Amazon's 2Q26 results reinforce the read-through signal that AI infrastructure demand remains strong

Overview

This report examines the development of global AI infrastructure, cloud capacity, AI chips, and AI application commercialization through Amazon's 2Q26 results and management's earnings call, and maps these trends to China's technology hardware supply chain. The core conclusion is that AWS growth and capacity reservations indicate that AI infrastructure demand remains strong, while management's comments on investment payback periods and long-term AWS revenue potential help ease market concerns about excessive AI investment.

Core views

First, Amazon's 2Q26 revenue and operating profit both exceeded market expectations, with AWS serving as the most important growth engine. Second, AWS capacity has been reserved through 2028F, and most of 2027 capacity has also largely been reserved, indicating strong persistence in cloud and AI computing demand. Third, both the AI revenue run rate and AI chip business ARR exceeded USD25bn and grew at triple-digit rates year over year, indicating that AI monetization is showing signs of scale. Fourth, servers and networking equipment pay back in less than three years on average, while servers have useful lives of at least five to six years and data centers have useful lives of more than 30 years, supporting the economics of elevated capital expenditure. Fifth, if global cloud capital expenditure remains strong, supply chain companies involved in optical modules and PCB/CCL may benefit.

Analysis framework

The report adopts an earnings read-through approach: it first compares Amazon's actual 2Q26 revenue, segment performance, operating profit, and capital expenditure with guidance and the Bloomberg consensus, then combines these results with management's comments on AWS capacity, AI chips, AI applications, investment payback periods, and future demand to assess the spillover impact of global AI infrastructure investment on China's hardware supply chain.

Methodology notes

• Earnings read-throughEarnings read-through

  Using leading cloud providers' earnings to validate AI infrastructure demand

  Amazon and AWS revenue, margins, backlog, capital expenditure, and management guidance are used to infer the impact of downstream cloud customer demand on upstream hardware segments such as optical modules and PCB/CCL.

• Relative valuationP/E multiple valuation

  Deriving target prices based on forecast EPS and target P/E multiples

  Zhongji InnoLight's target price of CNY1,325.00 is based on 20x 2027F EPS of CNY66.06; Shengyi Technology's target price of CNY195.00 is based on 43x 2027F EPS of CNY4.51.

• Operating economicsCapex ROI and asset life analysis

  Assessing the rationality of AI capital expenditure using payback periods and asset lives

  Management stated that servers and networking equipment pay back in less than three years on average, servers have useful lives of at least five to six years, and data centers can be used for more than 30 years. Therefore, the two to three years after payback may contribute to free cash flow.

Asset mapping & comparison

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

• Amazon (AMZN US)

  The core read-through subject, providing signals on global cloud and AI capital expenditure demand.

  Strengths

  2Q26 revenue and operating profit exceeded expectations; strong AWS growth, backlog, ARR, and AI revenue run rate all demonstrate demand resilience.

  Weaknesses

  The company is Not rated in the report, and higher capital expenditure may continue to raise market concerns about investment intensity and free cash flow.

  Comparison

  AWS's growth rate and margin are both significantly better than those of the overall business, making it the most important business line in this read-through.

  Risks

  Excessive capital expenditure, rising memory costs, weaker-than-expected AI demand realization, and a mismatch between cloud capacity construction and customer reservations.

• Zhongji InnoLight (300308 CH)

  A global leader in optical modules, benefiting from optical interconnect demand at cloud and AI data centers.

  Strengths

  Rated Buy by Nomura, with a current price of CNY864.00 and a target price of CNY1,325.00; strong AWS demand supports high-end optical module demand.

  Weaknesses

  Highly dependent on high-end optical module demand and the pace of product upgrades.

  Comparison

  Compared with PCB/CCL suppliers, Zhongji InnoLight has more direct exposure to optical module upgrades in AI data centers.

  Risks

  Lower-than-expected demand for high-end optical modules in datacom and telecom markets; intense competition in 400G and 800G; slower-than-expected upgrades to 800G and 1.6T; and intensified price competition affecting exports to global customers.

• Shengyi Technology (600183 CH)

  A major PCB/CCL supplier benefiting from demand for servers, AI infrastructure, and high-speed electronic materials.

