Margin Tops Its 2000 Peak as the Fed Tightens Again: Anatomy of Expensive Money on 800 Billion in Leverage

Taiwan's margin balance (money borrowed from brokers to buy stocks) has crossed a threshold not seen in a generation: about NT$8.1 trillion for listed plus OTC , with about NT$6 trillion on the main board alone . The last peak was around NT$5.9 trillion in 2000 and stood untouched for over two decades. Right as that record broke, the US Federal Reserve (Fed) delivered a unanimous 25 basis point hike — the first in almost three years — lifting the target band to 3.75%–4.00% . Leverage is at a historic high while the price of money turns up at the same time. Mr. Karl's live show frames the puzzle around that clash — appetite for debt is expanding while funding tightens, and the spread squeezes the market equation (funds + psychology) directly.

What the Dot Plot Really Says: 16 vs 4 Split

The dot plot is the Fed's map of where each participant thinks rates should be — not a promise, just individual expectations plotted as dots. Among 18 voting participants, 16 expect at least one more hike by year-end , with a median around 4.1% , implying another 25 basis point step in 2026 . The tell is the 2027 column: four members pencil in cuts next year . In the same table, hikes and cuts sit side by side. Think of drivers at an intersection where most step on the gas while a few hover over the brake — the split (divergence) itself is the signal. A hike majority with a dovish minority is often an early hint that the economy may not carry this tightness comfortably .

On inflation, the Fed's preferred gauge is PCE (personal consumption expenditures price index) . The 2026 median moved from 3.6% in June to 3.7% in September — direction up. More striking is the long run: the 2028 target was nudged from 2.0% to 2.1% . A tenth of a point sounds small but the message is large: the Fed is openly conceding that even in 2028, durable 2% stability is not assured under current conditions. Chair Kevin Warsh said the move was not tightening but taking a little accommodation away — still pressing the accelerator, just less hard . If inflation does not come down, a one-and-done hike is unlikely; the bias is toward a cycle .

Why Markets Price More: 2-Year at 4.75% and 10-Year at 5%

The bond market (where real money votes) is more hawkish than the statement. The US 2-year Treasury yield stood at 4.75% at the September 18 close, a full 100 basis points above the policy floor of 3.75% . The two-year yield is investors' bet on the rate path over the next two years ; if cuts were expected, it would sit below the policy rate. It sits above — markets are pricing at least three more hikes on top of the one the Fed guided . The 10-year touched 5%, its highest in three years and the highest since 2007 in a longer view, topping a volatile 18-month stretch. That is why commentary notes no cuts visible before 2028 : the Fed says one step, markets price a sequence .

Why is it expensive ? Because rates are not set by one committee but by the global pool of borrowers , and three large taps are open at once. First, AI companies : US$274 billion in corporate bonds from January to August 2026 , versus US$160 billion in all of 2025 and just US$36.1 billion two years earlier — data centers, power and chips funded largely with debt. Second, the US Treasury : fiscal deficit US$1.97 trillion through August , with interest alone topping US$1 trillion, above the defense budget . Third, the buyer base : foreign share of US government debt fell from about 50% in 2008 to about 30% now . More borrowers chasing fewer savers naturally lifts the price (the interest rate) . The story is not 25 basis points; it is this triple squeeze .

The Fed's Unspoken Third Point: On the Side Without Assets

The most debated line came on a follow-up question. Warsh recalled the Fed's duty to the least well-off , and when an NBC reporter asked who that meant he answered plainly: Americans without financial assets — roughly half the population , with no home equity and no 401(k), living paycheck to paycheck every two weeks . Without naming it, he said only with price stability can their real wages rise . Mr. Karl's reading, shared as personal interpretation, is that this cycle is calibrated for those without assets ; the implication whispered between lines is that a fall in stock or house prices would not be treated as the Fed's problem . Markets have relied for decades on the Fed put (the rescue insurance on a selloff) — the message now is no policy insurance this time . In Taiwan terms, it would be like a central bank saying it will act so young people without a home can afford one , and therefore a price correction is not its concern . For the psychology leg of the market , that is the stiffest headwind.

