Taiwan's Exports Are Strong and the Dollar Still Won't Come Home — MacroMicro on October's Messy Markets

Contents

1. An episode that starts from “October was a mess”
2. The Fed is aiming, and still not pulling the trigger
3. Money is flowing back into bonds, and it is being picky
4. Upstream earnings came in clean, midstream got put under the microscope
5. Two supply chains have torn the currency market in half
6. They owned a misjudgement in public
7. Earnings were fine and the share price traded like a bear market
8. Is my thesis broken, or did it only break at the last link?
9. Further reading
10. One thing to take away

Wind sharp, sky high, gibbons wailing their grief; clear islet, pale sand, birds wheeling home. Boundless, the falling leaves come rustling down; endless, the long river rolls on and on.

— Du Fu, “Climbing High” (Tang, 767; translated by the author)

MacroMicro’s After Meeting EP. 220, published on 1 November 2026 and recorded on the morning of 30 October, gives a whole episode to October across stocks, currencies and bonds: Treasury yields fell from a 5.3% high back below 5%, the Taiwan dollar weakened to 32.6 against the US dollar, and China — whose GDP has missed its target two quarters running — saw the renminbi post its best run since 2022. Research deputy director Vivianna reads the macro chain as intact — strong exports, inflows, a firmer home currency — with Taiwan breaking at the final link: exporters are holding their dollars for coming investment in the United States, and by September 2026 the dollars not yet converted reached 400 billion, up 52% on the year. In the same episode she owns a misjudgement: the team expected the central bank to tighten harder, and underestimated its tolerance for inflation. The whole reading rests on the economy staying strong, and third-quarter GDP, released on 30 October at an annualised 3.4% quarterly pace, holds that premise up for now.

An episode that starts from “October was a mess”

Host Roger opens by listing the mess item by item: earnings were not bad, yet AI supply-chain share prices traded like a bear market; well-performing Taiwan has a weak currency while slow-growing China has a strong one; the best-behaved thing all month was the Federal Reserve. Vivianna comes on to unpack it, and the two of them first wander off topic: she listens every week to a news podcast with three co-hosts, and the part she enjoys most is the arguing, because their politics differ and they clash directly over the news; she says it is how she gets a break when market research starts to blur. That made me smile.

The Fed is aiming, and still not pulling the trigger

October’s pause came as expected. Vivianna found the meeting uneventful, but pulled three things out of it as guidance. The statement upgraded its description of the economy from solid to strong, and swapped “inflation reaching target in a timely way” for a line about gathering more information first. The pause was not unanimous: the vote was 10–2, with Lisa Cook casting her first dissent since 2022 and Cleveland’s Hammack likewise arguing for another hike. Asked whether the climb in market yields had already tightened financial conditions for the Fed, Chair Kevin Warsh answered that he was unwilling and unable to define market rates at these levels as tight.

The three point the same way: if the goal is to press down on inflation, financial conditions are not tight enough yet. So Vivianna sees a decent chance of another quarter point in December, with CME FedWatch then putting the odds around 80%. She quoted a line going round the market: the Fed is aiming, and aiming again, and still not pulling the trigger.

What lets her call this precautionary is the data. Third-quarter GDP accelerated from 2.2% to an annualised 3.4%, and the composition is clean: consumption contributed 2.5 percentage points, private investment jumped from 4.6% to 19.4% quarter on quarter, and inventory rebuilding added another 1.6 points, which says firms were still restocking. The PCE price index released the same day went from 0.3% to 0.5% month on month, with the core measure flat — no second round of increases yet. Her test sits here: historically the hikes that puncture a bubble are the ones that surprise and keep coming, like 2022, with no gap to stop and look around. October stopped and looked; and hiking in starts and stops gives companies time to adjust their financing plans.

Money is flowing back into bonds, and it is being picky

Mid-October’s CPI print had the market worrying about hikes and valuations, and yields fell instead of rising, lifting bond prices. EPFR counted nearly 80 consecutive weeks of net inflows into global bond funds through late October, and the bid-to-cover ratio on US 10-year Treasuries has climbed for two months to 2.77, the highest since 2014.

