Memory price inflation is reshaping the hardware chain: raising pressure on cloud capex and lowering 2026 global PC shipment forecasts

Overview

This report focuses on the impact of continued memory price increases on the global technology hardware supply chain. Based on its latest memory price forecasts, server BOM calculations, hyperscaler capital expenditure models, PC shipment forecasts, and UBS Evidence Lab consumer surveys, UBS believes memory inflation is simultaneously driving up AI/general server costs and suppressing end-demand for PCs. The report's core conclusion is that AI and cloud capital expenditures still face upside revision pressure, but PC shipment volumes need to be revised further downward, leading to significant divergence within the hardware sector.

Core views

First, the magnitude of DRAM/NAND price increases has expanded further versus prior forecasts, with UBS estimating per-Gb prices will rise 458%/321% in 2025-2027. Second, memory costs will become a major driver of upward revisions to hyperscaler capital expenditures, with memory costs potentially contributing about 66% of capex growth in 2026. Third, the PC market in 2026 is expected to show a pattern of front-loaded purchasing in the first quarter, weakening from the second quarter, and a more pronounced slowdown in the second half. Fourth, PC brands can temporarily cushion pressure through price hikes, product mix upgrades, and low-cost inventory, but as inventory is depleted and prices continue to rise, risks to demand and margins will increase. Fifth, investment allocation should tilt more toward AI servers, semiconductors, and component supply chains rather than traditional PC brands.

Analysis framework

The report uses a combination of top-down and bottom-up methods: updating DRAM, NAND, and HBM price paths using UBS memory price forecasts; estimating memory cost pass-through using server and PC BOMs; assessing funding pressure using a hyperscaler capital expenditure and operating cash flow framework; revising 2026-2027 demand using a global PC shipment model; and validating replacement cycles, budgets, and AI PC interest through the UBS Evidence Lab survey of 1,500 PC users in the US and China.

Methodology notes

• Cost pass-through analysisBOM cost breakdown

  Rising share of memory costs

  Measures the impact of DRAM, NAND, SSD, and HBM price increases on total system cost, selling prices, and margins through server and PC bill-of-materials calculations.

• Capital expenditure modelHyperscaler capex model

  Incremental contribution of memory spending to cloud capex

  Combines memory configurations, unit prices, and shipment volumes for AI servers and general servers to estimate hyperscalers' incremental memory spending in 2026-2027.

• Demand forecastPC unit forecast revision

  High memory prices suppress PC shipments

  Revises global PC shipment forecasts based on price increases, front-loaded purchases, enterprise IT budget drawdown, the Windows 11/Windows 10 transition cycle, and changes in replacement cycles.

• Consumer researchUBS Evidence Lab Global Consumer PC Survey

  PC replacement cycles, budgets, and willingness to buy AI PCs

  Based on the semiannual January 2026 survey of 1,500 PC users in the US and China, it observes shorter replacement cycles, higher budgets, and sustained strong interest in AI PCs.

Asset mapping & comparison

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

• Semiconductors and memory supply chain

  Benefits from rising DRAM, NAND, and HBM prices and AI server demand

  Strengths

  Tight supply, strong AI computing demand, and a price upcycle extended through 2027

  Weaknesses

  High prices could eventually suppress end-demand and create risk of a cyclical reversal

  Comparison

  Compared with PC brands, semiconductors and components can benefit more directly from price and specification upgrades

  Risks

  Demand revisions, delayed customer purchases, peak in the price cycle

• Hyperscalers

  AI and general server memory costs significantly raise capital expenditures

  Strengths

  Strong demand for AI applications and operating cash flow still able to support current investment

  Weaknesses

  Capital intensity rises from around 10% before AI to about 35% in 2026, with the capex/operating cash flow ratio increasing significantly

  Comparison

  More able than traditional enterprise IT spending to absorb costs, but requires stronger AI monetization

  Risks

  Capital expenditures approaching or exceeding free cash flow capacity, insufficient AI commercialization speed

• PC brands

  Face pressure from rising memory costs, price pass-through, and weaker demand

  Strengths

  In the short term, can support revenue and profit through price increases, prioritizing premium products, and low-cost inventory

  Weaknesses

  In the second half of 2026, inventory costs rise, enterprise budgets are consumed early, and mainstream models may be downgraded in configuration

