Technology Is The Leverage

For most of the past year, the question we heard most often was a version of the same one. What did we think of Situational Awareness?

It was a fair question. The fund had become the most talked-about name in the AI trade. It grew on a simple idea: own the picks and shovels of the buildout, the chips, the memory, the power, and add leverage to turn a good return into a great one. For a while it worked brilliantly. Then, in a single month this summer, it came apart. The fund lost about two-thirds of its value in one month (apparently the largest loss in that timeframe in financial history) and was forced to sell its public holdings.

Two ingredients turn almost any sound technology thesis into a fragile one:

• A crowded trade. When everyone owns the same thing, there is no one left to sell to when you need to.

• Leverage. A copious amount of borrowed money requires both the direction and the timing to be impeccable.

This is not a new story, and it is not unique to AI. Every technology boom builds the same trap. Railroads, telecom fiber, and the early internet all drew enormous capital into the same infrastructure at the same time, and in each case the buildout was real, and the overcrowding was fatal to the people who financed it with borrowed money.

Retail investors crowded into the same handful of memory and AI names and bought them on margin, often through leveraged single-stock products. Those products amplified the rally on the way up and accelerated the decline on the way down through forced selling. More than a million accounts (in South Korea) faced margin calls, and several hundred thousand ordinary investors were wiped out.

So the real question is not what we think of one strategy. It is where we go from here, and whether any of this changes our conviction about technology and AI.

It does not. It sharpens it.

Here is where we stand and why it is different from almost everyone else on the board:

Illustrative only, based on limited publicly available information and internal analysis.

• We buy the users of a technology, not its sellers. In this cycle that means owning companies that put AI to work, not the companies selling the hardware or models it runs on.

• We avoid the most crowded and most cyclical layer of any buildout. We generally hold no net long position in semiconductors, semiconductor equipment, or the hardware complex.

• We do not use meaningful financial leverage. Our returns are meant to come from operating leverage, the kind a real business produces when technology lets it do far more against the same cost base.

• We would rather own the company that gets permanently better. A business that uses technology to widen its margins, take share, and pull away from its competitors is a more durable holding than a supplier hoping to sell the next generation of equipment into an uncertain cycle.

• We underwrite the twenty-year implementation cycle, not the five-year capital expenditure cycle. The buildout matters. It is simply not the whole opportunity, and it is the part the crowd has already bought.

The larger and more durable prize is the slow, unglamorous work of ordinary companies rebuilding how their industries operate around this technology. That work is quieter; it is far less crowded, and we believe it is where the real money will be made.

So no, the unwind of Situational Awareness does not make us less optimistic. It makes us more confident of a simple thing: in a gold rush, the most dangerous place to stand is in the middle of the crowd, with borrowed money, next to everyone who read the same memo. The surest returns from any technology come not from speculating on the buildout, but from owning the businesses that use it to become permanently better.

Net Performance Overview [1]

Featured Research [5]

Data Sovereignty

• AI models are becoming cheaper, more capable, and increasingly interchangeable. The lasting enterprise value sits in the workflows, proprietary data, and operating logic that make your business unique.

• The model still matters. But increasingly, the long-term value of enterprise AI resides in the operating layer and the proprietary data that make the model more relevant, personalized, and effective.

• The relevant question is not simply which model performs best overall. It is which model delivers sufficient performance for a specific workflow at the right cost, with the appropriate level of control.

• Explore the Data Sovereignty Series

Strategy Memo: Data Sovereignty Is Becoming a Competitive Requirement (Full Version)

• The technical challenge is turning institutional knowledge into a system that machines can reason across.

• What changed was its access to the structure, relationships, and operating context unique to the enterprise.

• Data sovereignty ensures that value accrues to the enterprise that created it rather than leaking to the model providers powering it.

5. Closing Comments

We welcome your questions, feedback, and perspective: ir@plutus21.com

Thank you for reading.

Onward and Upward,

Hamiz, Richard, Daniyal, Brad, and the Plutus21 Team

[1] Performance results are for informational purposes only and are not investment advice or a solicitation to invest.

All figures are net of applicable management and performance fees, trading costs, and expenses unless stated otherwise. Where exact fees, costs, or expenses cannot be confirmed, reasonable estimates are used and identified as such. Past performance does not guarantee future results, and actual outcomes may differ due to market conditions, portfolio composition, and other factors and risks. Returns for Blockchain Opportunities I (Fund), Blockchain Opportunities II (Fund), and Blockchain Genesis (Fund) reflect standard fees (2% annual management fee and 20% performance fee under a high water mark). 2025 figures are unaudited.

Calculations rely on NAV Fund Administration and are subject to revision. AI Inflection (SMA) and AI Core (SMA) reflect actual client accounts managed at Interactive Brokers (IBKR), net of fees as follows: AI Core (1.25% annual management fee, accrued monthly; 15% performance fee, accrued quarterly) and AI Inflection/Axiom (1.5% annual management fee, accrued monthly; 15% performance fee, accrued quarterly). Figures are unaudited. Calculations rely on IBKR Portfolio Analyst and are subject to revision. 2026 figures are unaudited. Calculations rely on the external multistrat fund’s administrator and are subject to revision. No representation is made that any investor will or is likely to achieve similar results. Annualized returns for periods of less than three years may be misleading and are shown for informational purposes only. Actual results may vary. Performance shown reflects results significantly. All data and sources may be subject to revision.

Performance shown reflects results from separately managed accounts traded using substantially the same strategy before the launch of the Axiom Series and is presented for informational purposes only. Such results are not the performance of the Axiom Series and may not be representative of, or indicative of, the results the Axiom Series will achieve.

[2] YTD = Year-to-Date

[3] ITD = Inception-to-Date

[4] ITD CAGR = Inception-to-Date Compounded Annual Growth Rate. For strategies with less than one year of track record, this figure is not annualized to avoid misleading results.

[5] For informational purposes only; not an offer to buy or sell any security or product. For full disclaimers, click here.