H1 2026 rewarded investors who stayed with the AI infrastructure trend, but only if they could tolerate sharp drawdowns. Memory, networking and compute leaders outperformed, while crowded positions often experienced steep corrections before recovering. Timing mattered far less than disciplined risk management.

For H2, I would not abandon AI hardware, but I would be more selective. As valuations become richer, I would favour companies with durable earnings growth, strong cash flow and pricing power over purely momentum-driven names. A balanced approach that keeps core AI exposure while gradually adding quality value sectors can help reduce portfolio volatility if market leadership broadens.

The key question is no longer whether AI spending continues, but which companies can convert that spending into sustainable profits over the next several years, rather than just the next earnings season.