Embaud

@BenBajarin Margins are expending—> rev grows faster than depreciation…. But the market is still slow to understand. It will take another 2 quarter befor noone questions capex anymore

Q on the MSFT call. The answer, my friend, is inference margins are VERY good.

@MarkosAAIG @BenBajarin If you look at it from absolute dollars of growth you get Azure +8 Bn, AWS +7Bn , Google +6Bn. Not saying Google is not doing a great job but % growth doesn’t mean the same depending on the base. For now Msft is taking market share

@BenBajarin GCP is insane.

Ok so: AWS - 28% GCP - 63% !!!!!!!! Azure- 40%

@FinnStockinger Very interesting. On the depreciation I think all huperscalers have moved to 6 years and given that V100/200 ate still running i think it’s a very fair assumption

$GOOG Google CAPEX Review While most market commentary focuses on Google’s ad revenue, the company’s true transformation lies buried in the footnotes of the 20-F filing. The 2026 financial data reveals a firm undergoing a fundamental shift from a software-first giant into a massive infrastructure conglomerate. This is no longer just about better algorithms; it is a battle for physical efficiency, power access, and capital management on an unprecedented scale. By looking past the PR narrative and into the balance sheet, we uncover the mechanical realities driving this race: from subtle accounting shifts to massive investments in energy foundations. Here is a calculated look at the numbers defining the new reality for Alphabet. 1️⃣The "Paper Billion": Financial Engineering as a Shield The most telling detail isn't in the press release - it’s buried in Note 4: Property and Equipment. Google has officially extended the depreciation cycle for its servers from 4 to 6 years. ➡️This single accounting maneuver "created" $980 million in operating income this quarter out of thin air. ➡️In an industry where AI hardware becomes obsolete in 18 months, claiming a 6-year lifespan is a massive gamble. Based on this, I conclude that management is using aggressive accounting to buffer margins against the staggering costs of the AI arms race. Without this shift, the profit "beat" would have been a whimper. 2️⃣ "Cramming" Over Building: The Time-Crunch Strategy Google’s Capex has hit a run rate of $50 billion annually ($12.01B this quarter). But look at the Management’s Discussion and Analysis (MD&A): they aren't spending on "Facilities" (buildings); they are spending on "Technical Infrastructure" (chips and racks). ➡️ This confirms a desperate time constraint. Building a new data center takes 3 years. Google doesn't have 3 years. They are gutting existing centers and "cramming" them with high-density TPU clusters. ➡️For the first time in a decade, "electricity availability" has overtaken chip shortages in the Risk Factors section as the #1 threat to growth. Google is hitting a physical ceiling. 3️⃣The TPU v6 Trap: Vertical Integration Isn't Free The filings reveal that the unit cost of the TPU v6 has surged by 22–25% over the previous generation. Moving to 3nm nodes and HBM3e memory has stripped away the "cheap" advantage of custom silicon. ➡️While names are redacted, the surge in "Purchase Obligations" for ASIC design points directly to Broadcom ($AVGO). Furthermore, massive prepayments for manufacturing slots suggest Google is effectively bankrolling TSMC’s adoption of ASML’s High-NA EUV machines to secure the upcoming TPU v7 on the 2nm node. 4️⃣Shadow Capex: The "Other Bets" Energy Play Spending in the Other Bets segment for energy projects (SMRs and Geothermal) is up 40% YoY. ➡️On-balance, I conclude Alphabet is "offloading" essential infrastructure costs into the Other Bets category to keep the Google Cloud margins looking artificially clean for Wall Street. These aren't "moonshots" anymore - they are life-support systems for the AI fleet. 5️⃣The $85 Billion Validation The only thing preventing a market panic over this spending is the Remaining Performance Obligations (RPO), which hit $84.7 billion. ➡️This is the "backlog" of signed, locked-in cloud contracts. It grew 20% YoY. This is the only metric that legitimizes the burn - Google is building the factory because the orders are already signed in blood. ➡️The Bottom Line: Google is a hostage to its own growth. They are aggressively "window-dressing" their balance sheet with depreciation tweaks to hide the fact that the machine is becoming exponentially more expensive to run. The biggest winners in Google’s current Capex cycle aren't the shareholders - it's the utility companies and thermal engineers keeping the servers from melting.

