JPMorgan weekly on tech hardware and semiconductors: MU pre-earnings focus on SCAs and ASPs, with AI capex and WFE expectations continuing to rise

Overview

This is a JPMorgan North America tech hardware and semiconductor expert sales weekly note. Its core topics are MU pre-earnings buy-side expectations, pre-quiet-period communication with tech hardware companies, the HDD pricing cycle, AI capex financing, upward revisions to the WFE market size, AAPL cost pass-through, AMD data-center opportunities, and AVGO's competitive position in ASIC/XPU. The report overall shows that AI data-center demand continues to drive the storage, server, networking, HDD, WFE, and custom-chip supply chains, but some stocks are already crowded and short-term trading still faces positioning and valuation pressure.

Core views

The key views are: first, MU is benefiting from improving DRAM and NAND demand, AI-driven ASP upside, and potential SCAs, but valuation has shifted to an EPS-upcycle framework, and the buy side is still hesitant to pay more than 7 to 8 times earnings. Second, the upside for HDD names WDC and STX comes mainly from pricing rather than exabyte growth, and supply constraints reduce the risk of price giveback. Third, the AI capex cycle has shifted from debates over revenue viability to questions about compute scarcity and financing structure; US$5.5 trillion of AI capex before 2030 and US$4.1 trillion of debt financing are now important medium- to long-term themes. Fourth, WFE expectations were raised again, with DRAM, TSMC, INTC, Samsung Foundry, and YMTC as the main incremental demand sources. Fifth, the market is underestimating AVGO's position in GOOGL TPU and multiple customer ASIC/XPU projects.

Analysis framework

The report combines expert sales coverage with buy-side conversations, drawing on MU buy-side earnings surveys, pre-quiet-period discussions with hardware and networking companies, sell-side model revisions, JPM's global technology team's AI capex deep research, positioning monitoring, and single-name views. The emphasis is not on a single financial model, but on using buy-side expectations, management communications, price trends, capital expenditure plans, and positioning changes to judge near-term earnings risk and medium- to long-term supply-chain opportunities.

Methodology notes

• earnings_checkBuy-side pre-earnings survey

  Use buy-side expectations for MU SCAs, ASPs, and shipment coverage ratios to judge what will matter on the earnings call.

  About 75% of surveyed buy-side participants expect MU to announce more SCAs, about 50% expect only qualitative commentary, and about 40% estimate forward bit shipment coverage in the 30% to 50% range.

• positioning_analysisUS Tactical Positioning Monitor

  Measure near-term market positioning risk using hedge funds, retail, ETFs, options, and style-based flows.

  The report says US one-week tactical positioning rebounded from roughly -2z the prior week to about +1z, but overall US positioning remains near the mean, so seasonal gross-leverage pullback after June options expiration should be watched.

• industry_tamAI Capex and WFE TAM upward revisions

  Estimate the AI infrastructure and semiconductor equipment cycle using capex, data-center capacity, debt financing, and equipment demand.

  JPM's global technology team raised total AI capex before 2030 to US$5.5 trillion, and the semiconductor team raised the 2026 WFE market size to US$155 billion and the 2027 figure to US$200 billion.

Asset mapping & comparison

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

• MICRON TECHNOLOGY INC (US.MU)

  Core coverage name and pre-earnings focal point

  Strengths

  Improving DRAM and NAND demand, AI-driven ASP upside, expectations for more SCAs, and potentially strong free-cash-flow generation.

  Weaknesses

  The stock is one of the most crowded longs in tech, and the market has already shifted to an EPS-upcycle valuation framework.

  Comparison

  Compared with a traditional storage cycle, this one could be longer because of 3- to 5-year SCAs and AI demand.

  Risks

  Insufficient disclosure of SCA terms, buy-side concerns about customer default in a down cycle, demand destruction, and unwillingness to pay more than 7 to 8 times earnings.

• APPLE INC (US.AAPL)

  A reference name for storage cost pass-through and consumer-electronics price elasticity

  Strengths

  High iPhone usage frequency, more subsidies, Apple Intelligence could drive a real replacement cycle, and consumers may be more accepting of tech inflation.

