CFO Biweekly: 2026/6/10
Two doors opened for USD to leave Taiwan without even touching the NTD. Karl Marx’s Ghost is laughing. No one in the financial press is looking at the Third Door.
1:00 AM. Two documents spread across the desk. I finished reading, then sat up straight. Not because it was good news.
One is the FSC directive dated April 21, 2026, allowing listed companies to distribute dividends in US dollars directly to foreign shareholders. The other is the TWSE guideline dated May 29, 2026, specifying the accounting treatment and internal control framework for listed companies holding stablecoins such as USDT and USDC.
The media reported them separately. The financial press said dollar dividends are a good thing, reducing FX costs for foreign investors, “internationalization.” The crypto media said the stablecoin guideline is a boon for Web3, “the dawn of compliance.” That’s not what I saw. Within a month, Taiwan opened two institutionalized channels for dollars to leave the country, channels that didn’t exist before.
I thought of the machine that makes NT$794 million a day for the government and where its fuel comes from.
*Central Bank and CBC will be used interchangeably throughout this article.
The First Door: A Path Where Over NT$100 Billion a Year Could Directly Leave
Let’s get the numbers straight first.
According to a report by local Commercial Times on April 21, 2026, listed and OTC companies had already announced NT$2.2 trillion in cash dividends for 2025, with analysts forecasting a full‑year total of NT$2.5 trillion. On the same day, the FSC announced that foreign ownership as a percentage of Taiwan’s total market capitalization stood at 49.99% as of the end of March 2026.
Upper bound of dividends that foreign investors can receive: NT$2.5 trillion × 49.99% ≈ NT$1.250 trillion.
I ran two scenarios (personal models, not official numbers):
Upper‑bound scenario (90% choose USD): Institutional investors make up about 80% of foreign ownership; funds, sovereign wealth funds, pensions have almost no need for NT dollars; I estimate their dollar election rate close to 100%. Foreign natural persons account for about 20%, assume 50%. Weighted average ~90%. Estimated outflow: NT$1.250 trillion × 90% ≈ NT$1.125 trillion. Converted to USD (spot rate around NT$31.5 as of early June 2026, my own estimate): about $357 billion USD.
Baseline scenario (50% choose USD): NT$625 billion, about $19.8 billion USD.
Both scenarios share a common weakness: the 90% and 50% are my own judgments, with no historical comparison. The dollar dividend system is new; there is no precedent for foreign investor behavior. So these numbers, rather than predictions, are the stress‑test bounds I set for myself. Empirically, the actual outcome could fall anywhere between them. But even using 50%, NT$625 billion is not a small number.
$357 billion USD would be about 5.9% of the April 2026 foreign exchange reserves of $602.488 billion (announced by the CBC on May 6, 2026). If both scenarios’ outflows are concentrated in the July‑August peak dividend season, neither can be ignored.
The rules are scheduled to take effect in 2027 at the earliest. (Our contacts hinted it could be as early as January) The actual election rate by foreign investors is unknown. But the direction is certain.
In the past, when foreign investors received dividends, they had to be converted into NT dollars first, then back into USD. Two currency swaps, with time lags, costs, and friction. More precisely, what is reduced is not the total volume of FX, but the passive NT$ residence time of foreign capital, money that used to briefly park in the NT system due to FX friction can now settle directly in dollars, skipping the NT landing step entirely. This is a structural change in the outflow path, from resistance to no resistance.
Three eras of friction:
Before: Two FX swaps, money had to stop, time lag, costs.
After dollar dividends: One step, direct into USD account, almost frictionless.
After stablecoin goes mainstream: Zero. Dollars move on chain, never touch NT.
A premonitory sign already exists: The CBC reported that in December 2025, remittances of foreign investor earnings and dividends reached $7.7 billion, and after offsetting inflows, net remittances were $3.6 billion (CBC, January 6, 2026). That was the natural state before the rule change. After the rule change, the floor under that number will only move higher.
The Second Door: Digital Dollars on Corporate Balance Sheets — Behind Them, the US Treasury’s Financing System
The TWSE’s stablecoin accounting guideline ostensibly deals with IFRS recognition (whether USDT should be classified as a financial asset or an intangible asset). As a CFO, I look at corporate behavior.
Before the guideline, holding stablecoins was a gray area for listed companies: accountants didn’t know how to sign off, boards didn’t know how to approve, audit had no framework. Now the rules are out, holding is legalized, and listed companies have a reason to allocate.
