Hi everyone,
Here is my portfolio allocation as of the end of January 2026.
2026 YTD Return: -6%
Current Portfolio:
| Ticker | Jan Allocation | Dec Allocation |
|---|---|---|
| IREN | 23.84% | 17.80% |
| APP | 15.30% | 22.99% |
| NVDA | 15.01% | 14.89% |
| ALAB | 12.44% | 10.54% |
| EOSE | 11.30% | 9.10% |
| NBIS | 11.20% | 10.96% |
| RBRK | 5.61% | 8.40% |
| NET | 4.21% | 4.79% |
| HIVE | 1.09% | 0.51% |
Transactions: During this month, I added to my positions in ALAB and NBIS.
Market Observations & Thoughts
1. The “Good Enough” Problem: AI’s Threat to General SaaS
I have been observing how AI is impacting Software-as-a-Service (SaaS) stocks, and my personal experience at work confirms some fears. I increasingly use AI for summarization and handling mundane tasks, bypassing traditional software workflows.
A real-world example: In the past, if I wanted to know what clients were demanding, I required my employees to manually input client requests into project management software and add specific classification tags. This created friction and required disciplined data entry.
Now, that manual step is obsolete. I can simply dump raw emails and historical documents into an LLM and ask it to summarize the themes. It saves labor and removes the need to integrate business processes into rigid software.
While the LLM’s output might be slightly less precise than a dedicated, structured software tool, it is vastly cheaper and frictionless. In productivity, “good enough” combined with “cheap and easy” is a category killer.
2. The Safe Harbor: Cybersecurity
Because of the trend above, I believe general productivity SaaS is vulnerable. However, Cybersecurity is the exception. In security, “good enough” is not acceptable. You cannot tolerate “slightly weaker” precision when protecting a network. Because of this high barrier to entry and the requirement for perfection, I view cybersecurity companies (like RBRK and CRWD) as the most resilient against the AI disruption wave.
3. Model Commoditization vs. Infrastructure
Another clear trend is the shrinking performance gap between open-source LLMs and top-tier proprietary models (OpenAI, Anthropic, Gemini).
The difference is no longer about raw intelligence; it is shifting toward user experience (UX) and data security. If the core intelligence becomes a commodity, companies like OpenAI, whose primary product is the model itself, face significant business risks—potentially even bankruptcy if they cannot differentiate.
This reinforces my conviction in the “picks and shovels” of this revolution. I believe the best risk/reward ratio lies in:
-
Upstream Hardware: (e.g., NVDA)
-
Downstream Inference: (e.g., IREN, NBIS)
A Counterpoint to myself: My thesis assumes intelligence is commoditizing. However, if the market shifts heavily toward robotics and multi-step reasoning agents, raw performance precision becomes critical again. For example, in a task requiring 10 sequential reasoning steps, a small difference in accuracy compounds massively (e.g., success rate 0.95^10≈60% and 0.95^10≈54%). I am watching this, but for now, I am betting on infrastructure.
Finally, a note on my participation here. In my journey to follow Saul’s method, I realized I have been too focused on short-term gains and comparing my monthly returns to others. This mindset often leads to noise rather than signal.
To foster a more long-term, stable strategic mindset, I may move my portfolio updates from monthly to quarterly.