AI trading needs to shift from semiconductor one-way leadership to a more balanced value allocation

Overview

This report is part of JPMorgan Global Markets Strategy's Flows & Liquidity series and discusses the need for rotation within AI trading, while also updating U.S. liquidity, cross-asset flows, positioning, ETF flows, short interest, and crypto market risks. The report notes that since last September, AI chip and storage-related semiconductor stocks have almost continuously outperformed hyperscalers, and this gap does not appear sustainable over the long term. JPMorgan's house view leans toward a more constructive convergence path, where after AI monetization and improvements in revenue and profits, hyperscalers, AI model providers, and end users can take a larger share of incremental AI value.

Core views

There are three core views. First, AI trading needs to shift from semiconductor one-sided strength to more balanced value allocation; otherwise, if semiconductor profitability is built on client pressure, it could dampen the willingness of hyperscalers and AI model providers to spend capex and eventually become a headwind for semiconductor demand. Second, the market has already started to express concerns about hyperscalers through short interest and credit spreads, especially with rising short interest in hyperscalers and some semiconductor stocks in May-June, while AI adoption beneficiaries have not yet seen comparable pressure. Third, U.S. money creation has been faster than nominal GDP, still providing liquidity support to U.S. financial assets, particularly U.S. equities, but non-bank investors have lower cash allocations, making the market more fragile when negative shocks hit.

Analysis framework

The report combines relative equity performance, Bloomberg bottom-up analyst capex consensus, JPMorgan's own AI capex forecast, market cap and sales share, equity and credit market performance, short interest, AI compute and LLM token pricing, U.S. M2 proxy indicators, the balance sheets of banks and the Federal Reserve, and cross-asset positioning monitoring to assess the AI trade and liquidity backdrop. Its analytical focus is not to make a single-stock recommendation, but to identify the marginal impact on asset prices from flows, positioning, and macro liquidity.

Methodology notes

• Cross-asset positioningCross Asset Positioning Monitor

  Cross-asset positioning percentile

  This framework aggregates indicators such as futures speculative positioning, momentum signals, mutual fund beta, risk parity positioning, hedge fund beta, client surveys, global non-bank investor asset allocation, and short interest into a current percentile from 0 to 1, used to gauge crowding by asset class.

• AI capexHyperscaler capex consensus comparison

  Comparison of consensus capex and JPMorgan forecast

  The report uses Bloomberg-compiled bottom-up analyst consensus to track hyperscaler capex paths and compares it with JPMorgan analysts' 2027 forecast of $1,150bn to assess semiconductor demand risk.

• Market sentimentShort interest monitor

  Short interest as a share of float

  The report compares short interest for hyperscalers, semiconductors excluding storage, JPM AI adoption beneficiaries basket, and JPM AI vulnerable basket as a proxy for the market's degree of concern at different stages of the AI trade.

• LiquidityUS M2 proxy

  U.S. money supply proxy indicator

  The report uses the sum of U.S. commercial bank deposits and U.S. money market fund AUM as a proxy for U.S. M2 stock, and evaluates how the pace of money creation supports financial assets.

• AI monetizationAI compute and LLM token pricing monitor

  Compute pricing and token pricing

  The report tracks NVIDIA Hopper GPU lease prices and an LLM token spending index, arguing that compute pricing is key to whether hyperscalers can maintain or improve margins from AI capex.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• AI semiconductor stocks

  Leading beneficiaries within AI trading, but the sustained outperformance versus hyperscalers is viewed as difficult to maintain over the long term.

  Strengths

  They benefit from AI chip, storage, and data center capex demand, and hyperscaler capex growth remained strong in 2026.

  Weaknesses

  If client capex clearly slows from 2027, both semiconductor demand and valuations could come under pressure.

  Comparison

  Their relative performance versus hyperscalers has been strongly strong, but this divergence has already raised market questions about sustainability.

  Risks

  Capex downgrades, AI monetization falling short of expectations, rising short interest, and transmission of pressure from credit markets.

• Hyperscalers

  The dominant demand-side of AI capex and a key lever for whether AI value allocation can rebalance.

  Strengths

  If AI monetization improves, revenue, profits, and valuation could catch up with semiconductors.

  Weaknesses

  Over the past year, stock performance was broadly flat, valuation compression appeared, and credit spreads widened relative to semiconductors, pushing up capital costs.

  Comparison

  Significantly behind semiconductors, hyperscalers are a potential beneficiary of AI rotation and also the source of pressure in negative scenarios.

  Risks

  A sharp slowdown in capex growth after 2027, rising financing costs, and insufficient improvement in compute pricing.

• U.S. equities

  A core risk asset supported by U.S. money creation and liquidity expansion.

  Strengths

  The U.S. M2 proxy is expanding faster than nominal GDP and has historically supported financial assets.

