13 companies here publish proof of their AI capability. They average 60 of 100 on that signal. 41 of 136 scored companies are in the published APAC 1,000.
| Measure | Value |
|---|---|
| Screened this cycle | 255 |
| Left too little public trail to assess | 112 |
| Cleared the evidence gate and were scored | 143 |
| Organisations counted in the market averages | 136 (7 duplicate rows for the same group counted once) |
| In the published APAC 1,000 | 41 |
| Domestic entities | 142 |
| Readiness score | 63 of 100 (range 56-70) |
| Average visibility | 41.2 of 100 |
| Evidence grade | C |
| Dossier | 20 pages, 4.6 MB |
The conditions companies here operate under score 63, and what those companies publish scores 41.2. Of the two, the second is the one companies here control, and the quicker to change: 90% of the 136 companies with enough of a public trail to score publish no proof of capability at all.
The 13 companies in Taiwan publishing proof of capability average 60 of 100, on a platform layer that is already built.
Buyers, investors and partners often treat a missing record as missing work. Most decisions about a company are shaped before anyone makes contact, based on what people can find. For most people, the public record is the only picture of a company they ever see. In Taiwan, 112 of 255 companies screened leave too little public evidence to assess.
Investors, corporate buyers and procurement teams build a shortlist from public material long before a meeting exists. A company absent from that material is not rejected on the merits. It never reaches the shortlist.
Prospective partners, joint-venture counterparties and senior candidates all check what an organisation says about its own AI work. Silence is read as inactivity, which is often wrong and rarely without cost.
When someone asks an AI system what a company is doing with AI, the answer is assembled from public sources. A company that publishes nothing lets other sources describe it.
“Most companies we look at are doing more with AI than their customers, investors and partners can see. The work is real, the evidence of it is thin, and people decide on the evidence.”
AIR APAC, Q2 2026 market reading
A visibility score is something a company here can move, because every signal behind it is something it already controls. Three steps separate it from a record that shows what it is actually doing.
Every scored company in Taiwan appears in the dataset with its five signal scores and an evidence grade. A grade C or D row is usually a company that has not said much, not a company that has not done much.
This index reads the outside only. Leadership alignment, governance, data quality and workforce capability are invisible to it by design, and that is where the real answer sits.
Proof of capability is the region's weakest signal, because companies announce intent faster than they show work. Job posts, named leads and dated case studies move the score honestly.
None of this requires a new AI programme. It requires publishing the one that already exists.
The five signals use different scales, so the longest bar shows how the Index works more than how Taiwan performs. Read against its own other four signals, Taiwan is relatively strongest on platform and data footprint and weakest on capability visibility. That weakest signal is the clearest place to start.
13 companies in Taiwan publish evidence of AI capability, with an average of 60 of 100 on this signal. Where this evidence exists, it lifts the score. 90 percent of scored companies have not published it yet, against 83 percent across the region.
Work published only in languages other than English is understated here, not absent.
| Signal | Score |
|---|---|
| Average visibility across the market | 41.2 |
| Sector-stratified signal | 40.7 |
| Supplier readiness signal | 43.2 |
| Domestic adopter signal | 53.1 |
The country context blocks come from other publishers with other methods. They are shown beside the AIR APAC reading, never merged with it.
| Source | Value | What it measures |
|---|---|---|
| Oxford Insights Government AI Readiness Index 2024 | 74.58 | Country level, government readiness |
| AIR APAC public-signal visibility | 41.2 | Outside-in reading of the organisations in this market |
Benchmark years: 2023 and 2024. This market has the economy-wide benchmark. No business benchmark covers it.