The interesting part is the picking. Within government bonds, short maturities beat long; within corporates, investment grade beats high yield, and high yield is still being sold. At the same time Bank of America’s fund manager survey put October cash at 3.5%, and below 4% is low — managers are not holding much cash.

Vivianna strings this together: if the market were waiting for recessionary cuts, money would pour into Treasuries regardless of maturity; if it expected stagflation, money would move into cash rather than buying corporate bonds. Neither fits, so it looks more like transitional positioning during a choppy stretch — parking money in something that yields while waiting for data. Put in cycle terms, she places us at the end of the middle leg or the start of the final leg: equities can still make highs, and the swings get more frequent.

Upstream earnings came in clean, midstream got put under the microscope

Upstream looked good, with TSMC, Micron and ASML all confirming again that money is concentrating up the chain. Intel was the exception: soaring memory prices hit PC demand and fourth-quarter guidance missed, while its data centre chip sales stayed strong — the damage landed on the non-AI part.

The cloud players midstream are another picture. Google fell 10% after reporting, Microsoft rose 3% after hours the same day, and Amazon the next day fared like Google. Yet the numbers show Google Cloud revenue up 103% year on year with operating profit three times the year-ago level. What got punished was search advertising: it is half of Google’s revenue, its growth slowed this quarter, and cash flow — negative since last quarter — slid further. Cloud is only 24% of revenue, so the biggest block naturally gets the scrutiny. Amazon has the same shape: online retail up only 2%, digital advertising up 11%, with e-commerce more than three times the size of the cloud business. Both explained themselves on their calls: Google cited currency headwinds, the end of the EU’s parcel duty exemption, and the World Cup base effect in the second quarter; Amazon cited the timing shift in Prime Day.

The information is in the box the market now checks. For several quarters the question was whether the core business earned enough to fund AI investment, and negative free cash flow was acceptable because revenue was racing; this quarter what got punished was growth in the main revenue line slowing. Vivianna treats advertising and e-commerce as cyclical businesses, and the last time big tech’s advertising and e-commerce growth peaked was the second quarter of 2021 — everyone remembers 2022 and 2023. Meta’s ad revenue slowed for a second straight quarter this time too. So the thing to watch in the fourth quarter is whether the slowdown persists once one-off factors like the World Cup and Prime Day are out of the way.

On magnitude, she cited one estimate: pure AI revenue for the three big public clouds in 2027 sums to around 500 billion US dollars, and holding big tech capital expenditure at today’s 1.3 trillion for next year without revision, that revenue alone covers half. On constitution, the large cloud operators are integrated from upstream to downstream, so a slowdown in their legacy businesses looks different from a company with no core business that funds capital expenditure entirely on leverage when financing conditions tighten.

Roger added a funny observation: the only one of the giants beating the index this year is Apple, the member of that group with the least AI and the fullest cash pile. He passed along a line he had heard — selling shovels has stopped paying, and selling fruit is where the money is.

Two supply chains have torn the currency market in half

The currency segment is the reasoning I valued most in this episode. Vivianna says the underlying macro logic has not changed — good fundamentals attract inflows and firm the home currency — and the strangeness comes from behaviour that last year’s tariff fight left behind. Most tariffs were withdrawn or never collected, but nobody knows whether another round is coming, so origin-washing has not gone away, front-loading of exports is more visible along the electronics chain, China has expanded its trade with Latin America and Africa, and the US-aligned chain has leaned towards America in response to investment policy there.

The renminbi half runs exactly as the logic predicts. Chinese export totals have kept making highs since the tariff fight, exporters took in large volumes of dollars, and China’s long-standing capital controls mean those dollars cannot leave. Onshore dollars pile up, exporters convert them into renminbi, and the currency rises into its best run since 2022.

The Taiwan and Korea half breaks at the third link. Their exporters also earned large volumes of dollars, and did not convert them home, because more investment in the United States is coming and converting back and then out again means eating the spread twice. By September, dollars not yet converted in Taiwan reached 400 billion, above last year’s level and up 52%. That is the root of the Taiwan dollar coming unstuck from exports and from Taiwan’s fundamentals: the earlier links are all running, and the last step never completes.