The next step is linear logic.
Export‑oriented companies like TSMC, MediaTek, and Hon Hai receive large amounts of US‑dollar receivables every month. In the past, they had only two places to put them: dollar time deposits, or convert to NT dollars. Now there is a third option: hold USDC. According to Circle’s Q4 2025 financial report, the yield on USDC’s reserve (mostly short‑term Treasury bills) is about 3.7% (corresponding to the 3‑month T‑bill yield, actual level around 3.71% as of June 5, 2026, Trading Economics data), which is about 3 percentage points higher than dollar demand deposits, and more liquid than time deposits. A CFO’s spread calculation doesn’t need a survey.
Open up the underlying structure of this instrument: USDC’s reserves are 100% cash and government money market funds managed by BlackRock (Circle official statement, 2026), mainly holding short‑term US Treasuries. For every $1 of USDC a corporate deposits, $1 of value flows into the US Treasury’s financing system.
The whole chain is:
Taiwanese companies earn USD → put it into USDC → Circle buys T‑Bills → US Treasury funds its deficit → Taiwanese companies earn 3.7% interest
Corporations earn the spread, the US Treasury gets financing. Each side contributes, each side takes what it needs.
As of May 2026, the global stablecoin market cap was approximately $320 billion (CoinMarketCap, May 2026), continuing to expand from around $220–240 billion at end‑2024 (BIS estimated ~$255 billion in June 2025). If Taiwanese listed companies collectively allocate just 1%, that would be $3.2 billion in implicit US Treasury purchasing power. After legalization, the direction of this number requires no prediction.
The actual scale of this effect, however, depends on a key variable that cannot currently be verified: how much of companies’ dollar revenues would have been converted into NT dollars, and how much was already kept in dollar form outside Taiwan. For companies like TSMC and MediaTek, dollar receivables are far larger than their onshore operating expenses; a significant portion was already staying in the dollar system. Stablecoins give them an additional option for dollar asset allocation, not necessarily a dramatic change in their NT conversion behavior. The actual magnitude could be smaller than the surface calculation suggests. There are currently no public statistics to support this — this is the least transparent link in the entire logic chain. But the direction is clear: even if only a few percentage points of conversion behavior are affected, the long‑term trend is downward.
This is not a conspiracy theory. The design makes everyone walk through it voluntarily. Every step makes sense, but money flows in only one direction.
The Third Door: The Machine That Makes NT$794 Million a Day
This is the part I must highlight — the layer you can only see after connecting the first two.
For fiscal year 2026, Taiwan’s Central Bank has a budgeted contribution to the national treasury of NT$200.046 billion (United Daily News, May 2026; a CBC deputy governor confirmed in the legislature that this target “can be achieved,” but faces three major uncertainties: US monetary policy, international capital flows, and Taiwan’s own monetary policy; the final figure will be confirmed by year‑end).
This is the third consecutive year above NT$200 billion, the highest since Governor Yang Chin‑long took office. Converted to daily: NT$200.046 billion ÷ 252 ≈ NT$794 million per day.
Where does this money come from? The CBC says itself: mainly from FX reserve investment income. In plain English: US Treasury interest. $602.488 billion sitting there. The higher the interest rates, the larger the reserves, the more interest income, the more contributed to the treasury. That’s it.
Where do the FX reserves come from? Taiwan is export‑driven. Exporters earn dollars, but to pay domestic salaries and taxes, they must convert those dollars into NT dollars. Those dollars hit the market and put upward pressure on the NT dollar. The CBC steps in, buys those dollars, caps the appreciation, and maintains export competitiveness, as C.C. Wei has stated on multiple public occasions, every 1% appreciation of the NT dollar reduces TSMC’s operating margin by 0.4%.
The CBC takes dollars, issues NT dollars, its FX reserves grow, it invests those dollars in US Treasuries, earns interest, and contributes to the treasury.
The premise for this mechanism to continue is a steady stream of dollar‑for‑NT exchange demand in the FX market — dollars for the CBC to buy.
In the same month, the FSC institutionalized two channels that allow US dollars to leave Taiwan without ever converting to NT dollars.
Dollar dividends go direct: Foreign investors receive dividends and no longer need to convert to NT dollars first — FX demand falls.
Corporate stablecoin holdings: Dollars move on chain, never landing in NT — FX demand continues to fall (provided that these companies would otherwise have converted some dollars into NT).
