# AI capex — X 热门讨论 (2026-09-18 03:00 UTC)
## @SureshKBN (Suresh K) · 09-18 00:26 · ♥110 ↻6 💬4 At any given point, I think in 3 scenarios — Bear, Base and Bull.
Among OFC players, I believe STL Tech is relatively better protected. I’ve tracked the company for nearly 2 years.
Look at what happened when the US imposed an additional 25% tariff on India — margins took a hit of nearly 760 bps, yet the company was still able to sail through it.
So I don’t look at STL Tech only through the next quarter.
For AI/Data Center capex, I look through the lens of FY28/FY29 earnings, capacity utilisation and further expansion plans. That future earnings potential is also why the market is willing to give these businesses a premium valuation today.
And STL Tech has another important structural advantage — its preform capacity in India, which gives it deeper backward integration than most OFC players.
Bear case: tariffs/margins hurt more than expected. Base case: AI/DC demand absorbs capacity and earnings scale. Bull case: strong utilisation + further expansion + sustained AI capex create a much bigger FY28/FY29 earnings base. That is what scenario analysis means. Not predicting one outcome — being prepared for all three. > 引用 @SureshKBN: at this juncture , if there is anything i need to bet on ofc it is STL TECH .. rest all lot of ifs and buts https://x.com/SureshKBN/status/2100743249724068275
## @StockOptionCole (Cole) · 09-18 00:32 · ♥53 ↻8 💬6 $GOOGL My 1st favorite Tech setup right now
Coiling like a giant spring on the Weekly
I may increase my position soon. Targeting 410+
The bull case for Big Tech like $GOOGL $META $MSFT $AMZN is so strong
QQQ is heading into Q4 with the most bullish chart pattern out
AI Capex supercycle is still accelerating close to $800B currently and pushing $1T in 2027. AI demand will likely outpace supply in 2027. Even Chip companies are saying supply will be WORSE
Q3 earnings in October - Cloud growth is reaccelerating and every hyperscaler is guiding capex up, not down. Fears may rise but the balance sheets & numbers are proving how strong AI demand is
These companies are also constantly announcing buy backs as well
Q3 earnings NEED to beat & surprise expectations now that we’ve raised rates and hyperscalers burning 92% of operating cash flow on capex or we can bust for a little. > 引用 @StockOptionCole: $GOOGL One of the strongest Tech setups right now is sitting on the 200 Day MA and 50 Week MA support
Backbone level
This monster setup can rip this 20%+ higher over the next 4-6 months
I just took a position in this one shared with Subscribers below https://t.co/HOycfOXgmo https://x.com/StockOptionCole/status/2100744744993436016
## @AssaadRazzouk (Assaad Razzouk) · 09-18 00:48 · ♥30 ↻4 💬6 The IEA has underestimated solar growth in every single annual World Energy Outlook it's published since 2014: Actual 2023-25 generation has already blown past every one of the last decade's forecasts
Analysts did the exact same thing with AI: late-2024 capex forecasts had to be revised upward month after month just to keep up with reality
Linear institutions can't model exponential technology. Plan accordingly.
Wires chewing pipelines
https://t.co/MMJmpSp4Fo https://x.com/AssaadRazzouk/status/2100748721512485162
## @CNBCTV18News (CNBC-TV18) · 09-17 08:42 · ♥31 ↻3 💬0 'My base case is that there will be massive capital destruction in the US from the AI capex boom'
Chris Wood of Jefferies tells @_prashantnair AI capex is accelerating - but shifting to debt & a funding break could trigger a sharp unwind.
#IranWar #Nifty #BankNifty https://t.co/XBUBVzgwuQ https://x.com/CNBCTV18News/status/2100505588962894212
## @EricJhonsa (Eric Jhonsa) · 09-17 18:53 · ♥31 ↻1 💬1 The memory and semicap weakness seems correlated. Investors worried memory capex won't moon as much as previously hoped due to HBM de-specing. Think that view undervalues both current demand drivers and future ones (e.g. higher accelerator shipments, continual learning), but it's prob driving selling among headline-driven investors.
Also, seems like semicaps were popular among compounder bro and long-only types who got AI religion late and now have cold feet. And memory was quite popular with retail momo-chasers who got burned. https://x.com/EricJhonsa/status/2100659407533359548