Alphabet raises capex again, benefiting the Asia-Pacific cloud semiconductor supply chain
AI summary card
Alphabet raises capex again, benefiting the Asia-Pacific cloud semiconductor supply chain
Morgan Stanley believes that Alphabet’s higher AI and cloud capital spending, the monetization of TPU, and improved Axion CPU performance create a positive read-across for AI supply chain companies such as Aspeed, Montage, TSMC, ASE, MediaTek, KYEC, MPI, WinWay, and Hon Precision.
- Alphabet raised FY26 capital expenditure guidance from US$180-190bn to US$195-205bn, with 2Q26 capex reaching US$44.9bn, up 100% YoY and 26% QoQ.
- Management reiterated that FY27 capital expenditure will increase significantly, indicating continued durability in AI compute demand and cloud infrastructure expansion.
- 2Q26 marked the first delivery of TPU systems to customer data centers with related revenue recognized; the report believes 2027 will be the key year for TPU sales ramp-up.
- Alphabet is optimizing its full computing stack across TPU, NVIDIA accelerators, Axion CPU, servers, networking, and data centers; the new Axion CPU offers about 30% better performance per dollar than peer products.
Report interpretation
Overview
This report is Morgan Stanley’s Asia-Pacific cloud semiconductor read-across commentary on Alphabet’s 2Q26 earnings call. The core conclusion is that Alphabet’s continued upward revisions to AI and cloud capital spending, ongoing compute capacity constraints, the start of external TPU deliveries, and an expected ramp in 2027 all support the view that AI infrastructure investment remains in an upcycle, sending positive signals to the Greater China and Asia-Pacific cloud semiconductor, AI server, and advanced packaging and testing supply chains.
Core views
The report believes Alphabet’s commentary helps ease investor concerns about a slowdown in AI investment. FY26 capex guidance was raised to US$195-205bn, with 2Q26 capex at US$44.9bn, of which about 60% of technical infrastructure spending went to servers and 40% to data centers and networking. TPU systems began delivery to customer data centers and recognized revenue in 2Q26, with only a small portion of contracted sales expected to be recognized in 2026 and most expected in 2027. Alphabet is also expanding TPU, NVIDIA accelerators, Axion CPU, servers, networking, data centers, and third-party capacity, aiming to lower workload costs and improve capital efficiency.
Analysis framework
The report uses a methodology of mapping earnings read-across from large U.S. cloud providers to the Asia-Pacific semiconductor supply chain: first observing Alphabet’s capital expenditure, AI compute demand, TPU delivery pace, and in-house CPU progress, and then assessing how demand transmits to cloud semiconductors, AI servers, wafer foundries, packaging and testing, test interfaces, and related component companies.
Methodology notes
Map Alphabet’s capital expenditure and compute deployment pace to the Asia-Pacific cloud semiconductor supply chain
Upward revisions to cloud provider capex usually imply stronger demand for servers, accelerators, networking, data centers, and related semiconductors; this report uses information from Alphabet’s 2Q26 earnings call to assess the durability of AI infrastructure investment.
Break down beneficiary segments across TPU, NVIDIA accelerators, Axion CPU, servers, networking, and data centers
The report decomposes the expansion and optimization of Alphabet’s full computing stack into potential demand for wafer foundries, packaging and testing, ASICs, server management chips, memory interfaces, test interfaces, and the AI server value chain.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Aspeed (5274.TW)Cloud semiconductor beneficiary
- Strengths
- Alphabet’s continued AI and cloud investment through 2027 is supportive of expectations for server-related chip demand.
- Weaknesses
- The report does not provide updated company-level financial forecasts or a target price.
- Comparison
- Along with Montage, it is explicitly named in the report as a direct beneficiary among cloud semi names.
- Risks
- If cloud capex slows, server deployment is delayed, or customer order timing falls short of expectations, the read-across may weaken.
- Montage (6809.HK)Cloud semiconductor beneficiary
- Strengths
- Benefits from continued investment in AI servers and cloud infrastructure and is explicitly named by Morgan Stanley as a positive read-across name.
- Weaknesses
- Changes in the company’s own orders, pricing, and margins are not discussed in this report.
- Comparison
- Like Aspeed, it is one of the report’s core cloud semi read-across names.
- Risks
- If demand for memory interfaces and server platforms is weaker than expected, the transmission effect may be affected.
- TSMC (2330.TW)AI supply chain beneficiary
- Strengths
- Expansion in TPU and AI accelerator demand usually supports demand for advanced-node foundry services.
- Weaknesses
- The report only provides supply chain read-across and does not offer standalone earnings sensitivity for TSMC.
- Comparison
- Compared with equipment or testing segments, TSMC has more upstream core manufacturing exposure.
- Risks
- The visibility of orders may be affected by the pace of customer in-house chip development, capacity allocation, and capex cycle changes.
- ASE (3711.TW)Packaging and testing supply chain beneficiary
- Strengths
- Rising demand for AI chips and server-related products may support advanced packaging and packaging/testing demand.
- Weaknesses
- The report does not quantify the specific revenue contribution from Alphabet or the TPU chain.
- Comparison
- Like KYEC, it is a read-across name in the downstream AI semiconductor supply chain.
- Risks
- Changes in packaging and testing capacity, pricing, and customer mix may affect earnings leverage.
- MediaTek (2454.TW)AI supply chain beneficiary
- Strengths
- The report includes it among AI supply chain companies that may benefit from positive commentary on TPU and Axion CPU.
