An investor who allocated RM10,000 to every one of the 39 tradable Bursa Malaysia IPOs in 2026 — Main Market and ACE Market, no selection, no exclusion — holds RM613,557 today against RM390,000 deployed, a blended return of 57.3%. Run as a blind, equal-weighted cohort study, the headline number resembles a market functioning well: broad participation, rewarded.
The composition of that return does not support the same conclusion. Nineteen of the 39 positions — 48.7% of the sample — are below their allocated purchase price as of this writing. The entire RM223,557 of profit generated by the cohort is concentrated in five issuers, representing RM50,000 of the RM390,000 deployed. This note sets out the ranked return data, quantifies the concentration, and identifies the underlying driver connecting the winning names.
Return by issuer, ranked — full cohort (n=39)
Returns are calculated against the allocated IPO price, not the first-day opening trade, on the basis that allocated cost is the relevant reference point for capital actually committed.
Profit concentration: the cohort's return is not broad-based
Of the RM223,557 in total gains generated across the 39-name cohort, the distribution by contribution is as follows:
| Segment | Capital | Profit contribution | % of total profit |
|---|---|---|---|
| Top 5 issuers | RM50,000 | RM174,001 | 77.8% |
| Remaining 34 issuers | RM340,000 | RM49,556 | 22.2% |
| Full cohort | RM390,000 | RM223,557 | 100.0% |
Top five and bottom five issuers by return
Top 5 — Ranked by Return
| Issuer | Return |
|---|---|
| Kee Ming Group | +594.7% |
| Ambest Group | +460.0% |
| Stratus Global Holdings | +273.8% |
| SkyeChip | +233.0% |
| Sum Technology | +178.6% |
Bottom 5 — Ranked by Return
| Issuer | Return |
|---|---|
| Hock Soon Capital | -55.8% |
| SBS Nexus | -50.0% |
| One Gasmaster Holdings | -42.0% |
| Enest Group | -34.6% |
| Manforce Group | -31.6% |
Findings
Issuer naming and headline sector convey no information about the winning theme
Kee Ming Group and Ambest Group — the two top performers, +594.7% and +460.0% respectively — are an electrical contractor and a precision machining/fabrication operator. Neither is classified or marketed as a technology or semiconductor issuer. Both function as suppliers into the physical build-out of fab and data-centre capacity, generating revenue independent of which specific chip or cloud platform ultimately wins share. The return data indicates business-model exposure to the build-out, not sector label, is the relevant screen.
Retail oversubscription is not a predictor of post-listing return
One Gasmaster Holdings was oversubscribed 25.24x ahead of listing — a demand signal reported across four separate outlets at the time — and is down 42.0% since. Oversubscription measures allocation demand at the IPO price on listing day; it carries no information content about business quality or valuation sustainability beyond that date. Treating oversubscription multiples as a quality signal for post-listing holding periods is not supported by this cohort.
Equal-weighting produced a return profile driven by concentration, not diversification
The 57.3% blended return is frequently the number cited when this cohort is discussed publicly. Per Exhibit 2, that figure is carried almost entirely by five of 39 positions; the other 34, unselected and unweighted, returned 14.6% in aggregate — below most benchmark alternatives over the same period. The headline return is a function of the sample happening to include the winning names, not of diversification generating the outcome.
IPO pricing independently converged on the same thesis underlying national industrial policy
Malaysia's National Semiconductor Strategy had attracted RM85 billion in approved investment as of December 2025, with Penang alone drawing RM4.9 billion in Q1 2026.1 The IPO market — without policy coordination — allocated 77.8% of total cohort profit to issuers exposed to that same build-out. Convergence between an uncoordinated capital market and a stated industrial policy target is, in our view, a stronger signal than either data point alone.
The binding constraint on this thesis is resource availability, not capital availability
Data centres across Johor, Selangor and Negeri Sembilan consumed an estimated 28.68 million litres of water per day as of January 2026, a figure projected to grow more than fifteenfold by 2030.2 Johor has suspended new approvals for two categories of water-intensive facility through at least mid-2027. Separately, the national semiconductor strategy's own stated target — 60,000 additional skilled engineers by 2030 — is itself an acknowledgment that the current talent pipeline is insufficient. Capital commitment has not been the limiting factor in this cycle; physical and human infrastructure has.