  Weaknesses

  Rising storage costs create product gross-margin and pricing pressure.

  Comparison

  The report thinks the Wall Street Journal's cited US$200 to US$300 price increase is unrealistic, and that about a US$50 ASP increase is more consistent with JPM's math.

  Risks

  Price increases could hurt unit sales, consumer price sensitivity may be higher than expected, and the Apple Intelligence replacement-cycle driver may disappoint.

• BROADCOM INC (US.AVGO)

  A core beneficiary of the ASIC/XPU and GOOGL TPU chain

  Strengths

  TPU v9 2nm remains on schedule, the GOOGL-AVGO agreement extends to TPU v11, and AVGO has one of the world's largest pipelines of winning ASIC/XPU designs.

  Weaknesses

  The stock is under short-term pressure from media noise around the Mediatek-related supply chain.

  Comparison

  The report says AVGO is at least 18 months ahead of GOOGL's COT team in advanced chips, packaging design, and IP.

  Risks

  GOOGL's diversified supply strategy, advanced-packaging and design execution risk, and changes in customer project timing.

• ADVANCED MICRO DEVICES INC (US.AMD)

  An opportunity name for agentic AI server CPUs and MI450/Helios data-center GPUs

  Strengths

  The server CPU TAM is expanding, the CPU:GPU attachment ratio is shifting from 1:4 to 1:8 toward 1:1 or even a CPU-heavy mix, and full-rack MI450 systems are already running in customer labs.

  Weaknesses

  It still needs to prove that the GPU business can turn into durable share rather than a one-cycle win.

  Comparison

  The report stresses that the market originally expected CY27 data-center AI revenue in the tens of billions of dollars, but it has become unclear whether CPUs or DC GPUs will account for the larger share.

  Risks

  ROCm ecosystem maturity, customer ramp timing, capacity coordination, and MI450 production ramp.

• NVIDIA CORP (US.NVDA)

  A core beneficiary of AI GPU infrastructure and a reference point for the financing chain

  Strengths

  JPM's AI capex report rates NVDA OW, and the table shows NVIDIA GPU shipments rising from 1.7 million units in 2023 to 9.9 million units in 2027E.

  Weaknesses

  GPU share of AI accelerators is expected to fall from 68% in 2025 to 47% in 2027E as ASIC/XPU share rises.

  Comparison

  GPUs remain the largest ecosystem, but ASIC/XPU is becoming the incremental financing and deployment theme.

  Risks

  Custom-chip substitution, capex-financing constraints, and customers pushing ahead with in-house chips.

• WDC / STX

  Beneficiaries of the HDD pricing cycle

  Strengths

  Supply constraints support prices, pricing rather than exabyte growth is the main driver of the upward revision, and about 90% sequential incremental gross margin remains at high levels.

  Weaknesses

  Some long trades have become crowded, and the short-term performance gap between WDC and STX is hard to fully explain.

  Comparison

  WDC has lagged year to date but has recently caught up; STX was previously better owned and more crowded.

  Risks

  Price expectations are too high, marginal buying weakens after short covering, and HAMR 2.0 gross-margin expectations are overstated.

• WFE / Semiconductor capital equipment

  Upward-revision theme for the semiconductor equipment cycle

  Strengths

  JPM raised its 2026 and 2027 WFE growth forecasts to 28% and 29%, respectively, with DRAM, TSMC, INTC, Samsung Foundry, and YMTC contributing the main increments.

  Weaknesses

  The report says buy-side expectations may still be ahead of JPM's revised numbers.

  Comparison

  WFE expansion is driven jointly by logic, foundry, DRAM, and NAND spending.

  Risks

  Customer capex cadence slows, AI demand fails to materialize as expected, or the storage price cycle reverses.