Policy can only do so much. The dominant factor in the exchange rate now is the adjustment in trade behaviour, with rate differentials and fundamentals waiting behind it, so policy gets as far as braking. China can cut rates and the money still cannot leave; South Korea hiked outright and its currency has shown signs of stabilising these past months; Taiwan, Vivianna admits, she cannot read — September inflation is already 2.7% where 2.5% used to look like the unwritten threshold for a hike, and Governor Yang Chin-long says the question depends on next year’s inflation.

They owned a misjudgement in public

A listener’s question was pointed: the Taiwan dollar just made a new low, weren’t you bullish on it all along? Vivianna named the part they got wrong: the team thought Taiwan’s real interest rate was too low and that, with the global hiking cycle, the central bank would tighten harder. Hiking cannot change the trend of investing in the United States, but it would have pushed back on the depreciation; what exceeded their expectation was the bank’s tolerance for inflation.

Another listener asked whether the renminbi’s strength could be the result of selling Treasuries to convert back, or of hoarding gold. She said the logic works but the links have to be checked one by one: over the same period China trimmed Treasuries, its foreign exchange reserves did not shrink but grew, which says the dollars stayed onshore, and gold is only one to two per cent of the central bank’s assets. Two links fail, so neither path explains the move. I found this more useful than the conclusion: one data point lining up does not make the whole path true.

Earnings were fine and the share price traded like a bear market

This was my actual confusion in October, watching my own holdings. The numbers were not broken, the prices fell anyway, and my first reaction was that the market had lost its mind. The box the market checks moves, as above, and that is far more concrete than “sentiment” — it also gives me two things I can copy.

So after a report now I do two extra things. First, find which line got punished — search advertising growth at Google, online retail growth at Amazon. Then listen to how the call explains that line and judge whether it is one-off or structural: the World Cup base and the Prime Day shift are one-off, advertising budgets following the economy are structural. Then write down next quarter’s check, which is what gives me something to mark against.

I have got this wrong myself: I took “revenue is still growing” as a reason to relax, without separating growth that is accelerating from growth that is decelerating. The quarter acceleration rolls over, the share price moves first and the absolute numbers in the report still look good.

Is my thesis broken, or did it only break at the last link?

Sitting on a losing position, or holding a view while the price goes the other way, the real question in your head is “was I wrong from the start?” The Taiwan dollar here is a breakdown you can copy.

Back to the third link in that currency chain: once you know that is where it broke, two responses are both wrong — throwing the chain away is wrong, because it reconnects when that link returns; waiting for appreciation is wrong too, because nobody knows which day it returns.

What I do is write a view out as three to five links and put a checkable observation next to each one; on this chain that observation is the volume of dollars not yet converted, updated monthly.

Further reading

• MacroMicro, After Meeting EP. 220 (recorded the morning of 30 October 2026, published 1 November), hosted by Roger with research deputy director Vivianna
• MacroMicro’s November 2026 monthly report on taking the market’s stress test apart, and its August subscriber report examining the Taiwan dollar’s disconnect from fundamentals
• US Bureau of Economic Analysis: third-quarter 2026 GDP advance estimate, September PCE price index
• CME FedWatch rate-hike odds, EPFR global fund flow data, Bank of America’s October fund manager survey
• Du Fu, “Climbing High” (Tang, 767, public domain)

One thing to take away

One idea: when a judgement turns out wrong, go find which link in the chain broke before deciding whether to throw the chain away. October’s Taiwan dollar makes it plain — strong exports and exporters earning dollars were both running, and the link that broke was “convert the dollars back home,” a link with its own number you can look up. Most people who got this wrong had skipped the links in the middle.

Here is something I have tried, and anything you recently misjudged will do, investing or not — assuming someone would reply to your message, assuming a habit would stick in two weeks, assuming a dinner would close the deal. Take a sheet of paper and write it as three or four links (I did A → they received B → they were willing C → outcome D), then write one line beside each link: “what sign was there at the time?” Most of the time you will see the early links held, only one broke, and that one had given a sign. The next time something like it comes round, you know which link to look at first.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.