Taiwan Puzzle: Rates Unchanged, Money Still Tight

The Central Bank of the Republic of China (Taiwan) (CBC) left policy rates unchanged for a tenth straight meeting on September 17, 2026 : discount rate 2%, secured-loan facility 2.375%, accommodation 4.25% . The next day, September 18, it lifted the second-home loan cap from 60% to 70% of value — for switchers who already own one home , down payment drops from 40% to 30% . At first glance easing, Governor Yang Jinlong framed it as flexibility within tightness : policy is slightly tight and that tightness is run through quantity and reserve tools — required reserves and bank bills — not the price (the rate) . Headline says no hike , but the quantity tap is already tightened .

M2 (cash plus demand plus time deposits) has a reference corridor of 2.5%–6.5% per year . The January–July average in 2026 was 6.6% and June hit 8.13% , above the band. It looks like plenty of money, but the context flips it. CBC's 2026 forecast: real GDP growth 11.48% plus CPI 2.03% = 13.51% nominal expansion , against M2 growth 6.75% . The economy is outgrowing money — M2 to GDP fell from 2.51 in 2023 to 2.07 in 2026 and velocity is rising . For comparison, in 2020 inflation -0.23% plus growth 3.19% = 2.96% while M2 grew 5.84% and money outran activity — that year was loose, this year is tight. Why did the corridor breach? Yang points to two drivers: strong AI exports and domestic investment appetite plus buoyant equity trading generating transaction demand. The P* model (a framework for whether money fuels inflation) shows the price gap near zero in Q2 — the rise funds faster activity, not inflation .

Where the cash went is clear in two channels. On the household side, secured revolving loans (loans collateralized for personal finance) grew 32% year on year in July and the margin balance stood at NT$8.035 trillion at end-June , with the live update putting it above NT$8.1 trillion — above the 2000 record , after NT$6 trillion in April and NT$7 trillion in May were breached in sequence. On the corporate side, AI-related manufacturers added NT$911.3 billion in working loans in the first half , and total manufacturing working capital reached NT$2.28 trillion, 46% of the sector . State banks added NT$104 billion in a single month . As a result lending grew 12% while deposits grew 9.6%; what went out outpaced what came in . The chart split is telling: corporate plus equity loans up 14.3% while real-estate share fell to 34.44% from 35.56% in March and the June 2024 peak of 37.61% .

The squeeze already shows in market rates . Taiwan's 10-year government bond yield rose from about 1.47% last year to 1.92% this year, the 5-year from 1.80% to 2.23% ; the CBC 364-day bill went from 1.39% to 1.63% in August and other tenors from 1.53% to 1.71% — roughly a 20% to 40% lift . The weighted average for new loans at the five major banks was 2.167% in July, up 0.012 point month on month ; new mortgages at the five majors are quoted around 2.29% and some large banks are queuing applications as quotas filled . Concrete math: a NT$10 million mortgage over 30 years at 2.29% costs about NT$2,600 more per month if rates rise 0.5 point and NT$5,300 more if up 1 point . Easing the second-home cap 60%→70% trims down payment but does not trim cost or queue ; the grace period (interest-only early years) feels easy at first but principal piles up later - and likely at higher rates . So affordability math, not flexibility, is the test .

The last piece is the growth path . CBC projects 2026 at 11.48%, the strongest in 39 years, then 2027 at 5.8% — almost halved . Why? High base effect : this year is so hot that keeping the same speed is mathematically hard. The issue is not contraction but what valuation is priced on . Multiples today embed earnings expectations (EPS) for an 11.5% year , yet will be weighed on a 5.8% year . Mr. Karl's simple formula: market = funds + psychology . Funds are already getting tighter, psychology still lives in the 11.5% year ; when both turn down, high-level thinking (staying defensive at the top) is needed. Not a crash call but positioning: trim leverage, keep some cash, avoid chasing new highs , favor dividend-paying names , test +1 point scenarios for both rent and mortgage, and watch two gauges weekly — the US 2-year yield and Taiwan's 10-year yield ; as long as they climb, money gets tighter . As a side note from the final part of the show, gold futures net longs sit at the same level as late 2025 while price slipped from the 5,200–5,300 zone to 4,300–4,400 — leaving potential headroom even while rates rise; a Baltic Dry Index (BDI) breakout also flags shipping names — even in a hiking cycle the diversification story continues .