  Strengths

  Rated Buy by Nomura, with a current price of CNY100.08 and a target price of CNY195.00; valuation is based on 43x 2027F EPS of CNY4.51.

  Weaknesses

  Margins may be affected by competition in the CCL market and fluctuations in downstream demand.

  Comparison

  Compared with optical module companies, Shengyi Technology is more oriented toward the materials and PCB/CCL chain, with benefits more dependent on expansion in downstream areas such as servers, 5G, and automotive electronics.

  Risks

  Lower-than-expected downstream PCB demand from 5G, servers, automotive electronics, and other sectors; intensified CCL market competition putting pressure on margins.

• Shengyi Electronics (688183 CH)

  A major PCB supplier listed in the report as a Chinese technology hardware beneficiary of AI infrastructure demand.

  Strengths

  Listed as Buy by Nomura in the main text, benefiting from server and AI-related PCB demand.

  Weaknesses

  The report does not disclose detailed target price or valuation methodology for this company.

  Comparison

  It is in the same PCB/CCL supply chain as Shengyi Technology, but the report provides less detailed valuation data than for Shengyi Technology.

  Risks

  Lower-than-expected downstream demand from servers and other sectors, intensified competition in the PCB industry, and weaker-than-expected transmission of AI capital expenditure.

Key data

• Amazon 2Q26 revenueUSD200.6bn, up 19.6% year over yearAbove company guidance of USD194-199bn and 2% higher than the Bloomberg consensus.
• Amazon 2Q26 operating profitUp 43% year over yearAbove the market consensus estimate of 16%.
• AWS 2Q26 revenueUSD42.2bn, up 36.8% year over year and 12.4% sequentially4% above the Bloomberg consensus.
• AWS 2Q26 operating margin39.4%Up 1.7 percentage points sequentially.
• AWS backlogUSD496bnTriple-digit year-over-year growth.
• AWS ARRUSD169bnThe report disclosed that AWS reached USD169bn ARR.
• AI revenue run rateAbove USD25bnTriple-digit year-over-year growth.
• AI chip business ARRAbove USD25bnTriple-digit year-over-year growth.
• 2Q26 capital expenditureUSD54.2bnThe report text also cites year-over-year growth of 51.5% and 68.4%; both statements indicate a significant increase in capital expenditure, primarily for AWS and generative AI.
• FY26E capital expenditure guidanceUSD220bnRaised by management from USD200bn due to factors including higher memory costs and strong customer demand.
• 3Q26E revenue guidanceUSD197-202bn, up 9-12% year over yearThe corresponding operating profit guidance is USD22.5-26.5bn.
• Graviton penetrationUsed by 98% of AWS's top 1,000 EC2 customersManagement said Graviton can provide up to a 30%-40% cost-performance improvement versus other options.

Impact & implications

The report is positive on the AI infrastructure chain. Sustained AWS demand, capacity reservations through 2028F, and rising AI revenue and chip business run rates indicate that large cloud providers' capital expenditure is not merely a cost burden, but may correspond to visible customer demand and investment payback. In China, investment opportunities are concentrated in optical modules, PCB/CCL, and server-related hardware supply chains, particularly Zhongji InnoLight, Shengyi Tech, and Shengyi Electronics.

Risks

• Returns on AI infrastructure investment fall short of management expectations, causing the market to again worry about excessive capital expenditure.
• Actual AWS demand or capacity reservations fall below 2027-2028F expectations.
• Continued increases in memory costs intensify capital expenditure pressure.
• High-end optical module demand falls below expectations, or upgrades to 400G, 800G, and 1.6T products progress more slowly than expected.
• Downstream PCB/CCL demand falls below expectations, or intensified market competition puts pressure on margins.
• Intensified price competition may affect the export performance of Chinese supply chain companies to global customers.

What to watch

• AWS revenue growth, operating margin, and backlog changes in subsequent quarters.
• Whether FY26E capital expenditure of USD220bn is raised further, as well as the intensity of capital expenditure in 2027.
• The realization of AWS capacity reservations for 2027-2028.
• Whether the AI revenue run rate and AI chip business ARR continue to grow at triple-digit rates.
• Customer adoption of internally developed chips such as Trainium and Graviton.
• The pace of 800G and 1.6T optical module upgrades, as well as PCB/CCL order and margin trends.