01
AMD and Anthropic sign multibillion-dollar AI infrastructure deal, expanding competition in AI compute
AI InfrastructureAMDAnthropic
On July 22, AMD announced a major AI infrastructure agreement with Anthropic involving the planned deployment of up to two gigawatts of AMD Instinct MI450 GPUs and related rack-scale systems. Initial deployment is expected to begin in 2027. AMD also agreed to invest up to USD 5 billion in Anthropic, subject to agreed milestones.
The agreement highlights how AI compute competition is moving beyond individual accelerator shipments toward broader infrastructure partnerships, customer commitments, and ecosystem positioning.
R&A View: AI chip competition is shifting from individual GPU performance to full-stack supply capability. Procurement teams should track AMD-based AI compute platforms, HBM supply, rack-scale power delivery, high-layer PCBs, connectors, cables, and thermal management parts as AMD’s AI ecosystem expands.
02
Wistron opens USD 700 million Texas facility for Nvidia AI system production
AI System ManufacturingNvidiaU.S. Localization
On July 22, Wistron opened a USD 700 million manufacturing facility in Fort Worth, Texas, to produce AI systems for Nvidia. The project reflects how Taiwanese electronics manufacturers are expanding U.S. production capacity to meet rising demand for AI infrastructure.
The facility adds another link to Nvidia’s localized AI hardware supply chain, following broader efforts to assemble advanced AI systems closer to U.S. customers and data-center deployment sites.
R&A View: AI system manufacturing is becoming more regionalized. This may increase demand for U.S.-based assembly, testing, power systems, cables, connectors, thermal parts, high-end PCBs, and related supply-chain services. Buyers should monitor whether localized production changes allocation, lead times, and delivery priority.
03
Alphabet raises 2026 capital expenditure forecast again as Google Cloud growth accelerates
Cloud CapexAI Data CenterTPU
On July 22, Alphabet raised its 2026 capital spending forecast by USD 15 billion to a range of USD 195 billion to USD 205 billion. Google Cloud quarterly revenue increased 82% year over year to USD 24.8 billion, supported by continued enterprise demand for AI and cloud infrastructure.
The higher capex outlook points to continued investment in data centers, networking, AI servers, storage, and custom chips such as TPUs.
R&A View: Hyperscaler capex remains one of the strongest signals for semiconductor demand. Sustained cloud investment may support server CPUs and accelerators, enterprise SSDs, memory, optical modules, PMICs, MOSFETs, connectors, liquid cooling, and power infrastructure components.
04
Texas Instruments issues stronger-than-expected guidance as industrial and data-center demand improves
Analog ICPower ManagementIndustrial Demand
On July 22, Texas Instruments forecast third-quarter revenue of USD 5.65 billion to USD 6.15 billion, above analyst expectations. The company also reported second-quarter revenue of USD 5.46 billion, up about 23% year over year.
The results point to improved demand conditions for TI, including industrial and data-center-related applications. However, they should not be read as confirmation that the entire analog and power-chip market has fully recovered.
R&A View: AI infrastructure investment may support demand for analog ICs, PMICs, signal-chain chips, interface ICs, isolation devices, amplifiers, ADC/DAC products, sensors, and industrial-control components. Buyers should monitor whether analog lead times and pricing begin to tighten selectively.
05
Zhongji Innolight seeks to raise up to USD 7 billion through Hong Kong listing
Optical Modules800G / 1.6TAI Interconnect
On July 21–22, Zhongji Innolight moved forward with a Hong Kong listing that could raise up to about USD 7 billion before any full exercise of the over-allotment option. The company makes optical transceivers used in data centers and AI computing systems, and its revenue has grown rapidly with AI infrastructure demand.
The offering has secured commitments from major cornerstone investors, highlighting strong capital-market interest in optical interconnect suppliers. Final proceeds still depend on pricing, demand, and the completion of the listing process.
R&A View: AI clusters require more high-speed data movement between GPUs, servers, and racks. Optical modules, lasers, DSPs, TIAs, driver ICs, optical engines, high-speed connectors, and precision assembly capacity should remain key watchpoints as AI systems scale from 800G toward 1.6T.
06
IQE raises annual sales growth forecast on AI and data-center demand for compound semiconductor wafers
Compound Semiconductor WafersEpitaxial WafersData Center
On July 21, UK semiconductor wafer supplier IQE raised its full-year revenue growth forecast after first-half trading came in better than expected. The improvement was supported by demand for compound-semiconductor epitaxial wafers used in AI infrastructure, data centers, and related applications.
The update shows that AI demand is moving further upstream into specialized wafer and compound-semiconductor supply chains.
R&A View: AI demand is not only a logic and memory story. It may also support compound semiconductor epitaxial wafers, optical communication wafers, RF materials, sensors, and power-related substrates. Procurement teams should watch upstream material availability, qualification cycles, and supplier capacity planning.
07
Supermicro says fiscal Q4 new orders exceeded USD 60 billion, lifting backlog to a record
AI SystemsNew OrdersBacklog
On July 21, Supermicro issued a preliminary business update stating that it had received more than USD 60 billion in new orders during fiscal Q4 2026 and that backlog had reached a record level at fiscal year-end.
The figure refers to new orders received during the quarter, not recognized revenue and not necessarily the exact backlog amount at quarter-end. The company indicated that these orders are expected to be delivered over multiple future quarters.
R&A View: AI hardware demand is increasingly visible at the system level. Beyond accelerators and HBM, procurement teams should track server chassis, motherboards, high-current power delivery, retimers, high-speed connectors, cables, thermal modules, fans, liquid-cooling parts, and testing capacity.
08
Nationwide U.S. data-center protests highlight growing power, water and permitting risks
Data CenterPower / WaterPermitting Risk
On July 18, coordinated protests against rapid data-center expansion were held at a reported 142 locations across 42 U.S. states, according to organizers cited in public reporting. The concerns focused on electricity demand, water use, land use, and local transparency as AI infrastructure expands.
This shows that the AI supply chain is increasingly affected by non-chip factors such as grid capacity, water access, permitting, community approval, and local policy risk.
R&A View: AI supply risk is no longer only about chip availability. If data-center projects slow due to electricity, water, permitting, or local opposition, the demand timing for servers, power systems, optical modules, storage, cooling equipment, and supporting components may also change.