MacroVoices #540 Adam Parker: Beyond the AI Bubble: Diversifying Portfolios in an Earnings-Driven Market

Adam Parker expects corporate earnings to remain strong while price-earnings multiples contract modestly, producing a choppy but upward-trending US equity market over the next six to twelve months. He argues the market is pricing a distribution of long-dated revenue outcomes, with companies showing rising out-year revenue revisions having outperformed those that do not. He is not pre-positioning for a stagflationary or growth-scare scenario and dismisses interest rate forecasts as a basis for equity positioning, noting that Morgan Stanley's rate strategists were wrong every single year he worked there.

Parker identifies the S&P 500 as having effectively become an AI semiconductor ETF, estimating that approximately 265 of the top 3000 US equities by market cap have meaningful AI revenue, defined as at least 5 percent of revenue or an imminent new product announcement. Every sector in that universe has AI revenue representation except consumer staples, and all major large-cap technology platforms except Apple qualify. He views compute as an above-GDP growing business for several years ahead and considers being bearish on AI equivalent to trying to time a cyclical top, which he regards as extremely difficult.

Parker warns that apparent diversification around AI concentration is often illusory. Power and utility stocks including GE Vernova, Vistra, and Constellation were previously considered anti-correlated to AI semiconductors but have become highly correlated to that basket. Eaton, Caterpillar, and GE Vernova carry approximately 0.9 correlation to AI semis despite sitting in industrial and utility sectors. To genuinely diversify, Parker recommends overweighting energy and healthcare alongside technology as the offensive core of a US equity portfolio. Energy represents roughly 4 percent of the S&P 500, sits at a 40-year low in its correlation to the tech sector, and has above-average estimate achievability, which Parker views as particularly valuable given that the penalty for missing earnings has been far harsher than the reward for beating. Healthcare is priced as though there is a zero percent probability it will be the best-performing sector over the next five years, while Parker estimates the true probability at 30 to 40 percent, and the sector is highly anti-correlated to AI semiconductors.

On stock picking, Parker states that valuation-based approaches such as buying cheap and shorting expensive have generated essentially no money for approximately 15 years. Price-to-earnings, price-to-forward earnings, and price-to-book have no informational value for stock picking over timeframes shorter than three years. By contrast, buying stocks that just became more expensive on price-to-forward earnings has a higher probability of beating estimates than buying cheaper stocks, and the probability of a second earnings beat given a first beat is higher than the unconditional probability, giving momentum informational value. High quality stocks as scored by Trivariate have not beaten lower-quality stocks for six years, and approximately 82 percent of current S&P 500 market cap is classified as top-half quality, meaning a long-only investor could own two-thirds top-half quality names and still be significantly underweight the index.

Parker's simulation work across 25, 50, 75, and 100 stock portfolios run over thousands of rolling one-month periods found that institutional managers running more than 50 or 75 names have seen significantly better performance in recent years, attributing this to drawdowns from concentrated portfolios being too extreme even for 75th percentile stock pickers since 2020. Trivariate used Claude AI to analyze three and a half years of its own research notes and found subsequent one-week predictive value when language was categorized as neutral, bearish, or very bearish, though Parker views AI tools as primarily valuable for data ingestion and more efficient coding rather than as autonomous trading systems.

On market structure, systematic sell triggers for the S&P 500 are estimated around the 7,300 to 7,350 level, with both S&P and Dow futures at the 100th percentile of long positioning on a one-year lookback, reflecting CTA exposure that could become disorderly if those positions begin to rebalance. The US dollar has broken out of a 15-month trade range and is holding well above the 100 level on the DXY, with positioning at extreme long levels, though this is not being treated as a contrarian signal until price action stops confirming the trend. Gold remains in a primary downtrend with key support near 4,000 dollars, and a failure there would point to the 3,600 to 3,700 range. In 30-year bonds, large speculators covered 85,000 contracts of shorts in a single week following a Fed president speech that briefly flipped market consensus toward pricing rate hikes, with commercials simultaneously flipping from net long to net short.