Ideas
Gold bull market has further to run
Gold remains in the middle innings of a structural bull market. Although it will eventually end badly, it has not ended yet: central banks are buying gold in large quantities, developed-world fiscal policy remains unsustainable, retail interest is only now perking up, and US-listed ETF holdings are still below 2022 levels. His three-month tactical quant model remains bullish because the medium-term trend is still upward even though the short-term trend has deteriorated.
Skeptical US tech works in 2026
US tech has been a great trade in 2025 and prior years, but he is skeptical it will be a great trade in 2026. The AI trade has supported capex and consumption, but valuations are stretched, AI capex returns are uncertain, and earnings may be exaggerated by unprecedented compute demand. He is not yet running for the hills; he would wait for a metaverse moment where a large-cap tech company announces huge AI spending and its stock falls, which would signal the end of the AI trade.
Nvidia earnings risk from compute supply
Nvidia's earnings may be exaggerated by unprecedented excess demand for compute. While Nvidia can make a lot of money selling chips now, capex is ongoing, supply is increasing, and eventually more competition will bring prices and earnings down. That makes current P/E ratios a poor signal of how expensive or cheap the stock is.
S&P defensive over 12 months
On the S&P 500 and overall tactical asset allocation, Peter is tactically neutral over the next three months but would be more defensive over a 12-month horizon. He is watching a checklist for turning more bearish, including a metaverse moment in AI and, more importantly, a rise in layoffs; until layoffs increase, he does not expect a recession. He notes S&P 500 price-to-sales valuations are about 50% more expensive than at the 2000 peak, so stocks could fall a lot if risks materialize.
Nuclear trade past expiration date
Nuclear power is a good technology in the current age, but the investment story is not undiscovered. People in the investment community know it well, and some stocks levered to the nuclear/uranium theme are very expensive and some have no revenue yet. He thinks that trade has gone past its expiration date.
Hyperscaler AI monetization bar too high
The investment thesis behind hyperscalers is essentially the AI dream: investors are hoping these companies can monetize huge AI investments. The bar is very high. By the end of the decade, annual capex could approach $1 trillion, requiring roughly $2 trillion of annual revenue to profitably monetize, but the combined trailing 12-month sales of Microsoft, Google, AWS, Oracle, and Meta are only about $1 trillion. Unless AI really takes off, that revenue will not materialize and the stocks are at risk of going down.
Buy cheap defensive staples and healthcare
Within a modestly defensive allocation, he would buy more consumer staples and healthcare. Those sectors have not done particularly well this year and are relatively cheap, making them attractive defensive holdings.
Emerging markets attractive on cheaper valuations
He would also look abroad, where emerging markets are coming up from behind. They are cheaper, have pretty good valuations in many cases, and have pretty good economic fundamentals.
This The David Lin Report video, published October 29, 2025,
features Peter Berezin
discussing GLD, US Tech, NVDA, SPY, URA, AI Hyperscalers, XLV, XLP, EEM.
8 trade ideas extracted by AI with direction and confidence scoring.
Speakers:
Peter Berezin
· Tickers:
GLD,
US Tech,
NVDA,
SPY,
URA,
AI Hyperscalers,
XLV,
XLP,
EEM
