CXMT's R&D Spending Ratio Exceeds Industry Peers for Three Consecutive Years
0xBroomberg
CXMT's IPO prospectus shows its R&D-to-revenue ratio led every comparable peer for three consecutive years, hitting 15.52% in 2025 — nearly double the peer average of 9.13% — yet the filing concedes the company still trails Samsung, SK Hynix, and Micron in process technology.
How high is this R&D ratio, exactly?
CXMT's R&D expense ratio: 49.75% (2023), 19.06% (2024), 15.52% (2025) — the highest among all comparable peers every single year.
The 2025 peer average was just 9.13%: Samsung 11.31%, SK Hynix 6.66%, Micron 10.16%.
This means → for every ¥100 in revenue, CXMT spent roughly ¥16 on R&D; peers averaged about ¥9.
That 2023 figure near 50% — real spending or a math illusion?
In 2023 CXMT's revenue was only about ¥9.1 billion. A small denominator inflated ¥4.52 billion in R&D spending to 49.75% of revenue.
In plain terms = the company wasn't spending wildly more that year — revenue simply hadn't scaled up yet, so the ratio looked extreme.
By 2025 revenue had grown to roughly ¥61.3 billion, a three-year CAGR of about 160%. The 15.52% ratio at that scale is the data point that actually matters.
What about the absolute dollar amounts?
R&D spending rose from ¥4.52 billion in 2023 to ¥9.59 billion in 2025 — nearly doubling in a single year.
Three-year cumulative R&D: ¥20.6 billion, or 21.67% of cumulative revenue over the same period.
As of end-2025: 6,259 R&D staff, 32.43% of headcount; 3,929 domestic patents and 3,043 overseas patents on file.
TSMC is at just 6.47% — does that mean TSMC under-invests?
TSMC is a pure-play foundry — it manufactures chips but doesn't design them. Chip-design R&D costs sit on the books of clients like Apple and Nvidia.
This means → TSMC's R&D ratio is structurally lower and not directly comparable to CXMT, which runs an IDM model (integrated device manufacturer — design through fabrication, all in-house).
The prospectus includes TSMC in its peer group based on business-model similarity, not head-to-head competition.
Does a higher ratio mean the gap is closing?
Samsung, SK Hynix, and Micron have decades of accumulated process knowledge — a kind of "sunk R&D" that never shows up in any single year's expense ratio.
In plain terms = money they already spent no longer appears on this year's income statement, but the technological moat it built is still there. A higher ratio is one of the few levers a late entrant can pull, but it does not automatically translate into a narrower gap.
The prospectus itself acknowledges: all three incumbents are simultaneously ramping capex, so the absolute resource gap does not shrink just because CXMT's ratio is higher.
What is the real bet behind this IPO?
CXMT's 2025 gross margin was 41.02%, roughly on par with Samsung's 39.38% and Micron's 39.79%, but well below SK Hynix's 60.41%.
This reflects a process gap that shows up directly in profitability — selling the same product category, SK Hynix captures nearly 20 percentage points more gross margin per revenue dollar than CXMT.
Whether the current AI-driven memory upcycle can give CXMT enough runway and cash flow to sustain this R&D intensity until the technology gap meaningfully narrows — that is the central wager behind the entire IPO.
Content is for reference only, not financial advice.