Key data

• MU buy-side expectations for additional SCAsAbout 75% expect more SCAsAbout 10% think the company may wait, citing near-term pricing leverage or ongoing negotiations.
• Expected forward bit shipment coverage for MUAbout 30% to 50% is the main range estimated by the buy sideAbout 50% expect the company to confirm SCAs but not disclose a specific percentage.
• DRAM ASP expectation44.6%The report says this is broadly in line with market consensus.
• NAND ASP expectation45.9%The report says this is broadly in line with market consensus.
• AI capex total estimateUS$5.5 trillion by 2030Raised from US$5.1 trillion in November last year.
• AI data-center capacity growth base case138 GWFrom JPM's global technology team's AI capex deep-dive report.
• Debt financing componentUS$4.1 trillionThe disclosed projects have an average LTC above 85%, roughly equivalent to 60% LTV.
• GPU/ASIC chip financing opportunityMore than US$3 trillion over the next five yearsThe report argues that chip financing is the next capital-markets story.
• 2026 WFE market-size forecastUS$155 billion, up 28% year over yearThe prior 2026 growth forecast was 21%.
• 2027 WFE market-size forecastUS$200 billion, up 29% year over yearThe prior 2027 growth forecast was 18%.
• Total AI accelerator shipments23.3 million units in 2027EThe table shows 3.3 million units in 2023 and 16.3 million units in 2026E.
• ASIC/XPU share of AI accelerator shipments53% in 2027EThe table shows this share rising from 32% in 2025 to 53% in 2027E.
• Estimated ASP increase needed for AAPLAbout US$50Used to offset more than US$100 of year-over-year storage cost pressure and the remaining gap after other components and vertical integration savings.
• AMD server CPU TAMMore than US$120 billion before 2030The report says agentic AI structurally expands the server CPU opportunity.
• AMD revenue opportunity per GWAbout US$15 billion to US$20 billionThe report believes this could skew higher over time.
• AVGO AI revenue growth expectation2x to 2.5x growth in 2027, then another doubling in 2028Based on the report's summary of Harlan Sur's view.

Impact & implications

The investment implication is that AI data-center buildout continues to expand the semiconductor cycle from GPUs into CPUs, ASIC/XPU, storage, HDDs, networking, WFE, and financing tools. MU and HDD companies benefit from rising prices and supply constraints, WFE companies benefit from upward revisions to DRAM and advanced-node capex, and AVGO and AMD benefit from custom-chip, TPU, and data-center GPU/CPU opportunities. At the same time, the market has already fully embraced some AI and storage longs, so investors need to distinguish genuine earnings upgrades from low-EPS winners driven only by sentiment, and remain alert to crowded trades and seasonal leverage unwind.

Risks

• MU and the storage chain are crowded long trades, and if SCAs disclosure or ASP guidance falls short of expectations, valuation downside could be significant.
• SCAs may lengthen the storage cycle, but the buy side still worries that customers may abandon commitments in a down cycle.
• AI capex financing needs are enormous, and there is a maturity mismatch between a five-year chip life and long-dated IG capital.
• Hedge-fund gross leverage rose unusually ahead of June options expiration, and history suggests a notable pullback may follow.
• If AAPL price increases exceed what consumers can absorb, iPhone sales may be hurt.
• Some hardware companies face storage-cost, tariff, product-mix, and gross-margin pressure.
• WFE and AI-infrastructure expectations have already been revised materially higher, so related stocks could come under pressure if demand or financing slows.

What to watch

• Whether MU's earnings call announces more SCAs, and whether it discloses coverage ratios, pricing structure, customer type, and cash commitments.
• Whether MU's comments on AI applications, ASPs, gross margin, and long-term free cash flow support the current upcycle valuation.
• Whether HDD names WDC and STX continue to maintain price improvement and high incremental gross margins.
• Whether MSFT issues large IG bonds in the second half of 2026 to fill its FY27-28 funding gap.
• Whether ASIC/XPU continues to gain share in the AI accelerator shipment mix.
• Whether JPM's revised 2026 and 2027 WFE forecasts are validated by equipment orders and customer capex.
• Whether AAPL actually raises iPhone prices, and whether the increase is close to the roughly US$50 estimate.
• Progress on AVGO's TPU v9 2nm program, advanced substrate facilities, and multiple ASIC/XPU customer projects.
• Whether AMD MI450/Helios initial shipments in early 3Q and a meaningful 4Q ramp actually materialize.