milkroad.com/d8cj6v

Anthropic just filed for its IPO and some of these numbers are absolutely insane. The company generated $16.3 billion in revenue during the first half of 2026, compared to just $1.1 billion during the same period last year. That's incredible growth but what I find even more interesting is how much money Anthropic is spending to sustain it. Anthropic reportedly lost $80.7 billion during the first six months of 2026, although roughly $72 billion came from non cash accounting charges related to convertible financing. So the actual cash burn is nowhere near that headline number. However, the company still spent nearly $9 billion on compute and infrastructure during that period and has at least $518 billion in longer term infrastructure commitments. And now we have Anthropic reportedly agreeing to spend up to $120.5 billion on computing capacity from SpaceX through 2029, with payments potentially reaching $3.1 billion per month once all the capacity is delivered. We're talking about a company spending billions every single month just to secure enough compute to remain competitive. Meanwhile, Anthropic is expanding its revolving credit commitments from $3.3 billion to $15 billion and increasing its minimum liquidity requirement from $500 million to $5 billion. Essentially, the company is securing access to significantly more financing while making sure it has enough cash available to support these enormous commitments. Another interesting detail is that Anthropic's seven co-founders will retain 50.1% of the voting power through a special Class F share that carries no economic rights. So even after going public, the founders will maintain control over major corporate decisions regardless of how much outside investors contribute. Personally, I think Anthropic is an incredible business, but I've also been saying for a while now that frontier AI labs are being valued way too aggressively. Growing revenue at this pace is impressive, but eventually investors are going to want to see how much of that revenue actually turns into profits after paying for all this infrastructure. And this is exactly why I continue to prefer the infrastructure side of AI. Whether Anthropic, OpenAI, SpaceX, or another frontier lab comes out ahead, they all need compute, memory, networking, and power to keep scaling. The AI labs are spending hundreds of billions competing against each other, and the infrastructure companies are the ones collecting those checks.