Global technology and AI equity de-rating: J.P. Morgan sees advanced tech de-rating creating renewed opportunities, led by semiconductors

Overview

This global equity-strategy report examines whether the summer technology and AI selloff has gone far enough to create opportunities. J.P. Morgan finds that positioning and valuations have reset materially, but differentiates sharply between semiconductors, where earnings and capex fundamentals remain strong, and software or AI-vulnerable businesses, where structural profitability concerns persist.

Core views

The broad technology and AI ecosystem stalled for roughly three months from June through the prior week, after the early-year rally ran into stretched positioning and concerns about the durability and monetization of AI capital expenditure. J.P. Morgan argues that the reset has made positioning cleaner, while earnings strength, intact capex plans and emerging monetization cases support renewed engagement with the group. The institution does not expect technology and AI to return to their former extreme market leadership, but believes there remain substantial opportunities within the complex. The report retains its preference for semiconductors over software. Semiconductor positioning has been unwound and the sector’s pullback is described as healthy because fundamentals remain robust: industry demand-supply balance is not expected before 2028, pricing is expected to keep rising into 2027, and company commentary indicates that demand is still accelerating. AI economics still support compute purchases, model advances are accelerating, and hyperscalers have room to fund strong capex growth over the next two years through operating cash flow, debt and equity funding. Since the June peak in Semis versus Software, semiconductor 12-month forward EPS rose 30%, whereas software saw virtually no earnings uplift; the report sees the resulting gap between relative price performance and earnings delivery as supportive of re-entering the Semis-versus-Software pair trade. J.P. Morgan does not advocate a standalone short in software because severe de-rating has made the group historically cheap and can support tactical rebounds. However, it continues to expect software, business services and media businesses exposed to AI cannibalization to lag other parts of the AI ecosystem. Lower model costs reduce barriers to entry for software developers and in-house development, while intensifying competition and persistent questions over profitability weigh on the longer-term outlook. Cybersecurity is a relative exception within software: its shares have rallied and some valuations are rich, but the report believes AI-related security threats should keep the theme relevant. The analysis also identifies changing AI infrastructure beneficiaries. Software-versus-semiconductor performance surged 12% in one day after a call to slow the AI frontier, illustrating sensitivity to frontier-model headlines. Yet J.P. Morgan doubts that material slowing will ultimately occur because the race remains highly competitive. Agentic AI should raise total compute needs while shifting workloads toward more CPUs: conventional LLM infrastructure has CPU/GPU ratios of roughly 1:4 to 1:8, compared with a potential 1:1 or CPU-heavier mix for agentic systems due to orchestration, data movement, tool use and state management. The report highlights semiconductor production equipment as a capex beneficiary, naming ASML as the key European SPE exposure and noting its inexpensive PEG valuation despite investor expectations of earnings downgrades. Agentic AI also creates risks for consumer-facing platforms and European companies that rely on consumer inertia, search friction, switching costs or traffic control. J.P. Morgan’s vulnerable baskets include travel, discovery and advertising, marketplaces, fintech and insurance platforms. If agentic AI becomes the default consumer interface, reduced site traffic and leads could pressure margins. These baskets sold off sharply on the relevant announcement, but the report stresses that their cheap valuations do not resolve the underlying risk to profitability. For Mag-7 and hyperscalers, the report views valuation compression as partly justified but potentially excessive. Mag-7 is flat year to date and trades at 10-year valuation lows; the hyperscaler index is down 8% year to date and its relative forward P/E is more than one standard deviation cheap. Higher AI spending is shifting business models from asset-light to asset-heavy, causing free cash flow to turn negative and increasing debt and equity issuance. The combined free cash flow of Amazon, Microsoft, Alphabet, Meta and Oracle is shown falling from $233.4bn in 2024 to $10.6bn in 2026 and negative $96.4bn in 2027. Token prices have declined sharply, creating monetization concerns, but the report says surging consumption has more than offset lower unit prices so far, total spend is rising and monetization cases are building. It therefore expects solid earnings growth to support hyperscaler performance even if further de-rating absorbs part of that benefit. At the broader market level, J.P. Morgan expects equities to resume advancing when oil and rate volatility fades and September seasonality passes, with Q3 reporting season expected to be robust. IT earnings growth is estimated at 30.8% for the S&P 500 and 35.2% for the STOXX 600 in Q3 2026. Technology’s high index weights—39% in the US, 44% in EM, 19% in Japan and 9% in Europe—mean improved technology trading would help equity markets. Nevertheless, the report argues that a market advance does not require AI to outperform: during the June-July momentum unwind, Korean equities and semiconductor indices fell sharply while broad equities held up and made fresh highs in August. Its broader allocation remains overweight equities, EM and the Eurozone, with semiconductors among favored sectors.

Analysis framework

J.P. Morgan combines positioning and relative-price analysis with forward EPS revisions, forward P/E and PEG valuation comparisons, sector demand-supply evidence, company demand commentary and capital-expenditure funding analysis. It then links technology developments to regional equity exposure, sector allocation and the broader outlook for rates, oil volatility and Q3 earnings.

