Another AI investment genius who can copy homework from the US stock market is the young genius Leopold Aschenbrenner, who was kicked out by #OpenAI. The fund he manages, Situational Awareness, has skyrocketed from $225 million to $5.5 billion in just over a year, increasing by 24 times.

His current holdings of 15 major US stocks are all related to AI. His logic has similarities to our usual research on public chains or AI, with a core focus on fundamental hard constraints. Invest in electricity when there's a lack of it, invest in memory when there's a shortage of it, and invest in data centers when there's a need for them.

By the end of 2025, the top 5 holdings disclosed by his fund are as shown in 👇 picture 2. These 5 companies have seen astonishing growth in 2026, bringing the fund close to a $2 billion increase, making them great representatives for copying homework, with homework links in the comments!

#BE addresses the "electricity shortage issue".

#LITE tackles the "data transmission issue".

#CRWV resolves the "cloud and token economy issue";

#SNDK solves the "data storage issue".

#CIFR deals with the "electricity borrowing issue for data centers";

1️⃣Bloom Energy (#BE), the "personal power bank" of the AI era

The core business is selling fuel cells. It doesn't rely on the power grid; it generates electricity on-site using natural gas or hydrogen, reducing the power construction cycle for AI data centers.

As everyone knows, the US power grid is old and outdated. Waiting for the grid to improve means AI data centers could miss their opportunity. Oracle is constructing ultra-large data centers for OpenAI, and it may take years to get power approval from the grid, but working with Bloom Energy can resolve it in months. Oracle is also one of BE's major clients. It is now the "off-grid artifact" for AI infrastructure.

2️⃣Lumentum Holdings (#LITE), the "highway" for computing power data centers, an absolute leader in optical communications, specializing in optical transceivers and lasers.

Today's GPU clusters, with tens of thousands of H100s, have an extremely horrifying data exchange rate. Using copper wires to connect would be slow and generate heat. Lumentum's optical modules convert electrical signals into optical signals, allowing tens of thousands of chips to work together like a brain. Without it, computing power clusters would be like scattered sand.

3️⃣CoreWeave Inc (#CRWV) is a "favorite child" computing power provider of NVIDIA. This company used to mine Ethereum, being all about GPUs, but after ETH transitioned to POS, it completely pivoted to GPU cloud services.

It is one of NVIDIA’s largest global partners and can even obtain top-of-the-line graphics cards (H100/H200/B200) ahead of Amazon and Google. It is a cloud service specifically designed for AI training, much more efficient than traditional public clouds, and is the "landlord" most AI unicorns love to rent from.

4️⃣Sandisk Corp (#SNDK) has recently surged wildly and is quickly approaching $1,000. As a storage giant, it focuses on flash memory and SSD hard drives.

Currently, everyone is competing for computing power, but the larger the AI model, the more data it needs to be fed, and the larger the "weight files" it produces. AI not only requires fast brains but also needs to store vast amounts of knowledge. Storage is a crucial segment in the entire AI chain.

5️⃣Cipher Digital (#CIFR), a "model of transformation" from a mining site to a data center, is a company that successfully transitioned from cryptocurrency mining to HPC.

First, this company was originally called Cipher Mining, and it just changed its name to Cipher Digital at the beginning of 2026. What is the most valuable asset for mining sites now? It is large-load power permits and existing distribution facilities. This company has dismantled mining facilities and transformed them into AI data centers, directly selling power capacity to AWS and Google. In the current "electricity shortage," the power approvals in its hands are like gold mines.

Young genius Leopold Aschenbrenner's investment logic is worth referencing and reflecting on. He uses the standard "first principles" approach to deduce what the AI market will lack most in the next 3-5 years, and then places bets in the capital market now!🧐

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。