The consequences of these two paths are not immediate. They are structural, slow — maybe 2 years, maybe 5 years — before they leave a recognizable trace in the CBC’s annual report. The floor of NT conversion demand in the FX market shifts downward over the long term. Upward pressure on the NT dollar diminishes. The CBC’s operational space to “buy dollars, hold down the NT” shrinks. FX reserve growth slows, interest income weakens, and the ceiling on the treasury contribution quietly moves down.
No one will write this in a policy explanation, because it would require three agencies — FSC, CBC, Ministry of Finance — to appear in the same frame. Each is doing the right thing in its own domain: market internationalization, Web3 regulation, reducing FX costs. All three are right. But together, they start shaking the machine that makes NT$794 million a day.
The above analysis has an implicit premise: the Taiwan‑US interest rate differential remains at current levels. But the huge gap between USDC’s 3.7% yield and Taiwan’s 2.1% government bond yield is not eternal. If the Fed enters a rate‑cut cycle in 2027‑2028, USDC yields could fall, narrowing the differential. That would have two effects simultaneously: stablecoins become less attractive to corporations, and the carry incentive for foreign investors to move out of NT dollars weakens. In that scenario, the “money flows only one way” trend might see a temporary reversal. This does not change the structural judgment of a long‑term downward shift in FX conversion demand, but it reminds us: trends are not smooth straight lines. When the interest rate environment reverses, there will be a window during which NT assets look attractive again, but that window is likely cyclical, not structural repair.
The analytical framework of this article also assumes a gradual, slow‑erosion process. But if the situation in the Taiwan Strait changes significantly entering 2027, the logic of foreign capital flows will completely break out of the “carry consideration” framework and enter a “safe‑asset flight” mode. In that scenario, foreign investors will remit out not because dollar dividends are convenient, but because they need to leave. For those managing large positions, tail risk is not a question of “if it will happen” but “whether you can survive it.” If that scenario is triggered, the role of the first and second doors will upgrade from “structural factors” to “accelerators”, the friction has already been removed, the path for rupture‑level outflows is wide open. This is a jump scenario that must be built into the gradual analysis framework.
Karl Marx’s Insights
We’ve talked about two institutionalized doors. Now I want to talk about the third door. It is not designed by any institution, but hidden in the logic behind all institutional design.
Why this phenomenon is Marxian
In Volume 1 of Capital, Marx analyzed capital’s motion form and presented a formula: M‑C‑M’. Money (M) transforms into commodities (C), and commodities transform into more money (M’). The key is never the commodity in the middle, the commodity is just a transitional form that capital must pass through; the key is that money returns, and returns with an increment.
Apply this to Taiwan’s FX mechanism, the structure is identical:
M (dollar capital inflow) → C (Taiwan’s labor and technology) → M’ (more dollars, settling as FX reserves)
Taiwan’s decades‑long FX mechanism is essentially an export‑driven M‑C‑M’: foreign investors and exporters bring dollars into Taiwan, use Taiwan’s engineers and production lines (C) to create value, and ultimately a larger quantity of dollars (M’) settles at the CBC as foreign exchange reserves, then flows back into the US system in the form of US Treasury holdings. The interest spread the CBC earns is contributed to the treasury, subsidizing Taiwan’s fiscal budget. That is the underlying logic of Taiwan’s development model.
Now, this circuit has gained two new exits and neither exit was opened by bad actors.
“The sum of rational choices” — Marx’s deepest insight
One easily overlooked but critically important point in Marx’s description of capital’s motion: the systemic consequences of a capitalist system do not require a centralized designer. No one needs to sit in a room drawing flowcharts. Every participant at every node is merely making their own rational choice: workers want a living wage, capitalists want profit, bankers want interest spreads, but the sum of these choices eventually constitutes a direction of capital motion that is undesigned yet stable.
That is the theoretical origin of the phrase that appears repeatedly in my analysis of the “two doors” — “Every step makes sense, but money flows in only one direction.”
The FSC opens dollar dividends: rational. Reduces FX costs for foreign investors, improves capital market efficiency.
The TWSE gives a stablecoin guideline: rational. Provides accounting clarity for corporations, promotes Web3 compliance.
Corporate CFOs put idle dollars into USDC to earn 3.7%: rational. That’s what their fiduciary duty to shareholders demands.
Foreign investors choose to receive dollar dividends: rational. They never needed NT dollars anyway, why do an extra FX swap?
