Most traders start with the wrong question: Are you bullish or bearish? The first question should be: Is this market environment even worth deploying capital into? You can have the right directional bias and still get chopped up if the market isn’t giving you clean conditions. My framework is simple: 1. In or out2. Then long or short Directional bias comes after the decision to participate. You don’t need 20 great trades a month. You may only need 2–3 big, high-conviction trades in a month or quarter to make your year. Everything else can be scratches. That’s fine. The goal isn’t to trade constantly. Participate selectively. Then press when the environment, direction, and catalyst all line up. And Q4 is starting to look like a period I want meaningful exposure to. No need to FOMO yet. But
Data centers get most of the attention around $Bloom Energy Corp(BE)$ . But factories have the same problem: they need a lot of reliable power, and they can't afford to wait years for the grid. A new chip fab can cost $20B+ to build. Leaving that facility idle for years because utility infrastructure isn't ready isn't an option. That’s where Bloom's on-site power systems become interesting. $BE is already deployed across major industrial sites, including: • $Intel(INTC)$ foundry expansions • $Ferrari NV(RACE)$ 's Maranello factory, cutting fuel use by roughly 20% • $Quanta(PWR)$ 's California facilities, with a $502M order
$Micron Technology(MU)$ reports earnings tomorrow, and the headline numbers may not tell the whole story. The biggest thing I’m watching is HBM demand tied to $NVIDIA(NVDA)$ ’s Rubin platform. Susquehanna analysts believe some HBM4 revenue originally expected this year could slip into 2027 as Rubin ramps. That could make Micron’s quarterly results look uneven, with some revenue shifting between quarters. But that’s timing, not necessarily a change in the demand trajectory. Rubin requires enormous amounts of high-bandwidth memory, and Micron is one of the few suppliers positioned to serve that demand. That’s why the market will be listening closely to management’s commentary on Rubin, HBM4 supply and demand
Michael Burry just moved up his AI crash bet, saying the “bubble in AI may burst sooner than later.” He closed out his short positions and moved into puts: • $Micron Technology(MU)$ — June $500 puts, stock around $1,080 • $NEBIUS(NBIS)$ — June puts below $100, stock around $230 • $iShares Semiconductor ETF(SOXX)$ — Sept 2027 puts in the low $400s • $Palantir Technologies Inc.(PLTR)$ — larger Sept 2027 put position in the low $100s His thesis points to research questioning how much AI revenue is actually proven, along with warnings that memory cyclicality could return. I see the setup differently. 1.Memory is cyclical. That
$NU Is Down 38%. Here’s What the Market Is Pricing In
$Nu Holdings Ltd.(NU)$ is now in a 38 % drawdown, with four major narratives weighing on the stock: 1️⃣ Trade War US tariffs are putting pressure on two of Nu’s most important markets. Mexico sends roughly 80% of its exports to the US, while Brazil has also faced significant US tariffs. Slower growth can pressure consumer spending, loan demand and repayment behavior. 2️⃣ Iran War The conflict and disruption around the Strait of Hormuz pushed oil above $100, adding pressure to fuel and food prices across Latin America. Higher inflation squeezes household budgets and raises concerns about consumer credit quality. 3️⃣ Brazil Macro Brazil’s benchmark rate is around 14%, keeping borrowing costs elevated and weighing on spending and credit growth. This ma
$Bloom Energy Corp(BE)$ is becoming increasingly interesting as AI data center expansion runs into a major bottleneck: physical power infrastructure. Off-grid microgrids can help bypass transmission and grid-connection constraints, giving Bloom a large addressable market as data center power demand continues to grow. That helps explain why Bloom’s backlog has surged to roughly $20B. But the bigger opportunity may be what happens after each system is deployed. Bloom attaches a service contract to 100% of its systems, with contracts typically running 10–20 years. That creates a long-duration recurring revenue stream on top of the initial equipment sale. Bloom estimates this contracted service opportunity at roughly $14B. So the thesis isn’t simply abo
$Nokia Oyj(NOK)$ at $10.29 is entering a very interesting stretch. Three catalysts are landing almost back-to-back: 🧠 Oct 20–22 — NVIDIA GTC$NVIDIA(NVDA)$ ’s biggest AI event of the year puts the entire AI infrastructure ecosystem in focus. 💰 Oct 22 — $NOK Earnings The timing couldn't be tighter. And there’s another connection: $NVDA invested $1B in $NOK, taking roughly a 2.9% stake. That gives the AI infrastructure story another layer of relevance as the market heads into GTC week. 📈 Technically, $NOK is getting tight. The level I’m watching is $11. A clean break above it could open the door toward $15+. 🎯 Risk level: $9.50 🎯 Upside target: $15+ The setup is simple: Tight chart → $11 breakout → AI catalys
$NU Is Starting to Look Like a Global Digital Bank
The potential $11–13B Monzo deal makes more sense when you look at where $Nu Holdings Ltd.(NU)$ is heading. $NU has already shown ambitions beyond the Americas with its Nu Global strategy. Now Monzo could give it a ready-made platform in Europe: 🇬🇧 ~15M retail customers 🇪🇺 UK base + expansion into Ireland and Spain 💰 $NU generates roughly $1B in quarterly net income Instead of building an EU customer base from scratch, Nu could potentially use Monzo as a launchpad and accelerate expansion across Europe. Near term, the math may not look great. Monzo reportedly commands a much higher valuation multiple than $NU, so the deal may be EPS dilutive initially. But the bigger question is what Monzo could become inside Nu’s global platform. And $NU is already
Meta’s Muse Is Creating Another AI Infrastructure Demand Signal
$Meta Platforms, Inc.(META)$ ’s Muse just launched, and demand is already putting pressure on capacity. Here’s the interesting part 👇 ☁️ Every Muse user gets access to a free cloud computer. Each instance reportedly runs with: 🧠 2 vCPUs → $Advanced Micro Devices(AMD)$ 💾 8GB RAM → $Micron Technology(MU)$$SK hynix(SKHY)$ 💽 100GB SSD → $SanDisk Corp.(SNDK)$ Scale that across millions of users and the infrastructure requirements start adding up fast. And this is just one AI application. The AI race isn’t only about training bigger models anymore. Every successful AI agent or applica