If you closed Thursday with $10,000 in Direxion Daily Technology Bull 3X Shares (NYSEARCA:TECL), you opened the weekend with about $8,000. TECL closed Friday June 5 at $202.59, down 19.93% from Thursday’s $253.01. The underlying it leverages, Technology Select Sector SPDR Fund (NYSEARCA:XLK), fell 6.66% from $193.17 to $180.30, its worst session in over a year. The math is clean, and it is exactly what TECL is built to do.
The Arithmetic of a 3x Wrapper on a Bad Day
TECL is engineered to deliver three times the daily return of the Technology Select Sector Index, the same index XLK tracks. When XLK drops 6.66% in a session, the wrapper is supposed to spit out roughly triple that, and it did. The 19.93% single-day loss is the leverage doing its job, not a malfunction or a tracking blowout. There is no asterisk and no fee story large enough to explain it. The whole event lives in two numbers, the sector index and the multiplier on top of it.
TECL is still up 72.62% year-to-date even after Friday, and up 182.33% over the past year. The ten-year return for the fund is a sci-fi 5,691%, against 810% for XLK over the same window. So the headline loss is real and so is the prior run. Both are products of the same machinery.
Why XLK Moves Like One Stock Some Days
The mechanism behind Friday is concentration meeting a catalyst. XLK’s top three holdings, NVIDIA (NASDAQ:NVDA), Apple, and Microsoft (NASDAQ:MSFT), make up 40% of the fund. Add Broadcom (NASDAQ:AVGO | AVGO Price Prediction) at 5.38% and you are at 45.38% in four names. The remaining 60-plus holdings cannot meaningfully offset a mega-cap shock when one happens. On Friday, two of those names took the index down by themselves.
Broadcom was the spark. Its June 3 earnings were strong on paper, with Q2 revenue of $22.19 billion, AI semiconductor revenue of $10.8 billion, up 143% year-over-year, and Q3 AI guidance of $16.0 billion. CEO Hock Tan called out "accelerating growth in AI semiconductor revenue". The problem was the bar. The Street had been priced for closer to $17.2 billion in Q3 AI semis, and Tan signaled Google may use multiple chip suppliers, which crimps the custom-accelerator story. AVGO finished Friday at $385.73, down 7.92% on the day and 13.66% on the week.
The second pressure was rates. Friday’s nonfarm payrolls printed at 172K versus 80K expected, sending the 2-year yield to a 16-month high of 4.16%. The 10-year sits at 4.47%, and the 10Y-2Y spread compressed to 0.38%, the lowest of the year. Long-duration growth stocks hate that combination, and the most expensive AI names sit at the long end of the duration spectrum. NVIDIA fell 6.20% to $205.10. Microsoft slipped 2.66%. Alphabet (NASDAQ:GOOGL) held up best, down only 0.98%, which is its own tell about where investors think the AI capex pie is going.
Pile a Broadcom guide-down on top of a hot payrolls print on top of a 45% top-heavy index, multiply by three, and you get TECL’s Friday. The Reddit reaction tracked the math. NVDA sentiment on wallstreetbets sat at 32, bearish by Friday evening, and the day’s most viral post was titled "Fuck everyone. 4.5 years to get to a million without options and I’m done!" with 15,729 upvotes. That is what capitulation looks like in 2026.
The Breadth Problem Underneath All of This
The deeper issue is that the rally carrying TECL up 72.62% year-to-date has been narrow. Only 43% of S&P 500 stocks rose in May, versus 64% in January. Strip out AI-linked names and the S&P 500 was up 2.4% through May, versus 11% with AI included. Put/call ratios going into the week were at the lowest readings since the 2021 meme-stock frenzy and the late-1990s tech bubble. That is the setup that makes a single bad guide cascade. If everything you own is four stocks and one of them sneezes about hyperscaler chip suppliers, the leveraged wrapper does not care about your thesis. It only cares about Thursday’s close versus Friday’s close.
The VIX closing at 15.40 on June 4, in the 15.6th percentile of the past year, tells you how unprepared the options market was for what came next. Cheap volatility plus narrow breadth plus a 3x daily reset fund is a recipe whose ending is always the same when the catalyst lands.
What Has to Hold for the Bounce, and What Has to Break for More
Friday tested two pillars at once, even if it did not close a regime. The AI capex pillar took its first real crack in months, with Tan’s commentary opening the door to multi-vendor sourcing at Alphabet, which is the customer the AVGO bull case has been leaning on hardest. The rate pillar shifted from cuts-are-coming to maybe-not, on a single payrolls report that the BLS itself marked preliminary.
Three things are worth watching, and they map directly to what broke. The first is hyperscaler capex commentary out of Microsoft, Meta, Alphabet, and Amazon at the next earnings cycle. Microsoft’s AI business is already running at a $37 billion annual run rate, and Alphabet just guided capex to $35.67 billion, up 107% year-over-year. If those numbers hold or go higher, Friday becomes a digestion event. If any one of them flinches, the AVGO story spreads. The second is execution on Alphabet’s $80 billion stock sale to fund AI capex, which is a real-world stress test of whether the market still wants to fund this buildout at current valuations. The third is the 2-year yield. If it stays above 4% and the payrolls revision does not soften, the duration trade does not reset, and the most expensive parts of XLK keep paying for it.
For anyone holding TECL specifically, the structural feature to remember is that 3x daily leverage compounds in your favor on a trend and against you in chop. Year-to-date it has delivered roughly 2.9 times XLK’s gain, which is close to the advertised ratio, because the tape trended. If June turns into a sideways grinder where XLK swings 2% on alternating days, TECL will bleed even if XLK ends flat. That is the part the chart never shows until it has already happened. The question for next week is whether the four stocks that dominate this index get any room to breathe. Watch the capex talk and the 2-year yield. Everything else is noise getting multiplied by three.
Top Gaining Stocks
Top Losing Stocks