🎮

LostInConsoles

Your Gateway to Retro Gaming Reviews

Investigations

Macronix's Game-Card Revenue Rose Two Straight Quarters — the 26%-to-5% "Collapse" Was a Denominator

By Nia 7 min read Industry Hardware Console Economics AI Memory
Macronix's Game-Card Revenue Rose Two Straight Quarters — the 26%-to-5% "Collapse" Was a Denominator

On 4 November the story of Nintendo's cartridge supply chain was written in four headlines. "Nintendo's biggest ROM chip supplier is bowing out of the business," Ars Technica reported, "and what that could mean for physical Switch game costs." Tom's Hardware put the destination in the headline — Macronix quits to focus on chips for AI data centers. PC Guide went straight to the consequence: "Nintendo Switch cartridges could face shortages as major ROM supplier reportedly ends production."

Every one of them rested on the same two numbers from Macronix's third-quarter release: ROM was 5% of net sales, down from 26% a year earlier. Both numbers are correct. But a share is a ratio, and nobody published the other half of it.

The table, with the denominator restored

Macronix reports net sales and a four-line product mix every quarter, and the releases are still on its servers going back to the start of 2025. Put them side by side and the segment that "collapsed" looks different from the share that was quoted.

Quarter Net sales (NT$m) ROM share ROM revenue, derived (NT$m) ROM, QoQ as reported ROM, YoY as reported
Jan–Mar 2025 6,137 14% ~860 −37% −17%
Apr–Jun 2025 6,799 13% ~884 −2% −4%
Jul–Sep 2025 8,214 26% ~2,136 +144% −5%
Oct–Dec 2025 7,729 12% ~927 −56% −34%
Jan–Mar 2026 10,469 8% ~838 −9% −4%
Apr–Jun 2026 19,125 6% ~1,148 +40% +37%
Jul–Sep 2026 24,237 5% ~1,212 (see note) +9% −39%

Derived column = net sales × the printed share. Macronix rounds shares to the whole percent, so the derived figures carry an error of a few percent; the QoQ and YoY columns are the company's own figures, computed on unrounded revenue.

Read the last two rows against the first two. ROM revenue in July–September 2026 was about NT$1.2–1.3 billion a quarter. In the first half of 2025 it was running at about NT$870 million. The business that is supposedly being abandoned is, in absolute terms, roughly half again the size it was eighteen months ago, and it has now grown for two consecutive quarters — +40% in April–June and +9% in July–September, on the company's own sequential figures.

The cross-checks tighten the last row upward. A printed 5% means somewhere between 5.0% and 5.4% of NT$24,237 million, and the −39% year-on-year figure implies the same quarter sat at NT$1.28–1.31 billion. The honest statement is that the 5% is really about 5.3%, and the line is worth about NT$1.3 billion — not the NT$1.2 billion you get by multiplying the rounded share.

What actually fell was the share

The arithmetic of the share is not subtle once the denominator is on the page. Total net sales went from NT$8,214 million to NT$24,237 million in a year, an increase of 195%, because NAND revenue rose 768% and NOR rose 158% — both sold into the AI build-out. A segment that held perfectly flat would have seen its share fall from 26% to about 8.8%. Add a real 39% decline against an unusually large base and you land on 5%.

That base matters, and it is the second thing the coverage left out. ROM did not glide from 26% to 5%. It spiked: +144% sequentially in July–September 2025, to NT$2.1 billion, then −56% in the following quarter. A one-quarter doubling and immediate collapse is the signature of a build against a dated release, not of a market finding its level. The comparison that produced the "collapse" headline is a quarter measured against the top of that spike.

Why the exit is still real, and still not about demand

None of this makes the exit fiction. Macronix's chairman told the third-quarter call that NOR demand from data centres is "very robust" and that the company would phase out ROM production, and the numbers support the logic: gross margin was 70.5% in the quarter, operating margin 53.4%, net income NT$10,951 million, and gross profit up 1,436% year on year. Every wafer of capacity that can be moved from a mask-ROM line serving a toy-and-cartridge market into a 300-layer NAND or data-centre NOR line earns several times as much.

That is a capacity decision, not a verdict on physical games. It is the difference between a supplier that has lost its customers and a supplier that has been outbid for its own machines. The first means Nintendo's cards are dying of neglect; the second means they are becoming a commodity that has to compete for foundry time against the most profitable memory market in a decade.

The distinction has a price tag, and Nintendo is already paying it. The 64GB full game card was, at launch, reportedly the only full-size format developers could get, and publishers have said the cost of a real card run adds roughly €10–15 to a title's shelf price. The game-key card — a cheap plastic token that carries a licence instead of a game — is what that price difference looks like from the publisher's side.

The demand side is not the problem

Nintendo's own interim filing, published the day after the Macronix coverage, describes a software market for the older console that the card line still serves. In April–September 2026 Nintendo sold 24.49 million Switch 2 games and 65.73 million Switch games; 72.9% of everything it shipped ran on the machine whose cards are ROM-based. Tomodachi Life: Living the Dream sold 8.81 million copies and Rhythm Heaven Groove 2.89 million in the half, both Switch titles, both card-bearing releases.

A supplier leaving a business whose largest customer just sold 65.7 million units of software in six months is not watching demand evaporate. It is choosing a better use of its fabs.

What this does not prove

ROM is not only Nintendo. Macronix sells mask ROM into other applications, and the segment line is not a Nintendo line. The correlation with Nintendo's release calendar is suggestive, not demonstrated — the July–September 2025 spike is consistent with a large first-party build, but Macronix does not name customers and neither does Nintendo.

Revenue is not units. A shift toward larger-capacity cards raises revenue per card without any increase in card count. The two quarters of growth are consistent with more cards, bigger cards, or both; the disclosure cannot separate them.

The exit is still an exit. Two quarters of growth do not guarantee a third. Macronix has said it will phase ROM out, and a supplier can grow a line while winding it down if the remaining customers are willing to pay for priority.

The share is rounded. Every derived NT dollar figure in the table above inherits the whole-percent rounding of the printed share. The direction of the series is robust; the exact values are not.

Confidence

Claim Label Confidence
ROM was 5% of 3Q26 net sales, 26% a year earlier Observed (company table) 99
Net sales rose 195% YoY on NAND/NOR Observed (company release) 99
ROM revenue bottomed in 1Q26 and rose in 2Q26 and 3Q26 Observed (company QoQ figures) 92
ROM revenue is roughly NT$1.28–1.31bn in 3Q26 Inference (share rounding + YoY cross-check) 85
The exit is capacity reallocation, not demand collapse Inference (margin, mix, chairman's stated rationale) 80
Physical card risk is now a price/allocation risk Analysis 72

The story the industry told on 4 November was that Nintendo's cartridge supplier is leaving because the cartridge business is shrinking. The table says the business grew in each of the last two quarters, and that what shrank was its share of a company that found something more profitable to make. Those are different risks, and only one of them is about whether people still buy game cards.

Sources

More Guides