Methodology notes

• Industry AnalysisSupply-demand framework

  Semiconductor demand-supply and pricing analysis

  The report assesses semiconductor demand, capacity balance and pricing to argue that industry fundamentals remain favorable, with balance unlikely before 2028 and pricing expected to rise into 2027.

• Valuation methodsP/E and PEG Valuation

  Forward P/E and PEG relative valuation

  The report compares forward valuation multiples with historical ranges and earnings expectations to identify advanced de-rating in software, Mag-7, hyperscalers and ASML.

• Corporate Fundamentals and FinanceFree cash flow analysis

  Hyperscaler free-cash-flow and funding analysis

  The report uses projected free cash flow and debt/equity issuance to show how AI capex changes hyperscalers’ financial profiles and helps explain part of their valuation compression.

• Quantitative, Factor, and Portfolio TheoryStyle factor analysis

  Relative performance and earnings-revision comparison between semiconductors and software

  The report compares sector price, EPS and valuation moves to support a relative Semis-versus-Software preference rather than a directional software short.

Asset mapping & comparison

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

• Semiconductors

  Preferred sector exposure within the AI ecosystem

  Strengths

  Strong earnings revisions, healthy demand-supply fundamentals, rising expected pricing and sustained AI capex.

  Weaknesses

  Recent performance was hurt by crowded positioning and questions about compute demand durability.

  Comparison

  Preferred over Software in J.P. Morgan's Semis-versus-Software pair-trade framework.

  Risks

  Lower compute requirements, regulation, China competition, lower-grade-chip demand and hyperscaler funding constraints.

• Software

  Structurally challenged AI-cannibalization exposure

  Strengths

  Record-cheap valuations and normalized short positioning may support tactical rebounds.

  Weaknesses

  Limited earnings upgrades, greater competition and continuing uncertainty over long-term profitability.

  Comparison

  Less favored than Semiconductors despite dramatic de-rating.

  Risks

  AI-enabled in-house development and lower model costs may undermine existing business and revenue models.

• Hyperscalers

  AI capex funders and compute-demand beneficiaries

  Strengths

  Solid earnings growth, rising AI consumption and building monetization cases.

  Weaknesses

  AI capex has reduced free cash flow and increased reliance on debt and equity financing.

  Comparison

  Relative forward P/E levels are more than one standard deviation cheap despite a shift toward asset-heavy models.

  Risks

  Further multiple compression, model commoditization, monetization shortfalls and financing pressure.

• ASML

  European semiconductor-production-equipment beneficiary of higher capex

  Strengths

  J.P. Morgan sector analysts expect solid group fundamentals and identify ASML as the key European SPE exposure.

  Weaknesses

  Its valuation reflects investor expectations of earnings downgrades.

  Comparison

  The report describes ASML as cheap on PEG relative to its fundamental outlook.

  Risks

  Earnings downgrades if semiconductor spending or demand disappoints.

Key data

• Semiconductors 12-month forward EPS since June peak+30%Forward earnings continued to rise while the relative price reversal unfolded.
• Software 12-month forward EPS since June peak+3%The report characterizes this as almost no earnings uplift.
• Mag-7 valuation10-year lowsThe group has de-rated to nearly one standard deviation cheap on a relative forward P/E basis.
• Hyperscaler index performance-8% YTDThe index includes Amazon, Alphabet, Microsoft, Meta and Oracle.
• Hyperscaler combined free cash flow$10.6bn in 2026; -$96.4bn in 2027Compared with $233.4bn in 2024, reflecting the effect of rising AI capital expenditure.
• Q3 2026 estimated IT earnings growth30.8% for S&P 500; 35.2% for STOXX 600The report cites robust earnings as support for its broader equity outlook.

Impact & implications

J.P. Morgan sees a more selective post-de-rating technology opportunity set: semiconductors and AI infrastructure are preferred because earnings revisions and capex demand remain strong, while software and AI-vulnerable businesses may rebound tactically but face longer-term competitive and profitability pressure. Improved technology trading would support major equity indices, particularly Korea and emerging markets, but is not considered essential to a broader equity-market advance.

Risks

• AI compute requirements could be lower than expected as models become more efficient or memory intensity falls.
• Regulatory risks, China competition and customer migration to lower-grade chips could weaken semiconductor demand.
• Hyperscaler capex could pressure free cash flow, require greater financing and face monetization or model-moat concerns.
• AI adoption could reduce traffic, leads and margins for consumer platforms dependent on direct user interaction.
• Higher oil prices, bond yields and escalating geopolitical uncertainty could delay the expected broader equity recovery.

What to watch

• Q3 earnings results and the direction of EPS revisions, particularly for technology and semiconductors.
• Hyperscaler capex plans, funding sources, token consumption and evidence that monetization offsets lower token prices.
• Semiconductor demand commentary, supply-demand conditions and expected pricing through 2027.
• Oil and rate volatility, including the path of US Treasury yields and policy expectations.
• The impact of agentic AI on CPU demand and on traffic-dependent consumer, marketplace, fintech and insurance platforms.