  Comparison

  Compared with component suppliers and AI server ODMs, PC brands are in a less favorable position on pricing power and demand elasticity

  Risks

  Shipment declines, margin pressure, consumers delaying replacement due to price increases

• Server ODMs and AI server supply chain

  Benefits from AI server volume growth and upward revisions to cloud capital expenditures

  Strengths

  High growth in AI server demand, rising content value, and some ODMs with high AI server exposure

  Weaknesses

  ODMs with low AI server exposure benefit less, and traditional PC-related businesses may weigh on performance

  Comparison

  UBS prefers ODMs with high AI server exposure such as Quanta, Hon Hai, Wistron, and Wiwynn, while being cautious on Compal and Pegatron

  Risks

  Changes in customer capex timing, supply chain bottlenecks, pullbacks after valuation gains

• Component suppliers

  Benefits from server specification upgrades, tight supply, and cost pass-through

  Strengths

  Rising content value, tight supply, and relatively less direct exposure to PC end-demand pressure

  Weaknesses

  Valuations have already been significantly rerated, with some component valuations above historical ranges

  Comparison

  Component valuations are about 33x forward P/E, above the historical 15-25x range; brands are about 10x and ODMs about 13x

  Risks

  Valuation pullback, specification upgrades below expectations, customer order adjustments

Key data

• Forecast increase in DRAM/NAND pricesUp 458%/321% respectively in 2025-2027Further raised from the February forecast of 289%/144%.
• 2026 global PC shipment forecastAbout 242 million units, about -10.8% to -11% year over yearLowered again from the previous roughly 255 million units and -4% year over year.
• 2026 PC revenue forecastAbout +1% year over yearShipment declines are partly offset by ASP rising about 13.5% year over year to $834.
• AI and general server memory spendingAbout $68 billion in 2025, about $524 billion in 2027Annual memory spending increments in 2026 and 2027 are about $209 billion and $247 billion, respectively.
• General server memory cost shareAbout 35% rising to 67%Per data center server memory cost could rise to about $23,000.
• Hyperscaler capital expendituresAbout $840 billion/$926 billion in 2026/2027EUBS expects further upward revision pressure remains.
• Consumer PC replacement cycleShortened from 3.17 years to 3.03 yearsMainly driven by China shortening from 3.36 years to 3.07 years, while the US was largely flat.
• Interest in AI PCs67% of respondents are very or somewhat interestedAmong them, 77% said they may upgrade to or purchase devices with AI PC features and are willing to pay a premium of about 15%-16%.

Impact & implications

The supply chain impact shows clear divergence: memory price increases boost the outlook for the storage supply chain in the short term and raise AI server and cloud capital expenditures, but they also burden PC brands and end-demand. For portfolios, the report more strongly supports allocation to AI servers, semiconductors, networking equipment, server ODMs, and component suppliers, while reducing optimistic expectations for traditional PC brands. If the pace of AI monetization by cloud vendors cannot keep up with capex expansion, free cash flow pressure around 2027 could become a market focus.

Risks

• Memory price increases may be too large, causing PC demand to weaken more than expected.
• After further upward revisions to hyperscaler capital expenditures, insufficient AI revenue monetization could create free cash flow pressure.
• Short-term earnings resilience at PC brands may mask demand and margin pressure in the second half of the year.
• Consumers and enterprises may respond to price increases by delaying replacements or lowering configurations.
• Highly valued components and AI server-related assets face correction risk if order or capex expectations are revised downward.
• Renewed memory supply or demand revisions could trigger a reversal in the price cycle.

What to watch

• Whether PC shipments show a sustained below-seasonal slowdown from the second through fourth quarters of 2026.
• Whether DRAM, NAND, and HBM contract prices continue to be revised upward, especially server DDR5 and HBM prices.
• Whether hyperscalers' 2026-2027 capital expenditure guidance is raised again.
• The pace of AI revenue monetization by cloud vendors, changes in operating cash flow, and the capex/operating cash flow ratio.
• The pace of low-cost inventory depletion at PC brands, their ability to pass through price increases, and gross margin performance.
• Order visibility and valuation changes for AI server ODMs, networking equipment, and component suppliers.
• Whether US and Chinese consumer PC replacement cycles, budgets, and willingness to pay AI PC premiums remain sustained.