No decisions that deserve blame. But when four rational choices are stacked together, an irreversible structural direction emerges.
From “friction” to “lubrication” — how institutions change the direction of capital flow
In Volume 2 of Capital, when discussing circulation time, Marx makes an observation: the turnover velocity of capital determines how many times a given quantity of capital can complete the M‑C‑M’ cycle in a given period. Any institutional change that shortens circulation time essentially increases capital’s efficiency of motion and increasing efficiency naturally directs capital toward the path of least resistance.
The machine at Taiwan’s Central Bank that makes NT$794 million a day relies for its fuel supply on one key friction: dollars must pass through NT dollars to land in Taiwan. When foreign investors receive NT‑denominated dividends, they need to swap back to dollars to leave. When exporters receive dollar revenues, they need to swap into NT dollars to pay local salaries and taxes. These FX conversion steps create a passive residence time for dollars in the NT system, and that residence time is precisely the operational foundation on which the CBC can continue accumulating FX reserves.
Now, two institutions within the same month have simultaneously lubricated two nodes of that residence time:
Node 1 (dividend payment side): The dollar dividend system allows foreign investors’ dividend income to bypass NT dollars entirely. The residence time for capital to return is zero.
Node 2 (corporate side): The stablecoin guideline allows companies’ dollar revenues to park on chain in Treasuries, without ever landing in NT dollars. The residence time for capital to enter is zero.
This is no one’s conspiracy. This is Marx’s law of capital motion as observed in the institutional field of 2026: capital will always find the path of least resistance. And institutions, often inadvertently, remove the stones from that path, one by one.
The endpoint question: When capital flows in only one direction, who determines that direction?
Marx’s M‑C‑M’ formula has an implicit endpoint problem: if capital’s motion requires repeating the M‑C‑M’ cycle, where does the incremental money (M’) go at the end of each cycle?
In Taiwan’s past version, the answer was clear: M’ settles as Central Bank FX reserves, buys US Treasuries, interest goes to the treasury. The loop was complete.
But now the circuit has two extra exits. Dollar dividends take one path: M’ is directly remitted out, the loop no longer passes through the Taiwan treasury. Stablecoins take another path: M’ parks on-chain in Treasuries, the loop not only bypasses the Taiwan treasury — it never even entered Taiwan’s monetary system.
The underlying asset of both new exits is the same: US Treasuries. Whether it’s dollars remitted from dividends that foreign investors later use to buy Treasuries, or dollars that Circle uses from stablecoin reserves to buy Treasuries, the ultimate destination is the US Treasury’s financing system. The difference: in the past, Taiwan’s Central Bank held US Treasuries centrally, and the interest income was distributed to Taiwan’s fiscal budget through the treasury contribution mechanism. Now, that income chain is being skipped.
This is not a capital competition between China and the US. This is what Marx described as the essence of capital: capital has no nationality. It simply flows toward the path of least resistance and highest return. The path of least resistance used to go through the Central Bank. Now, institutions have unintentionally opened two shorter paths.
Watch friction, not just flow. Macro analysis habitually stares at numbers (how many billions were remitted, whether reserves increased or decreased). But Marx reminds us that what truly changes the direction of capital motion is the coefficient of friction on the path. The dollar dividend rule changes not whether foreign investors remit (they would anyway), but whether the remittance path has resistance or not. The stablecoin guideline changes not whether companies hold dollars (they already do), but whether the form of holding is grey or fully compliant. The removal of these frictions is the real structural change, and the removal of friction is irreversible, you will never see a government announce “we’ve decided to restore FX friction.”
The endpoint matters more than the start. The core of the M‑C‑M’ formula is where M’ ultimately settles. Your three CFO decisions: allocating to USDC, reducing NT time deposits, and setting up early‑warning metrics look like risk management on a financial statement. But in Marx’s framework, they are all doing the same thing: ensuring that your own capital does not stand on the side of the path that disappears, as the endpoint of M’ moves from the Central Bank’s balance sheet to somewhere else. This framework tells us: rather than staring at quarterly foreign investor remittance numbers, stare at the structure of that machine’s fuel supply. When the structure changes, the numbers will eventually follow.
Below are three specific decisions I suggest, with concrete numbers. Including: the three trigger conditions for USDC, the percentage to which I’d reduce NT time deposits, and the two numbers I use as alarms when the CBC’s machine starts to stutter. I have not seen anyone else do these three things anywhere.
