Finlo

Semi/Memo trade is way too crowded. As I said before, even if we are right about the AI thesis (and I think we are), you can only make a tiny amount of return since it's already a consensus. Being contrarian and right is where the alpha comes from. @aleabitoreddit made a lot of money because he found Memo/Photics before everyone knew about it. I really believe it's time to start looking at SaaS names that are mispriced by the market. The thesis is the turnaround trade like when $BRK.A bought $UNH at the lows last year: Market sold $IGV hard these months. People believe SaaS will suffer from workforce reduction so their seat-based revenue will drop; and companies could use AI to create their own tools to replace those SaaS companies. I agree. But there must be some companies that are able to survive through the cycle. What we need to find is a company: a) Has the ability to transition from seat-based to consumption-based pricing b) Can't be replaced by AI since they have their own moats. $NOW is my pick. Why? a) $NOW is already transforming to consumption-based pricing b) Decades of enterprise-level data and internal data relationships that model companies can't replicate Proof: Now Assist ACV Trajectory: Surpassed $600M in ACV in 2025, more than doubling year-over-year Entered Q1 2026 at $750M; management then raised the full-year 2026 AI ACV target from $1B → $1.5B Renewal rate: 97% As AI infra trades are taking a big hit, I believe the market will look back and reevaluate $NOW. I will keep increasing my position as I see accelerating Now Assist ACV.

Today the market kept selling AI infra： $NBIS -6.09%. $CRWV -3.41%. $MU -2.61%. $NVDA -1.20%. $AAPL +4%. $MSFT +1.5%. $AMZN +1.5%. The names with a moat outside AI: ads, commerce, office got bought back. We got a June NFP came in at just 57K vs 115K expected. Participation fell to a 2021 low. From Polymarket API, the chance of 2026 hike drops from high 62% to 47%, I don't think market is switching back on risk-off since crypto is up today. I think some of the early investors in semis & memory are starting to trim their positions a bit. Long run, I'm still bullish on AI. But near term, we might see more corrections. The first thing I'm watching is the antitrust lawsuit against the Big 3 of memory.

@zpqE8DxPeP21081 @aleabitoreddit Humanoids definitely aren’t good at this point.

@heyfinlo @aleabitoreddit 因为清洁机器人根本就不好用 我买的都在闲置

I told you making the robots look hot might be a bottleneck: Today, UBTech has now announced its first full-size hyper realistic humanoid robot girlfriend. For adults only to purchase for only ~$145,717.90. -"The U1 series emphasizes attractive appearance and is only available for adults to purchase." - "It Can Imitate Most Human Movements and Expressions, and Also Has Long-Term Memory" - "U1’s surface uses bionic materials and a gel structure to simulate skin texture, elasticity, and a warm touch." UBTECH founder, chairman, and CEO Zhou Jian said on stage that just before going up, he learned that orders had exceeded 11,000 units. Half an hour later, UBtech said the number had been updated to more than 13,361 units. Demand for attractive robots is uncomfortably high.

@weaverobotics How long will Issac take to do all the chores? And how much? I hope it's not gonna take 8 hours.

Today, we’re launching our home robot Isaac 1. Isaac 1 deliveries will begin this fall. Order yours below.

@daniel_koss Nothing is INFINITE. We have to admit margin pressure will come eventually. We can't price in infinite deman when investing since otherwise the stock price should be♾️

Let me remind all $NBIS bulls that just a few weeks ago Nebius' CCO Tom Blackwell literally told me "demand from hyperscalers is INFINITE". Nebius "only" makes 50% of their revenue from hyperscalers by choice. The margin from selling to enterprises directly is higher. Hyperscaler demand is definitely not weaking. just "trust me bro" on this one. That being said I have no idea how long market will panic irrationally over a nonissue. Sometimes it just does its thing 🤷🏻‍♂️

@CaesarCapitalz Even though there's a possibility that Big 7 don't need that much computing power, I think this really only applies to $META. Because unlike AWS, Azure, or Google Cloud, their compute has no external cloud revenue to amortize it. It's pure internal spend.

If you’re selling $NBIS because of this news, you’re never going to make it.

@ValueInvestShow I think SaaS names like $ADBE have a real advantage： they're sitting on a cash pile. Even though they don't have enough AI exposure at the moment, they can always go ahead and acquire it.

"AI is killing Adobe" Adobe just bought another AI company. $ADBE is down 44% YTD on fears that Midjourney/Runway/Topaz-style tools gut Creative Cloud. But in fact, Adobe agreed to acquire Topaz Labs: the exact AI upscaling/video tool bears cite as the threat. How? Well, when you talk about operating leverage you talk about $ADBE -Q2 FY26: record revenue $6.62B, +12.7% YoY. -Non-GAAP operating margin: 44.5%. If a moat were really under siege, the company wouldn't have the cash, or the appetite, to buy the siege engine.

Today the market is hitting the brakes on AI infra. $MU -8.98%. $NBIS -13.4%. $NVDA -1.68%. But the software names with foreseeable cash flow caught a bid. $NOW +7%. $NFLX +2.74%. Even $DUOL +6.78%. The fundamentals of those AI infra names haven't changed. But I think it's a good time to start evaluating how badly AI actually hurt those SaaS names, and take a starter position. I put 10% of my position in $NOW when it dipped below 90. I think an AI model company has very little impact on a company that already has 10+ years of enterprise data. And they don't have the incentive, the whole point of an AI model company is intelligence. If they fall behind in that race, it doesn't matter if they ship a plugin or a new feature. Their competitor will eat them alive. Some would argue AI workforce replacement makes SaaS worth less. But as I see Now Assist revenue keep accelerating (100%+ YoY, Q1 2026), I think my thesis still holds.

