Artificial intelligence and infrastructure refreshes support hardware demand, but the margin for earnings error has narrowed after the sector’s sharp rally

Overview

The report previews C2Q26 earnings for U.S. technology hardware vendors. Goldman Sachs believes enterprise data center modernization, aging server refreshes, on-premises deployment of agentic artificial intelligence, and demand for artificial intelligence training and inference continue to improve, giving server, storage, and networking hardware revenue upside potential. Because key components are in short supply, recent growth is more likely to be driven by price rather than volume. Meanwhile, the sector has already rallied sharply in the second quarter, investors have a high bar for earnings beats, and near-term risk-reward has become more complex.

Core views

Traditional servers will benefit from both aging installed-base refreshes and new agentic artificial intelligence workloads, with DELL and HPE as the main beneficiaries. Artificial intelligence server demand is gradually expanding from hyperscale cloud providers and neocloud providers to enterprise and sovereign customers, and DELL’s leading share in enterprise and neocloud markets is an advantage. Agentic artificial intelligence will also increase demand for unified enterprise data, unstructured storage, and high-performance data platforms, benefiting NTAP, DELL, and HPE. The PC industry faces risks from memory and storage shortages, shipment declines, and price sensitivity. DELL, with higher commercial customer exposure, is relatively more resilient, while HPQ’s consumer business carries greater risk, though tariff refunds may provide a one-time earnings upside for its printing business.

Analysis framework

The report combines company earnings forecasts, FactSet consensus estimates, IDC industry data, 650 Group market forecasts, 451 Research data center capacity data, CIO spending surveys, discussions with company management, and interviews with former industry executives to evaluate hardware vendors across demand, pricing, shipments, supply chain, market share, and earnings scenarios.

Methodology notes

• Industry demand analysisTotal addressable market and customer structure analysis

  Assess demand potential through the market size, growth rate, and customer mix of traditional servers, artificial intelligence servers, enterprise storage, and data center networking.

  The report cites industry forecasts such as 650 Group and distinguishes among hyperscale cloud providers, neocloud providers, enterprise customers, and sovereign customers to identify demand exposure for each hardware vendor.

• Earnings forecastBottom-up earnings forecast and consensus comparison

  Compare revenue, gross margin, operating profit, and earnings-per-share forecasts with company guidance and market consensus expectations.

  This method is used to assess the scope for earnings beats and evaluate the impact of pricing, product mix, and cost changes on the profitability of DELL, HPE, NTAP, and HPQ.

• Scenario analysisTariff refund scenario analysis

  Estimate the potential impact of IEEPA tariff refunds on HPQ’s printing business profit and earnings per share under different refund ratios.

  The report sets refund scenarios of 25%, 50%, 75%, and 100%, showing that refunds may bring one-time earnings upside not yet included in company guidance.

• Supply-demand analysisVolume and price contribution decomposition

  Distinguish whether hardware revenue growth comes from shipment volume or average selling price, and analyze the different effects of component shortages on each.

  DRAM and NAND shortages suppress server and PC shipments but enhance vendors’ pricing power, so near-term revenue growth may be driven mainly by price.

Asset mapping & comparison

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

• DELL

  Buy; comprehensive beneficiary of traditional servers, artificial intelligence servers, enterprise storage, and commercial PCs

  Strengths

  Leads in market share in enterprise traditional servers and artificial intelligence servers, is tied with SMCI as a leader in the neocloud artificial intelligence server market, and has scale procurement and supply chain advantages.

  Weaknesses

  The PC business remains affected by shipment declines and rising component costs, and commercial PC market share has recently been under pressure.

  Comparison

  Has greater commercial customer and data center exposure than HPQ; compared with HPE and NTAP, has a broader business scope covering servers, storage, and PCs.

  Risks

  Supply constraints, artificial intelligence server inventory iterations, customers shifting to ODM procurement, declining cloud computing rental prices, and elevated sector valuations.

• HPE

  Buy; benefits from enterprise and sovereign artificial intelligence servers, traditional servers, networking, and consumption-based infrastructure

  Strengths

  Has higher exposure to enterprise and sovereign customers, and offers artificial intelligence servers, high-performance networking, and the GreenLake consumption model.

  Weaknesses

  Storage benefits are expected to be lower than for DELL and NTAP, and networking spending intentions have weakened sequentially.

  Comparison

  Benefits alongside DELL from enterprise on-premises computing, but DELL has a greater advantage in server market share; storage platform differentiation is weaker than NTAP’s.

  Risks

  Insufficient component supply, slowing networking demand, storage product mix improvement falling short of expectations, and delays in large project delivery.

• NTAP

  Buy; core beneficiary of the unified data platforms and enterprise storage required by agentic artificial intelligence

  Strengths

  ONTAP can unify management of hybrid multi-cloud environments, AIDE provides artificial intelligence contextual memory, metadata cataloging, semantic search, and governance capabilities, and the company has a leading share in unstructured data platforms.

  Weaknesses

  The total enterprise storage market is currently expanding more slowly than servers and networking, and the neocloud high-performance storage platform remains in the early stages of commercialization.

  Comparison

  More focused on storage and data management than DELL and HPE, with a more direct product differentiation opportunity from agentic artificial intelligence’s need to manage data silos.

  Risks

  Sharp increases in NAND costs, slower-than-expected enterprise artificial intelligence adoption, competition from parallel file systems, and slow commercialization progress for AFX and AIDE.