Global context: a regional allocation within a larger capex cycle
The four largest hyperscale operators — Amazon, Google, Meta, and Microsoft — guided combined 2026 capital expenditure to approximately US$725 billion, up roughly 77% year-on-year, with approximately three-quarters allocated specifically to AI infrastructure.3 That expenditure requires physical fulfillment: chip design, fabrication, testing, packaging, and data-centre construction and operation. Malaysia's semiconductor-services IPO cohort represents one market-priced, geographically specific slice of where that capital is being physically deployed.
A necessary qualification: Malaysia's role in this chain remains predominantly back-end — assembly, testing, and packaging — rather than chip design or front-end fabrication. The national strategy's stated ambition is to move into IC design and advanced packaging; this is a stated target, not a demonstrated capability shift. Four of the five top-performing IPOs in this cohort are exposed to the physical construction of semiconductor and data-centre capacity rather than the design layer. SkyeChip is the exception, operating in the higher-value design segment the national strategy is targeting.
Investment considerations
Malaysia holds more than four decades of accumulated back-end semiconductor operating history, English-language commercial infrastructure, and comparative political stability within the region. These factors positioned the existing ecosystem to absorb hyperscaler supply-chain diversification demand without requiring capacity build from zero — a structural advantage relative to newer regional entrants.
Public debate continues regarding the sustainability of hyperscaler AI capital expenditure. Michael Burry has publicly alleged that depreciation schedules understate true equipment cost by an estimated US$176 billion across 2026–2028.4 Moody's has flagged approximately US$662 billion in signed but not-yet-commenced data-centre lease commitments industry-wide.5 A correction in hyperscaler capital spending would affect the entire winning segment of this cohort simultaneously, as none of the covered issuers generate demand independent of that spending cycle.
Talent. The national strategy's own 60,000-engineer shortfall target by 2030 is a direct acknowledgment of an inadequate current pipeline. Water. Johor-corridor data-centre water usage is projected to grow more than fifteenfold by 2030; new approvals for select water-intensive categories are suspended through at least mid-2027. Power. Johor's contracted data-centre pipeline totals 8,542 MW against 0.7% colocation vacancy, with the binding constraint being substation and grid infrastructure rather than generation capacity. Budget 2027 ties new incentive eligibility to efficiency requirements, indicating policy is responding to an existing, not hypothetical, strain.6
Malaysia is one of several regional beneficiaries of supply-chain diversification, not an exclusive destination. Singapore continues to anchor the upstream research and IP layer of the regional ecosystem; Vietnam announced its first domestic chip fabrication facility in 2026. Of the three Southeast Asian IPOs exceeding US$500 million in the first half of 2026, Malaysia produced the single largest transaction, but Singapore and Vietnam each produced a comparably sized listing in the same period.
Forward view
Bursa Malaysia raised its full-year 2026 IPO market-capitalisation target from RM28 billion to RM34 billion following a strong first half — 36 listings, RM5.4 billion raised, and RM26.1 billion in incremental market capitalisation added, ASEAN-leading on both count and proceeds raised — and has indicated one to two additional large-capitalisation listings are expected in the second half.7 If the pattern observed in this cohort persists, semiconductor- and data-centre-adjacent issuers should continue to price at a premium and continue to diverge sharply from the balance of the listing calendar, given that divergence is a direct pass-through of capital allocation decisions made by US hyperscalers rather than a Bursa-specific phenomenon.
Two factors could interrupt this pattern. First, a correction in global hyperscaler capital expenditure would affect the entire winning segment of this cohort concurrently, as noted above. Second, Malaysia's own water and grid constraints — concentrated in the Johor corridor specifically — could become a binding limit on the pace at which already-announced capital is physically deployed, a timing risk rather than a capital-availability risk, though one that would present through earnings on a comparable schedule regardless of cause.