Invest1 distinct publisher3 min readUpdated
The round trip through Direxion's Semiconductor Bull 3X reads as leveraged momentum rather than a view on chips, and one fund now absorbs most of Korea's net US equity flow.
The Investor · Invest desk

Compiled by The InvestorSomething wrong?How this is made
Korean retail investors bought a net $662.85 million of the Direxion Daily Semiconductor Bull 3X Shares ETF on the 12th and 13th, days after selling $664.39 million of it in a single session on August 3 and $1.63893 billion cumulatively through the 11th [1][3]. Roughly $2.3 billion of gross turnover in one triple-leveraged fund inside two weeks is not a position on the semiconductor cycle; it is a flow that follows price [1].
The buying was front-loaded: $544.46 million on the 12th, then $118.39 million on the 13th [2]. The two-day purchase was within a quarter of a percent of the size of the single-day sale on the 3rd [9]. Even after it, the fund remains a net sale for the month, with cumulative net selling from the 1st through the 13th of $976.08 million [8]. The reversal did not undo the exit; it halved it.
Scale is what makes this a market event rather than a curiosity. Korean retail bought a net $1.1567 billion of US stocks in total through the 13th, so the two days in this one ETF equal about 57 percent of the entire month's net buying [11][2]. It was roughly 13 times the $62.09 million that went into Alphabet, the second-most-bought name over the same days, according to Korea Securities Depository data cited by Seoul Economic Daily [6]. July was more extreme: $3.78586 billion into the ETF against $4.66862 billion of total net US buying, about 81 percent [7][12][3]. Named single stocks are rounding: Amazon at $193.60 million led August, then SpaceX at $175.91 million and Alphabet at $136.30 million [12].
Holdings have compounded accordingly. The position was worth $6.55903 billion as of the 12th, fourth among Korean retail's US holdings behind Tesla at $19.15557 billion, Nvidia at $18.29608 billion and Alphabet at $8.42701 billion, up from tenth at the end of April [9]. It sits $1.87 billion below third place [7]. Part of that climb is marking rather than buying: the fund's price went from $114.72 at the end of July to $145.36 on the 13th, a gain of about 26.7 percent [10][4].
The instrument matters. It tracks three times the daily return of the Philadelphia Semiconductor Index, which means gains and losses both compound off a daily reset [4]. The index fell to 10,447.49 on July 29 as peak-out arguments spread, then rebounded to 12,456.00 on the 13th, about 19.2 percent [5][5]. Three times that would be roughly 58 percent, well above the fund's realised 26.7 percent, though the disclosed price series starts two sessions later than the index low, so the gap cannot be attributed cleanly to reset drag from these figures alone [6]. What can be said is that investors trading in and out of a daily-reset product across a chop pay for the round trip twice.
Domestic leverage is rising in parallel. Investor deposits recovered to 100.0683 trillion won on the 13th, up 91.9 billion won on the day, while margin loan balances rose for a seventh straight session to 30.9262 trillion won, equal to about 31 percent of those deposits [13][14][8].
Watch whether the month-to-date $976.08 million of net selling flips positive, whether July's $3.79 billion pace is approached again, and whether the margin balance keeps extending [8][7][14]. If the index gives back the move, the same reflexivity runs in reverse, faster.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Korean retail investors bought a net $662.85 million (about 936.6 billion won) of the Direxion Daily Semiconductor Bull 3X Shares ETF on the trading days of the 12th and 13th, according to the Korea Securities Depository's SEIBro portal on the 15th.
They bought $544.46 million of the ETF on the 12th and a further $118.39 million on the 13th.
On the 3rd, the first trading day of August, Korean retail investors sold $664.39 million of the ETF in a single day, and cumulative net selling through the 11th reached $1.63893 billion.
The ETF is a leveraged fund tracking three times the daily return of the Philadelphia Semiconductor Index; gains can be roughly tripled when the index rises and losses can widen sharply when it falls.
The Philadelphia Semiconductor Index fell to 10,447.49 on July 29 amid arguments that the sector had peaked, and rebounded to 12,456.00 on the 13th.
The two-day net buying of the ETF was about 13 times larger than the $62.09 million of net purchases of Alphabet, the second-most-bought stock over the same period.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Precise registry figures, single outlet
Flow, holding and liquidity numbers are specific and attributed to institutional sources (KSD SEIBro, KOFIA), which lifts evidence quality above typical narrative reporting. But the cluster contains exactly one article from one publisher, with no independent confirmation, no fund-level or brokerage disclosure, and mismatched measurement windows for the index-versus-fund comparison that block verification of the leverage math.
Large, documented real money at work
Adoption is directly observable in settled flow and holding data rather than intent: $662.85 million bought over two sessions, $3.78586 billion in July, and a $6.55903 billion holding that moved from tenth to fourth among Korean retail US positions in four months, alongside rising margin borrowing. The scale and concentration are unambiguous; what is missing is holder counts, position durations or realised returns.
Slightly overstated causal framing
The numbers themselves are solid and, if anything, more striking than the headline implies, so the gap is small. It is positive because the article asserts a sentiment cause - eased peak-out fears pulling money into leveraged products - without any named analyst, survey or positioning evidence, and presents the 3x tracker's 26.7% gain next to a 19.2% index move without noting that the windows are not comparable. The risk side (daily-reset drag, realised losses on the round trip, concentration) is described only generically.
Low promotional pressure, no participant voice
Data originates from neutral institutions (Korea Securities Depository, Korea Financial Investment Association) rather than from the ETF issuer, a brokerage or an asset manager, and no participant is quoted or given space to promote the product - which keeps commercial incentive low. The residual incentive is editorial: a business outlet has a readership interest in dramatic retail-flow reversals, and no conflict-of-interest disclosures are provided either way.
Facts firm, interpretation thin
Confidence is moderate: the quantitative core is registry-based, internally consistent and arithmetically checkable, so the flow reversal and concentration are reliable. It is capped by single-publisher sourcing, absence of any second view on why the money moved, and the unresolvable leverage-tracking comparison.
invest
Korean retail money quit the 3x chip trade: $4.42B in July, $106M so far in August1 distinct publisher
invest
The AI moat is now a balance sheet, so price the financing and not the model1 distinct publisher
product
Waymo opens the trunk: a 5nm ASIC, a quadrillion ops, and a supplier list rivals can price3 distinct publishers
invest
The 81% Problem: AI's Star CEOs Are Polling Badly With The People They Need To Hire1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
en.sedaily.com
1 article · August 15, 2026