A tier is a position in the published regional order, not a grade anyone awards. It says how much of a company's AI work a buyer or an investor can verify, and nothing at all about whether that work is any good.
| Rung | Label | Organisations here | Regional order | What it means |
|---|---|---|---|---|
| Tier 1 | Proven | 2 | Region order 1-100 | Publishing enough, recently enough, and across enough of the five signals that the public record alone establishes an AI position. |
| Tier 2 | Building | 9 | Region order 101-400 | A consistent public trail, usually strong on two or three signals and thin on evidence of hiring. |
| Tier 3 | Findable | 30 | Region order 401-1,000 | Credibly active and possible to verify, but sparse enough that it takes a deliberate search to find. |
| Tier 4 | Scored, not listed | 95 | Beyond the published 1,000 | Clears the evidence threshold but does not reach the published list. Real companies, scored and unlisted, and the largest group in almost every market. |
2 of the 136 companies in Taiwan that clear the threshold reach tier 1, and 95 are scored but not listed, 70% of those scored.
Moving up a tier is a disclosure decision, not a technology one, because every signal this index reads is something a company already controls.
From the published APAC 1,000, at the same scores and the same evidence grades as the regional table. Each one has its own page. Shown in published order. No position can be bought.
| Order | Organisation | Sector | Tier | Score | Evidence |
|---|---|---|---|---|---|
| 28 | Chunghwa Telecom | Tech/Fintech | Pacesetter | 70.0 | B |
| 48 | Cathay United Bank | Financial Services | Pacesetter | 68.0 | A |
| 160 | ASUS | Manufacturing | Momentum | 62.0 | B |
| 212 | TSMC | Manufacturing | Momentum | 62.0 | B |
| 216 | Wistron | Logistics/Supply Chain | Momentum | 62.0 | B |
| 218 | Yuanta Commercial Bank | Financial Services | Momentum | 62.0 | B |
| 305 | KKCompany Technologies | Professional Services | Momentum | 59.0 | A |
| 309 | LINE Bank Taiwan | Financial Services | Momentum | 59.0 | C |
| 328 | Taiwan Secom Co., Ltd. | Professional Services | Momentum | 59.0 | B |
| 350 | eCloudvalley | Professional Services | Momentum | 58.0 | B |
| 373 | Soft-World International Corporation | Professional Services | Momentum | 58.0 | B |
| 411 | MediaTek | Tech/Fintech | Emerging | 57.0 | B |
| 453 | Delta Electronics | Professional Services | Emerging | 56.0 | B |
| 455 | First Financial Holding | Financial Services | Emerging | 56.0 | A |
| 464 | Hua Nan Financial Holdings | Financial Services | Emerging | 56.0 | A |
| 475 | LITE-ON Technology Corporation | Manufacturing | Emerging | 56.0 | B |
| 482 | Nan Shan Life Insurance | Financial Services | Emerging | 56.0 | B |
| 505 | Taishin International Bank | Financial Services | Emerging | 56.0 | B |
| 512 | Winbond Electronics Corporation | Manufacturing | Emerging | 56.0 | B |
| 515 | Yuanta Financial Holding | Financial Services | Emerging | 56.0 | A |
Two known gaps apply to Taiwan this cycle. Evidence comes from English-language search only. The country context uses one of the three planned source types, because no business benchmark covers this market.
A high visibility score means a company publishes a lot that can be found about its AI work. It does not mean the work is good, in production, or making money. A company here running serious AI quietly will score low here, and it should: this index reads the public record and nothing else.
Three improvements are planned and documented in the method: fuller source coverage for each market, a tested coverage weighting, and merging companies that use several websites. The next edition, the APAC 5000, is published once a year.
Every limitation the Index carries is on the limitations page, and the method itself is on the methodology page.
The dossier
AI Readiness Index, Q2 2026 results, Taiwan
PDF, 20 pages, 4.6 MB. Open, ungated, no form. In the dossier PDF, Emerging refers to the group labelled Scored, not listed on this page.
The data behind it
The published 1,000, as a table
CSV, 1,000 rows. Open, ungated, no form. The first column is the order each organisation holds in the published list for this edition. No position is awarded and none can be bought.
A correction
If a row here misreads your organisation, the route to correct it is the contact page.