The AI Trade – Gold Debate

The AI trade looks similar to precious metals.

• Leadership in the AI trade has passed from the main beneficiaries to the second tier and the adjacencies, making the investment look very much like precious metals, indicating that we are moving into the second half of the investment cycle.
• The precious metals investment cycle is now reasonably well established, where a large rally in the gold price kicks everything off, followed by the miners and then market leadership passes to silver for the second half of the cycle.
• I became interested in precious metals in 2020 when an unprecedented amount of fiat currency was printed to pay for the COVID-19 pandemic.
• This is similar to what happened in the 1970s, which also triggered a bull market in precious metals as savers became concerned with the preservation of purchasing power in fiat instruments.
• In these sorts of cycles, the commodity rallies first, followed by the highest quality names that are exposed to that commodity and in the second half, leadership moves to the more speculative end of the spectrum, such as the junior miners and the explorers that don’t have any of the commodity but claim to be able to find it.
• Precious metals are very similar, with the added complication that gold usually rallies first and then, about halfway through, leadership passes to silver.
• That is not to say that gold stagnates; it doesn’t, it merely goes up much more slowly than the more speculative end of the investment spectrum, of which silver is at the apex.
• Silver is one of the most volatile, unpredictable and impossible to measure commodities in the investment spectrum, meaning that while it is a monetary metal, it should be seen as a speculative investment rather than something solid and reliable.
• What is happening in the AI trade right now looks an awful lot like precious metals, as the gold-plated investment in AI (Nvidia) is now stagnating, while the second in line are making new highs almost every day.
• Memory is one of the most cyclical and volatile segments of the technology industry, and it has now taken over performance leadership in a very similar way that silver does in the precious metals cycle.
• We are also seeing performance spread into the adjacencies, such as AI at the edge, where MediaTek has doubled this year, and the market has just remembered that Qualcomm exists.
• The market’s attention has also passed from chip supply to electricity supply and CPUs, as these are rapidly becoming a bottleneck to compute expansion and in energy’s case, a focus of protest and dissent against datacentres.
• The net result is that while many commentators are stating that this is just the beginning, the market is telling us that the cycle has moved into the second half of its upswing.
• Given that the first half lasted for 3 years, I am not about to start calling the top, but it does appear to be coming into view.
• Memory is sold out for 2026, and unless there is a large exogenous shock, by the time we get to 2027, I suspect that they will be sold out for 2027 as well.
• Furthermore, their unprecedented market power is allowing them to start demanding 2-year commitments, which won’t stop memory from being cyclical, but it will delay and potentially smooth it out somewhat.
• Q1 2026 saw capex plans increase yet again, with first-time commentary of even more in 2027, but still with little attention being paid to how one will earn a return from the spending binge.
• This is why I continue to think that the place to be is in the companies that make money by supplying the equipment for the datacentres (picks and shovels), almost all of whom are making money hand over fist at the moment.
• At the tip of the spear, I would only consider two companies, which are Anthropic and Nebius, as both appear to have found a way to earn more money per GW of deployed capacity and therefore are in a good position to actually start making money.
• I don’t own either of them at the moment as I remain fully invested in Samsung, Qualcomm and nuclear power, which are now coming to the fore.
• I am also looking for AI to lift the fortunes of ServiceNow and Adobe, both of which have been hammered by AI fears but are in a good position to thrive from it rather than die.