@aleabitoreddit This is the market noise that long-term fundamental investors should avoid. What I read from $META is that cloud computing is making money. This should be a bullish signal for $NBIS.

Basically pre-market right now with the Neoclouds like $NBIS in simpler terms. Costco: sells $5 chickens. Walmart: mass buys those rotisserie chickens and sells excess as “Walmart Chickens”. Stock Market: sells off Costco.

My question is how much more they're going to spend building this cloud. Their FCF is pretty dangerous now. TTM free cash flow already dropped to $1.2B, and 2026 capex guidance is $125-145B，raised again, mostly on memory costs. Building a cloud business is basically them finally trying to answer where the return comes from.

Finally, $META to build a Cloud Business to sell excess AI compute per Bloomberg. Thought Meta should go up against GCP, Azure, and AWS eventually to increase revenue streams. Meta up 6.65% premarket.

My 3 investment principles: 1. Quality over quantity in making decisions. 2. Do less, think more. 3. Hold winners longer.

@aleabitoreddit Right now it's mainly logistics robots that found the cost case, warehouses run them 20 hrs a day against real $18-25/hr labor. I'm curious how humanoids ever find that same cost advantage inside people's homes.

For robotics companies, I have a favorable view on Unitree and Agility Robotics, two humanoid players. And I have largest concentration in Agility Robotics, since I personally prefer US humanoid players. For upstream component exposure, I currently own: - Harmonic Drive (6324) given high content BOM on things like harmonic reduction gear. - Vishay Precision for sensors and a possible candidate for Telsa Optimus. LeaderDrive (688017) and Schaeffler I have favorable views on but don't own personally. Many of the other AI DC players, I have indirect exposure to robotics like memory. Don't recommend anyone to copy, just sharing personal positions/thoughts. Humanoid sector is large, and as seen with Goldman Sachs report that "Korean companies will command a 30% direct and indirect share of global humanoid robot production". Lot of players out there globally. This was an older report but just linking it again since you see all of them pop up in GS institutional reports + what they cover. Agility Robotics is my personal favorite as of now.

@Pirat_Nation I think the potential inflation driven by insane memory prices is the primary reason the government is getting into this. Even though consumer electronics aren't supposed to have a big impact on CPI, I don't think they want any surprises right before the midterm elections.

Samsung, SK Hynix, and Micron are facing a new federal antitrust lawsuit in the US that accuses them of working together to keep DRAM production artificially low, leading to higher RAM prices. The lawsuit claims the companies limited supply while demand continued to grow. According to the complaint, this allowed memory prices to rise much faster creating a “RAMpocalypse.” The lawsuit seeks class-action status and asks for damages on behalf of businesses and consumers who allegedly paid inflated prices for products containing DRAM memory.

@masason said "The AI race will be won on the ground." I think your post is exactly why data centers will need a space solution. It's not just the noise. Data centers need massive, stable power. SMR is the cleanest answer, but you can't put one next to a neighborhood no matter how safe they claim it is.

Residents in New Jersey and Michigan are suing data centers. Not over water. Not over power. Over the noise. A 300MW facility never stops humming. There's been no federal noise rule since the 1980s. But here's the part that should scare you. These centers register under wrapper names. Residents can't even find out who really owns the thing humming next door. You can't sue what you can't name.

I think companies won't have that many tasks that require a frontier model. For most of the work, a cheaper model will do the job. While I don't think US companies end up using Chinese models. But I think the F1 won't be common in reality we have way more Hondas than Lambos on the road lol.

Utter nonsense. Not all tokens are created equal. The quality and intelligence behind each token vary, and the outputs of leading U.S. frontier models still maintain a clear edge over their Chinese counterparts. Some people ask, “Why drive a Lamborghini just to go grocery shopping?” But remember: AI adoption among enterprises is still below 1%. There will be many more tasks going forward that require a Lamborghini. And there’s a good chance even a Lamborghini won’t be enough, we may end up needing a Formula 1 car.

@aleabitoreddit Feels like Elon's using SpaceX as SPAC for acquisition.

$SPCX / Elon Musk acquires Mesh, an optical networking startup. Which is working on 1.6T OSFP (pluggable). It’s seems they own the optical engine/packaging side of things, but likely sources CW DFB lasers off merchant companies. $SIVE is one of the more startup friendly plausible merchant suppliers as seen with Ayar to $POET? Maybe $LITE and $MTSI that were have a little history too, but less so. Regardless it’s very positive a lot of startups recently like Celestial have been acquired by Marvell. For both merchant laser supplier revenue (working with startups -> having hyperscalers like SpaceX drive revenue after being designed in already). As well as valuations from M&A desirability. Goes to show how optical interconnects is the right direction if Elon Musk is directly buying these companies.

@citrini I believe that most companies will use cheaper models for their daily jobs. Only a small amount of work that requires high intelligence will use frontier models like GPT 5.6 Sol. If that’s the case, cheaper models like DeepSeek will still take a big part of the market share.

For every SOTA model release going forward, it’s likely the government will decide which companies have access to them. Below is from OpenAI to a company that was early access for GPT 5.6 Sol

Watch the money, not the spending. Forget capex headlines and the dot plot. Watch the cost to borrow. Watch if bond sales weaken (Meta's last drew $96B, down from ~$125B). Watch if private lenders step back. And watch AI spending versus real AI revenue.

Even firms with a lot of cash are not safe. Meta in 2022. Rates were rising. The metaverse was losing money with nothing to show. Zuck cut Reality Labs and laid off staff. Costly money plus no profit forces a cut, even when you are not short of cash.

This is how an AI bubble actually breaks🧵