• HPQ

  Sell; PC and printing businesses face structural and cyclical pressures

  Strengths

  Price increases, premium PCs, artificial intelligence PCs, and potential tariff refunds for the printing business can support earnings; artificial intelligence PCs have higher gross margins than ordinary PCs.

  Weaknesses

  Higher exposure to consumer PCs, pressure on volumes and market share, and price increases that may weaken end demand.

  Comparison

  Lacks data center infrastructure exposure on the same scale as DELL and is more sensitive to weak PC and printing markets.

  Risks

  Worsening price elasticity, component shortages, market share losses, memory configuration constraints for artificial intelligence PCs, and failure to realize tariff refunds.

Key data

• Technology hardware sector second-quarter performance+111%The SPX rose 13% over the same period; the sector’s significant outperformance makes pre-earnings risk-reward more challenging.
• 2030 traditional server market forecast$164 billion650 Group expects the market to nearly double from current levels, 55% above its previous 2030 forecast; this quarter’s 2026–2030 forecasts were raised by about 31% on average.
• Potential uplift to CPU servers from agentic artificial intelligence5–6xA former Intel executive interviewed expects that as workloads shift from training to inference and agentic workflows, the CPU-to-GPU ratio may gradually move from 1:4 toward 1:1.
• Enterprise server spending intention60%In the June 2026 CIO survey, 60% of respondents expected to increase server spending, up from 37% previously.
• Enterprise storage spending intention75%The share of respondents expecting to increase storage spending rose to 75% from 47% previously.
• 2030 global data center capacity forecast217GWAbove the previous forecast of 168GW; compared with 101GW in 2025, this implies 116GW of additional capacity and about $6 trillion in capital expenditure.
• 2030 artificial intelligence server market forecast$1.2 trillion650 Group expects the market to expand about sixfold by 2030, and raised its 2026–2030 forecasts by about 19% on average this quarter.
• External storage market growth rate2025–2030 compound annual growth rate of 6%Below CPU servers at 13%, artificial intelligence servers at 45%, and data center switching equipment at 31%.
• 2026 NAND flash price forecastUp more than 200% year over yearHigh costs test vendors’ pricing and product mix, but DELL and NTAP’s multi-tier storage portfolios, as well as NTAP Keystone and HPE GreenLake consumption models, can partly cushion the impact.
• C2Q26 PC shipments67 million units, down 4% year over yearThis is the first decline after nine consecutive quarters of growth; the industry average selling price reached $1,049, up 19% year over year and the highest level in 24 years.
• DELL FQ2'27E revenue forecast$44.803 billionGoldman Sachs expects 50% year-over-year growth, close to the FactSet consensus estimate of $44.843 billion.
• HPQ earnings per share under tariff refund scenarios$3.12–$3.46Corresponds to refunds of 25% to 100% of relevant fiscal 2025 tariffs and expenses, versus baseline earnings per share of $3.00.

Impact & implications

Industry revenue and earnings may still see upward revisions, but share prices already embed strong growth expectations, so investors should focus more on backlog, supply assurance, pricing durability, and margins rather than revenue growth alone. Over the medium term, agentic artificial intelligence will extend compute demand from GPU training further into CPU orchestration, local inference, enterprise data governance, and high-performance storage, expanding the serviceable markets for DELL, HPE, and NTAP. In the near term, component shortages are favorable for pricing but limit volumes; if supply improves faster than demand, the pricing tailwind may weaken. Although PC vendors can offset part of the shipment decline through price increases and premiumization, demand elasticity and market share losses remain key constraints.

Risks

• The technology hardware sector has significantly outperformed the broader market, and any earnings or guidance that fails to meet elevated expectations could trigger a valuation pullback.
• Shortages of DRAM, NAND, and other key components may limit server, storage, and PC shipments.
• Average selling prices may fall after component supply improves, making the current price-driven revenue growth difficult to sustain.
• Enterprise and sovereign artificial intelligence investment still lags neocloud providers, creating uncertainty around the pace of demand diffusion.
• Large neocloud customers may shift to ODM-style procurement, weakening branded server vendors’ market share and margins.
• The launch of next-generation artificial intelligence platforms may lead to impairment of older server inventory or delayed orders.
• Continued PC price increases may trigger demand destruction and market share losses.
• Data center projects may be delayed due to insufficient power supply even if end artificial intelligence demand remains strong.
• Declining cloud computing rental prices may weaken the economics of enterprises building on-premises artificial intelligence infrastructure.
• Incremental enterprise storage demand, tariff refunds, and commercialization of new high-performance storage products all carry execution risk.

What to watch

• The respective contributions of volume and average selling price to traditional server growth for DELL, HPE, and SMCI.
• Supply allocation, long-term purchase agreements, and price changes for DRAM, NAND, and other key components.
• The extent to which server and storage backlogs support revenue in the second half of 2026 and 2027.
• Whether artificial intelligence server projects among enterprise and sovereign customers continue to broaden.
• Order and market share progress for DELL, HPE, and NTAP’s high-performance, disaggregated storage platforms.
• Customer adoption of NTAP AIDE, ONTAP, AFX, and Keystone.
• The impact of new platform launches such as Vera Rubin on legacy artificial intelligence server inventory and purchasing cadence.
• Whether PC price increases can continue to offset double-digit shipment declines in the second half of 2026.
• Changes in DELL and HPQ’s commercial and consumer PC market share.
• Whether HPQ can obtain IEEPA tariff refunds and their actual contribution to printing business profit and earnings per share.