  Weaknesses

  Global non-bank investor cash allocation is relatively low, so major negative shocks could force rapid cash rebuilding.

  Comparison

  Relative to bonds and cash, liquidity expansion directly supports risk assets such as U.S. equities.

  Risks

  Macro shocks, policy repricing, geopolitical risks, and inflation or tariff shocks raising cash demand.

• Bitcoin and crypto assets

  Affected by MicroStrategy's capital-structure policy and Bitcoin trading behavior.

  Strengths

  Weak sentiment can form a contrarian bullish signal; if MicroStrategy raises USD reserves and U.S. market-structure legislation is approved, conditions could improve in the second half.

  Weaknesses

  MicroStrategy's sale of BTC to pay dividends shifts the market from a one-way buy expectation to two-way flow risk.

  Comparison

  Compared with depreciating-hedge assets like gold, Bitcoin is additionally exposed to policy changes by large holders.

  Risks

  MicroStrategy continuing BTC sales, insufficient reserve coverage, legislative delays, and higher crypto-market volatility.

Key data

• U.S. money creationabout $1.6tr in 2025, about $1.8tr annualized year-to-date in 2026The report sees U.S. money creation accelerating from $1.6tr in 2025 to an annualized pace of around $1.8tr in 2026.
• U.S. money supply expansion in 2024about $1.2trThe U.S. M2 proxy, measured by commercial bank deposits plus money market fund AUM, rose strongly in 2024.
• Cumulative U.S. money supply increase from May 2023 to end-2024about $2.2tr or 9.6%The report says cumulative additions from May 2023 to end-2024 were about $2.2tr.
• Hyperscaler 2027 capex outlookJPMorgan forecast $1,150bn, consensus forecast $925bnIf the consensus-embedded deceleration in capex from 2027 onward is realized, semiconductors and the AI trade may come under pressure.
• MicroStrategy Bitcoin holdings impactholds about 4% of Bitcoin supply; bought $13.7bn BTC year-to-dateThe report says its year-to-date purchases amount to about 70% of JPMorgan's estimated total digital asset flow.
• MicroStrategy USD reservescurrently about $2.55bn, covering about 17 months of dividendsJPMorgan believes investors would be more reassured if reserves covered 24 to 36 months, reducing the need for near-term Bitcoin sales.
• Cross-asset positioning percentileequities 0.72, government bonds 0.73, USD 0.81, European equities 0.90, Japanese equities 0.81, gold 0.40, Bitcoin 0.51As of June 30, 2026, higher percentiles indicate a position nearer the high end of its historical range.

Impact & implications

For investment implications, the risk in AI trading is not that the AI theme disappears, but that after value becomes too concentrated in semiconductors, it can trigger reverse effects via client capex, funding costs, and sentiment. The constructive path requires hyperscalers and AI model providers to prove capex returns through better compute pricing, token pricing, and commercialization revenue; the negative path is capex cuts leading to lower semiconductor demand expectations and dragging down AI-related equities and credit assets. U.S. liquidity expansion still supports risk assets, but low cash allocations raise vulnerability to external shocks. In crypto assets, MicroStrategy's shift from a one-way buyer to a potential seller raises Bitcoin market volatility and uncertainty.

Risks

• Hyperscaler capex slows materially from 2027 onward, leading to downward revisions in semiconductor demand expectations.
• The valuation and performance divergence between semiconductors and hyperscalers continues to widen, creating technical and sentiment issues.
• AI compute pricing or LLM token pricing improvements are not sustained, weakening hyperscalers' returns on AI capex.
• Low cash allocation makes risk assets more prone to de-risking and cash rebuilding pressure under negative shocks.
• MicroStrategy either continues selling BTC or fails to raise USD reserves to 24- to 36-month coverage, prolonging two-way flow risk in Bitcoin.
• If U.S. market-structure legislation remains stalled in Congress, it could dampen crypto recovery in the second half.

What to watch

• Whether hyperscaler capex guidance for 2027 and beyond and Bloomberg consensus continue to be cut.
• How the gap between JPMorgan's 2027 $1,150bn hyperscaler capex forecast and market consensus of $925bn converges.
• AI compute lease pricing, especially NVIDIA Hopper H100 and H200 GPU price trends.
• LLM token pricing and commercialization revenue changes of AI model providers.
• Changes in short interest for hyperscalers, semiconductors, AI adoption beneficiaries, and AI vulnerable companies.
• Expansion in U.S. commercial bank balance sheets, changes in the Federal Reserve balance sheet, and the growth rate of the U.S. M2 proxy.
• The number of months covered by MicroStrategy's USD reserves, whether it continues selling BTC, and whether it raises reserves through common-equity issuance.
• Progress on U.S. crypto market-structure legislation.