- Weaknesses
- This report does not explain the specific product linkage or order path.
- Comparison
- Compared with pure cloud server names, MediaTek’s business is more diversified, so the read-across transmission may be more indirect.
- Risks
- If the share of AI-related business or the pace of customer adoption is below expectations, the positive read-across may be diluted by volatility in other businesses.
- KYEC (2449.TW)Testing services beneficiary
- Strengths
- Growth in AI semiconductor shipments may drive testing demand.
- Weaknesses
- The report does not provide specific data on testing volume, pricing, or utilization rates.
- Comparison
- Like ASE, it is in the downstream supply chain, but with more exposure to the testing segment.
- Risks
- Changes in customer product mix and test outsourcing ratios may affect the degree of benefit.
- MPI (6223.TW)Test interface and equipment-related beneficiary
- Strengths
- Growth in AI chips and high-speed computing demand may support demand related to test interfaces.
- Weaknesses
- The report does not elaborate on company-level order or capacity details.
- Comparison
- Compared with wafer foundries and packaging/testing, MPI has more exposure to the testing tool chain.
- Risks
- If the expansion in AI chip testing demand falls short of expectations or customer qualification is delayed, leverage may be limited.
- WinWay Technology Co Ltd (6515.TW)AI supply chain beneficiary
- Strengths
- The report lists it as a supply chain company that may benefit from positive commentary on TPU and Axion CPU.
- Weaknesses
- The report does not provide updated forecasts, rating changes, or target price changes.
- Comparison
- Like MPI, it is a read-across name related to test interfaces.
- Risks
- Demand realization depends on customer adoption, product specification upgrades, and industry capacity timing.
- Hon Precision (7769.TW)AI supply chain beneficiary
- Strengths
- Included by the report within the scope of positive read-across for the AI supply chain.
- Weaknesses
- This report does not quantify the specific path of revenue contribution.
- Comparison
- Like MPI and WinWay, it is a higher-beta supply chain read-across name.
- Risks
- Order visibility, customer concentration, and the production ramp timing of AI projects may affect realization.
- Alphabet (GOOGL.O)Source company for the read-across
- Strengths
- Higher capex, the start of external TPU deliveries, and improved Axion CPU performance together indicate the intensity of AI infrastructure investment.
- Weaknesses
- The significant increase in FY27 capex implies continued pressure on free cash flow.
- Comparison
- As a U.S. cloud provider, its capex signals are used to map demand for the Asia-Pacific semiconductor supply chain.
- Risks
- External TPU supply depends on customer demand, industry constraints, and Alphabet’s internal AI demand; if demand or capacity changes, the supply chain read-across will change.
Key data
- Alphabet FY26 capital expenditure guidanceUS$195-205bnRaised from the previous US$180-190bn, mainly driven by faster compute deployment, Google Cloud customer demand, and internal business demand.
- Alphabet 2Q26 capital expenditureUS$44.9bnUp 100% year over year and 26% quarter over quarter.
- Technical infrastructure spending mixServers 60%; data centers and networking 40%Reflects the direct boost from AI and cloud infrastructure expansion to servers and data center segments.
- TPU revenue recognition paceStarted in 2Q26; 2027 is the main ramp year2Q26 was the first delivery of TPU systems to customer data centers with related revenue recognized; 2026 is expected to account for only a small portion of contracted sales.
- Axion CPU performanceAbout 30% better performance per dollar than peer productsAlphabet said the new agent-optimized Axion CPU helps lower workload costs and improve capital efficiency.
Impact & implications
In terms of investment implications, the report’s main message is that demand for AI infrastructure has not shown a clear slowdown and is instead being reinforced by Alphabet’s continued capex increases and stronger expectations for 2027 spending. The cloud semiconductor and AI server supply chains are likely to benefit, especially companies related to server management chips, memory interfaces, advanced process technology, packaging and testing, test interfaces, and server hardware. However, the significant increase in FY27 capex also implies Alphabet’s own free cash flow may remain under pressure, while external TPU supply will still depend on the balance among customer demand, industry capacity constraints, and Alphabet’s internal AI demand.
Risks
- Alphabet or other cloud providers may deliver AI capital expenditure below current guidance or FY27 incremental spending may fall short of expectations.
- External TPU sales may be delayed by customer demand, industry capacity constraints, or crowding out from Alphabet’s internal AI demand.
- A slower deployment pace for AI servers, data centers, and networking could weaken order transmission to the semiconductor supply chain.
- The actual degree of benefit to supply chain companies may differ from the report’s read-across, especially for companies without disclosed direct order relationships or revenue contribution.
- The report includes disclosures of Morgan Stanley’s investment banking, shareholding, or service relationships with multiple covered companies, and investors should consider potential conflicts of interest.
What to watch
- Whether Alphabet’s FY26 capital expenditure continues to land within the US$195-205bn range or is raised further.
- The growth rate of FY27 capital expenditure, pressure on free cash flow, and management’s latest commentary on AI compute demand.
- The quarterly pace of TPU system deliveries to external customer data centers and related revenue recognition.
- The cost-efficiency performance of Axion CPU in real workloads and its impact on the server CPU ecosystem.
- Subsequent orders, capacity utilization, and earnings guidance for Aspeed, Montage, TSMC, ASE, MediaTek, KYEC, MPI, WinWay, and Hon Precision.