Why Macro is “Pretty Risk-On” for Equities | Tian Yang of Variant Perception

Variant Perception's macro risk indicators on a three-to-six-month horizon show a risk-on regime for equities. US real GDP growth is running close to two percent, inflation is elevated but well below 2022 levels, policy tightening is not a major problem, and liquidity on their models remains fine given the long lead lag before tightening hits markets. Credit spreads adjusted for sovereign risk premiums are in a normal range, and equity breadth is broadening, with the value line arithmetic index making higher lows, which Tian Yang says is inconsistent with net money leaving equity markets.

A market top checklist covering behavioral, corporate, and economic factors shows mostly amber warnings rather than full red alerts. The two consistent signals at every generational top are monetary tightening sustained for six to nine months and breadth narrowing while concept stocks keep rising, a pattern seen in the nifty 50 era, dot-com, and 1929. Red alert signals currently present include circular financing around Nvidia and retail excesses in markets like Korea, but the missing red signals are economic slowdown and deteriorating market internals. Insider buying has flooded in globally, which Yang says is not typical behavior seen at an imminent top.

LPPL, or log periodic power law, is a bubble and crash detection tool with roughly a one-month useful horizon. LPPL exhaustion signals fired in early June on semiconductor and AI-related names ahead of the correction. The current semiconductor price action resembles gold and Bitcoin tops, where fundamentals remain intact but price exhausts, and when a narrative tops it is typically hard for the asset to regain its highs. LPPL bottom signals have appeared on intraday timeframes but not yet on daily frequency, and semiconductor positioning has moved from extremely crowded long to uncrowded but not yet extremely short. Yang views AI-related stocks as likely dead money for a while rather than a crash, analogous to Bitcoin since early 2025. The agentic AI narrative is exhausted and the trade needs a new catalyst such as real-world modeling before it can return to prior highs. Meta raised capex guidance only slightly and Microsoft held it flat, yet semiconductors had one of their biggest rally days in response, which Yang reads as a sign markets topped before fundamentals changed. Variant Perception is keeping some semiconductor exposure but rotating to energy, financials, and health care in July.

Yang argues the AI capex cycle differs from railway mania or dot-com because government involvement backstops private spending through a sovereignty lens, and the US will step in to finance AI-related capex if the private sector exhausts capacity. Using Perry Merling's hierarchy of money framework, the first sign of cycle risk will be when counterparties begin demanding cash settlement instead of accepting compute or equity-like instruments. Mega IPOs like SpaceX signal that demand for cash settlement is rising, and investors must sell Nvidia shares to fund purchases of SpaceX and future listings like Anthropic. SpaceX stock has declined nearly every day since its IPO, the space business loses money while Starlink is profitable, and heavy spending on a new data center business obscures any valuation floor. A potential OpenAI IPO around end of 2025 or Q1 2027 could mark the symbolic moment that lines up with a generational top if macro conditions deteriorate by then, drawing a parallel to the AOL Time Warner deal at the dot-com peak.

Variant Perception is running a large overweight in energy initiated around May. Oil prices do not need to rise to justify energy fundamentals because crack spreads are very wide, and integrators and refiners should generate strong free cash flow as long as they return capital. Energy also serves as a natural ballast against prolonged inflation or supply shocks that keep long-end yields elevated. Yang warns that traditional defensive positioning in nominal bonds and consumer staples will not work in an inflation or supply-shock regime. WTI would need to average approximately 120 dollars for the rest of the year to tip the US into recession, while Brent averaging between 110 and 120 dollars would put Europe at the tipping point. Yang predicts oil is in a structurally bullish market over the next two years because China needs to replenish petroleum reserves drawn down to buffer the recent supply shock.

China's structural problem is that industrial policy has for ten to fifteen years funneled credit toward manufacturing and technology at the expense of the household sector, and falling real estate values have further destroyed household wealth. The involution problem in Chinese tech, where Alibaba, Tencent, Meituan, and Pinduoduo destroy each other's profitability through intense price competition, cannot be resolved because cutting excess supply would require mass layoffs the government will not permit. China's AI strategy mirrors its manufacturing playbook, commercializing rapidly and compressing prices, and its commitment to open-source models is accelerating price compression across the industry globally. Yang expects the AI market to settle into three tiers: a mass-market open-source tier at near-zero marginal cost, a mid-tier for small business adoption, and a high-end enterprise tier accessible only to governments and large organizations, with the middle layer of horizontal SaaS and consulting most at risk.

Japan's macro situation is described as untenable, with Tokyo real estate growing double digits, Shunto wage rounds at five percent or more for three consecutive years, and base effects from prior yen depreciation now mechanically pushing inflation higher.