Institutional Research

The workshop summary describes how tool-enabled, RAG-grounded and manager-orchestrated agents can automate repeatable financial-research work. Its central conclusion is that permissions, evidence controls, testing, review gates and cost limits must be designed into the workflow from the start.

• Agentic AI is framed as systems that plan, act and reflect, rather than chatbots that only answer questions.
• Live-data tools and approved-document RAG can ground research outputs in current or controlled information.
• Manager-led routing can improve specialization while directing routine tasks to cheaper, faster models.
• The proposed autonomous research team includes goal refinement, evidence retrieval, compliance review, termination checks and a constrained reporting agent.
• Testing, traceability, guardrails and explicit iteration limits are presented as prerequisites for production deployment.

The call reinforced Goldman Sachs' view that Murata is well positioned for sustained MLCC growth, supported by strong orders, high utilization and capacity expansion. The firm maintains Buy (on CL) and a 12-month ¥12,900 target price.

• June MLCC orders rose 25-30% month on month, and the July-September book-to-bill ratio may remain near the April-June 1.47x level.
• AI/data-center applications are the strongest demand area, while smartphone demand is seasonally improving and automotive demand is flat.
• Murata is targeting MLCC capacity growth of slightly above 10% in FY3/27 and 20-25% in FY3/28.
• AI/data-center applications could account for 20-25% of MLCC sales in FY3/27 and just under 40% in FY3/28.
• Goldman Sachs expects a second upward revision in the current fiscal year at interim earnings.

Goldman Sachs raised its target price to Rmb157 from Rmb143 after stronger 2Q26 results and higher revenue assumptions. The report expects PMIC and signal-chain expansion to drive growth, while viewing the valuation as fair.

• 2Q26 revenue was Rmb1,497mn, 25% above Goldman Sachs’ forecast and up 45% year on year.
• 2026E-28E revenue estimates rise 10%/13%/18%; net-income estimates rise 1%/13%/22%.
• The 12-month target price rises to Rmb157 from Rmb143, implying 28.2% upside from Rmb122.45.
• The target is based on a discounted 2030E P/E approach using a 37.4x target multiple and 10.2% cost of equity.

Goldman Sachs expects higher oil, LNG and diesel prices to raise September import and producer prices across much of Asia-Pacific. Headline and core CPI are generally at or above policy targets, although broader core inflation has risen only modestly so far.

• Goldman Sachs raised its December Brent forecast to $85/bbl.
• LNG and diesel prices reached new highs for the year, while refined-product prices remain elevated.
• Energy-price pass-through is expected to be stronger in import and producer prices than in consumer prices where subsidies or regulated prices apply.
• Goldman Sachs is above 2027 consensus most notably for India and Malaysia, and below consensus for Japan, Vietnam and the Philippines.

Goldman Sachs interprets the PBOC's Q3 MPC statement as not providing a fresh broad-easing signal and maintains its expectation of no policy-rate or RRR cuts in 2026. The notable change is stronger language aimed at preventing one-way RMB appreciation expectations ahead of year-end settlement activity.

• The September 19 Q3 MPC meeting retained a moderately accommodative policy stance; its statement was released on September 24.
• Goldman Sachs does not view revised language on counter-cyclical adjustment or monetary tools as evidence of stronger near-term easing.
• The report maintains a baseline of no policy-rate or RRR cuts in 2026.
• New language on preventing herd behavior and self-reinforcing irrational expectations points to concern over excessive RMB appreciation.
• The report sees targeted housing support as a more likely area for policy action than broad monetary easing.

The report argues that Fanuc’s procurement problems are company-specific rather than sector-wide, threatening production, shipments and market share. Lower earnings estimates and a reduced EV/EBITDA valuation premium lead Goldman Sachs to retain Sell.

• 12-month target price cut to ¥4,600 from ¥5,600, implying 18% downside to the stated share price.
• Channel checks indicate Fanuc’s production and shipment constraints persist while competitors report more positive orders, production and shipment conditions.
• FY3/28 and FY3/29 operating-profit forecasts were cut to ¥195 billion and ¥180 billion, respectively.
• The sector EV/EBITDA multiple was lowered to 9x from 10x and Fanuc’s relative premium to 60% from 70%.

The report argues that AI agents will increasingly capture consumer intent from discovery through payment over the next 3–5+ years, benefiting platforms with distribution, trust, transaction infrastructure and merchant participation. Adoption should begin in lower-risk, recurring and commoditized purchases, while liability, authorization and merchant-control issues remain major constraints.

• Goldman Sachs identifies $2.6tn of near-term US spending in high-likelihood categories for agentic commerce.
• About 2% penetration of high-likelihood card-present spending could add roughly 1 percentage point to eCommerce growth.
• The report favors platform, payment-network, commerce-infrastructure, identity and security providers over the long term.
• Trust, consumer intent verification, fraud liability and merchant willingness to expose inventory are central adoption hurdles.

Global raw-wafer demand is now forecast to grow at a 14% CAGR in 2025-28, while ex-China supply grows only about 5%. The firm sees a tactical opportunity in Sumco after its correction, but retains Market-Perform pending late-2027 contract renewals and potential 2028 pricing gains.

• Global wafer-demand forecast rises to a 14% 2025-28 CAGR from 12% previously.
• Ex-China supply is projected to grow only 5% annually through 2028.
• The ex-China market could enter shortage from late 2027 if Chinese suppliers remain focused on domestic demand.
• China domestic wafer demand is forecast at a 24% CAGR, above 17% supply growth, limiting potential exports.
• Sumco remains Market-Perform with a ¥3,290 target price despite potential 2028 pricing leverage.

The report frames the AI-capex debate around required future monetization rather than near-term margin pressure. It sees strong demand, constrained capacity and growing cloud backlogs as support for the return potential of the current investment cycle.

• A 15% ROIC hurdle implies roughly $1.42 trillion of cumulative revenue for six US hyperscalers in CY2028-30, or about $11.6 billion per GW annually.
• Goldman Sachs estimates AI infrastructure capex of about $1.3 trillion in 2027 and $2.0 trillion in 2028.
• The combined cloud backlog of AWS, Azure and Google Cloud was about $1.69 trillion as of C2Q26, versus $1.00 trillion of revenue required for those three companies under the framework.
• Consensus CY2026-27 capex estimates for five public US hyperscalers have risen about 66%, or roughly $750 billion, since the start of 2026.

UBS expects Cloud AI build-outs to keep TSMC leading-edge capacity and CoWoS packaging tight through 2027 while improving demand, pricing and earnings prospects across selected Asian semiconductor supply chains. The report highlights TSMC, MediaTek, ASE, UMC and a range of packaging, test and equipment beneficiaries.

• TSMC's N3 capacity is forecast to run above 100% loading in 2026-27 despite expansion to 210kwpm by end-2027.
• For each incremental 1GW AI-server project, UBS estimates US$1-2bn of TSMC revenue potential and roughly 2-5kwpm of leading-edge capacity demand.
• CoWoS industry capacity is projected to rise from 160kwpm at end-2026 to 270kwpm at end-2027, but demand is expected to remain strong.
• Google TPU shipments are forecast to rise from 4.9m units in 2026E to 10.1m in 2027E, with MediaTek gaining share.
• UBS expects CPO to become a meaningful but gradual opportunity, while pluggable optical transceivers remain the main solution for the next two to three years.
• Mature-node utilisation is forecast to improve in 2027 as Chinese competition becomes more rational and AI servers lift power-semiconductor demand.