Research Packs
Historical compatibility record

Taiwan Monthly Revenue Update

Historical data for 2026-09, compared with 2026-08. This is not a current investment thesis or demo. Current research industries are CPO, Physical AI, and space infrastructure. Check the registry for approved Packs.

Period2026-09
As-of date2026-10-17
Compared with2026-08
Coverage3 companies
EvidenceOfficial TWSE
Next releaseby 2026-11-10
Investment view effect

Current thesis effect

The July revenue data for EquityLayer's tracked Taiwan AI-materials slice does not show broad near-term deterioration relative to June.

investment view effect: insufficient evidence — The 2026-09 Pack contains new evidence. It does not reuse the registered 2026-07 tests. Review the new evidence before changing the view.

Open the full thesis

What changed

2026-09 replaces 2026-08; 3 tracked symbols moved in reported monthly revenue.

  • 2383.TW · 台光電: +6.9% vs prior snapshot; current MoM +6.9%, YoY +168.9%.
  • 3037.TW · 欣興: +5.1% vs prior snapshot; current MoM +5.1%, YoY +64.7%.
  • 8046.TW · 南電: +4.4% vs prior snapshot; current MoM +4.4%, YoY +57.8%.
Claim checks

Evidence status and thesis effect

  • materials-2383-july · verification: Confirmed · thesis effect: insufficient evidence — Resolved 2026-08-13 against the official July MOPS snapshot: EMC reported NT$19.2067B, above the ≥ NT$17.7B Confirmed band. The July read supports the continuation view; it does not establish valuation or causality.
  • materials-kyber-scope · verification: Confirmed · thesis effect: insufficient evidence — Resolved 2026-08-13 against the official July MOPS snapshot: Nan Ya PCB (8046.TW) was +16.13% MoM and Unimicron (3037.TW) was +9.11% MoM, both above the ≥ −3% Confirmed band. The timing issue remains contained in this test; it does not prove the full supply-chain thesis.
Limits and next check

What this update cannot prove

Unknowns: Reported revenue does not identify shipment volume, mix, margin, or customer concentration. A three-company slice is not evidence for the full Taiwan market or an entire industry. Issuer notes remain management statements and do not establish causality.

Falsifier: A later official MOPS release that shows broad deterioration in the tracked names, or a company disclosure that contradicts the revenue read-through, falsifies the current continuation view.

Next trigger: Refresh when the official MOPS 2026-10 monthly-revenue dataset is published; preserve this snapshot and append a new diff.

Follow it from your agent: create_research_follow {"pack_id":"tw-monthly-revenue"}. Your agent checks again on 2026-11-10 and tells you when a new version is out. Connect your agent

Cite as: EquityLayer Research. "Taiwan AI Compute monthly revenue." 2026-09, as of 2026-09. https://equitylayer.io/research/monthly-revenue

Evidence and sources

Official TWSE data

2383.TW
2026-09

台光電

NT$21.53B

Month over month+6.9%
Year over year+168.9%
YTD year over year+103.7%
Issuer note (management statement)受惠於AI伺服器成長力道
3037.TW
2026-09

欣興

NT$18.65B

Month over month+5.1%
Year over year+64.7%
YTD year over year+37.7%
Issuer note (management statement)產能及客戶需求增加
8046.TW
2026-09

南電

NT$5.59B

Month over month+4.4%
Year over year+57.8%
YTD year over year+41.8%
Issuer note (management statement)因受市場需求成長影響。

Reported facts: Monthly-revenue fields are issuer-reported. Changes are deterministic comparisons against the preserved prior snapshot.

Management claims: Issuer notes are company statements reproduced from the official dataset. They do not establish demand causality or a research conclusion.

  • 2383.TW · 受惠於AI伺服器成長力道
  • 3037.TW · 產能及客戶需求增加
  • 8046.TW · 因受市場需求成長影響。

Inference: These are deliberately narrow operating read-throughs derived from reported fields. They are not causal, valuation, or investment inferences.

Taiwan Stock Exchange (臺灣證券交易所) 2026 上市公司每月營業收入彙總表, dataset t187ap05_L, as published. The Open Data is made available to the public under the Open Government Data License. Open Government Data License: https://data.gov.tw/license

Open TWSE machine-readable source · Open-data licence

Agent/API details

Recommended workflow: discover_theses → test_thesis → get_thesis_update.

Remote MCP: https://equitylayer.io/mcp

JSON: /api/research/monthly-revenue?symbols=2383,3037,8046

Research Pack JSON: /api/research/pack?symbols=2383,3037,8046

Changes JSON: /api/research/changes?since=2026-06&symbols=2383,3037,8046

Research Update JSON: /api/research/update?pack_id=tw-monthly-revenue&since_version=taiwan-monthly-revenue:2026-06&symbols=2383,3037,8046

Backward-compatible tools: get_taiwan_monthly_revenue, get_taiwan_monthly_research_pack, get_research_changes_since

Add it to Claude Code:

claude mcp add --transport http equitylayer https://equitylayer.io/mcp

Or paste into any MCP client config:

{ "mcpServers": { "equitylayer": { "url": "https://equitylayer.io/mcp" } } }

Pack version: taiwan-monthly-revenue:2026-09 · Snapshot hash: 64dea69a5c27579cda94295dda593f5871bc9e96373f24526039efc1b9c9748e

Prior snapshot hash: 8cf5b9818d5c28a79a188c569164e6eb6984e7d3ab9462757ad02ae2eb78fa01

Each query returns the same dated snapshot, source attribution, and limitations. User-specific interpretation stays inside the caller's own agent, so EquityLayer incurs no per-question model cost.