LWAYS OPEN
The 17 July 2026 portfolio actions were based on a central research conclusion: recovering a drawdown does not necessarily require adding more capital to the positions that originally created the portfolio’s growth potential. When concentration becomes excessive and several lagging positions consume capital without sufficient relative strength, the more rational response may be to reduce opportunity cost and redirect part of the portfolio toward independent return drivers.
The portfolio review began with a distinction between company quality and portfolio usefulness.
A company may remain fundamentally credible while no longer representing the best use of capital over a specific investment horizon. This distinction was particularly important in the case of Oracle and Marvell Technology.
Both companies participate in major structural themes. Oracle benefits from cloud infrastructure and enterprise data demand, while Marvell supplies technologies used in networking, data centres and specialised semiconductor applications. Nevertheless, their portfolio positions had suffered deep drawdowns and had not demonstrated the same recovery strength as the strongest holdings.
The question was therefore not whether Oracle or Marvell could eventually recover. The relevant question was whether retaining the full positions offered a better expected outcome than reallocating part of the capital elsewhere.
Our conclusion was that the opportunity cost had become too high.
Why the sales were staged
The reductions were not executed through immediate full liquidation. Staged limit orders were used because both stocks remained capable of short-term rebound movements.
Oracle sales were completed at USD 124.90 and USD 126.90, with a further tranche positioned at USD 129.90. Marvell’s first reduction was completed at USD 188.90, with additional tranches positioned at USD 191.90 and USD 194.90.
This method recognised two competing realities:
- the portfolio needed capital released;
- the stocks could still recover further before the end of July.
Staged selling therefore provided a compromise between risk reduction and continued participation.
Why the Micron warrant was fully closed
The Micron warrant required a different decision because its risk profile was materially different from that of an ordinary share.
A leveraged warrant is affected by:
- movement in the underlying stock;
- remaining time to maturity;
- implied volatility;
- issuer pricing;
- spread;
- market liquidity.
A delayed recovery can therefore still result in losses even when the underlying company eventually performs well. The full sale of 42,100 units removed this asymmetric timing risk.
The exit also prevented the portfolio from depending on a single semiconductor recovery event to compensate for earlier losses.
Why Arista and NVIDIA were not increased
Arista Networks and NVIDIA remained among the highest-quality holdings in the portfolio. However, both already represented major allocations.
The research conclusion was that additional purchases would have improved neither diversification nor resilience. The portfolio already had extensive exposure to:
- AI data-centre spending;
- semiconductor demand;
- high-performance networking;
- hyperscaler capital expenditure.
Increasing Arista or NVIDIA would therefore have added more of the same risk factor.
The reinvestment search was intentionally directed toward companies whose earnings could be supported by different economic forces.
Palo Alto Networks: cybersecurity as an independent technology driver
Palo Alto Networks remains connected to the digital economy, but its principal demand driver is security rather than semiconductor capacity.
AI adoption may strengthen cybersecurity demand because more data, automated systems and cloud workloads require protection. However, Palo Alto’s investment case does not depend solely on AI accelerator sales or semiconductor pricing.
This made PANW a suitable bridge between growth and diversification.
Danaher: quality, recurring revenue and healthcare exposure
Danaher provides exposure to diagnostics, life sciences and research-related products. Its business profile offers recurring demand, high-quality operating assets and a significantly different cycle from semiconductor and networking companies.
The investment case was not based on the expectation of an explosive two-week price increase. Danaher was selected because it could stabilise the overall portfolio while still participating in a technically improving trend.
Dollar Tree: consumer value and turnaround potential
Dollar Tree offered a third, independent return driver.
Its investment case rests on:
- value-oriented consumer spending;
- improving technical momentum;
- operational recovery potential;
- reduced correlation with AI infrastructure stocks.
The company carries execution and consumer-demand risks, but it provides a form of diversification that the existing portfolio lacked.
Research conclusion
The 17 July actions were not an attempt to predict every market movement correctly. They were designed to improve the portfolio’s structure under uncertainty.
The resulting framework can be summarised as follows:
- retain the strongest strategic AI holdings;
- avoid increasing already oversized positions;
- reduce lagging positions during recoveries;
- fully remove time-sensitive leverage where the risk-reward balance has deteriorated;
- deploy capital gradually through limit orders;
- add companies whose earnings depend on different sectors.
This approach does not guarantee that the portfolio will recover by the end of July. It does, however, create a more rational structure from which recovery can occur. The portfolio becomes less dependent on one sector, less exposed to time decay and better positioned to benefit from multiple forms of market leadership.