@MarkosAAIG I guess it’s also a question of value extracted by endcustomers from AI. We don’t hear uet enough about it but if demand is so high it means they are getting huge value? So they could be ready to absorb some costs hike from the hyperscalers?

*passing trough costs to customers

Big Tech and my million-dollar question. So guys, before we dive a little bit into the numbers, my million-dollar question for the CapEx raise is: is the CapEx for more growth, or for the same growth and expansion but at higher prices because of supply chain stress? And I think that’s the million-dollar question. So, hypothetical case: if Meta, for example, wants to build out a hypothetical 5 GW and has CapEx planned for that in the $115 to $135 billion area, how much of their guidance — their upward guidance to $125 billion to $145 billion range — is related to higher expenses due to supply chain stress on that 5 gigawatts? Or did 5 gigawatts become 5.5 gigawatts, just to keep it simple? That, for me, is the crucial question I want to hear between the lines in the calls. And you all should watch that. Going into the numbers: cloud accelerated really well. Azure printing steady 40%, Google with a massive beat on 63% cloud revenue growth, and AWS on 28%. Very strong. Industry experts, by the way, think AI adoption at enterprise level is around 7%. Microsoft commented and said its AI business is now at a $37 billion annual revenue run rate, up 123% year over year. Of course, you need the exact definition of “AI business” compared to that 7% enterprise adoption, but still — we are very early innings in the whole AI buildout. Also, OpenAI diversification into Trainium with their 2 gigawatt Amazon commitment is very interesting. And it’s also very meaningful for my HBM thesis, given the 2 gigawatts of Trainium and the implications for HBM3E and HBM4 pull-through for SK hynix and Micron starting in 2027. And that brings us back to the beginning of my post again. The million-dollar question for me is: how much of the CapEx raises are directly tied to supply chain stress and pricing — not only HBM and memory, of course, but much more. Update: I just read that Meta’s CapEx outlook increased to the $125 billion to $145 billion range, reflecting higher component pricing and additional data center costs to support future-year capacity. So my million-dollar question is already partially answered, but we still need more exact breakdowns, because it impacts short-term ROI. Lets see way the calls bring. $MSFT $GOOGL $GOOG $AMZN $META $NVDA $MU

@KookCapitalLLC @HYPEconomist A very honest question: would you sell Sol/Sui/ a few memes at big loss (bought near atm, lesson learned on liquidity exit) and port the proceeds to hype or wait for the next bull run to recover? —> = who has more upside % potential?

@HYPEconomist for some reason everyone in crypto wants to chase a beta instead of buying the obvious winner btc is the best example hype is the second best example

i see a lot of people saying they don’t wanna buy hyperliquid:native because it’s already so big and the upside is capped these are the same people who said the exact same thing about $BTC at $4,000 we’ve never seen hyperliquid go through a euphoria-driven bull run

@absidy1234 @ContrarianCurse Never regret taking profits!

@ContrarianCurse Bought at $20, sold half at $85 🫠

*BLOOM ENERGY 1Q REV. $751.1M, EST. $535.3M *BLOOM ENERGY SEES FY ADJ GROSS MARGIN ABOUT 34%, SAW ABOUT 32% Explosive growth, low capital needed for capacity, AND margins? My my my this is going to be a massive fucking company

@rubicon59 Seriously?

$BE out with great results. But the valuation is silly, and there is still no real earnings 🤷‍♂️

Yes but find it a small retate compared to what we are seeing on speculative no-revenue names in semis or photonics. When a company that is proving to deliver and is becoming the defacto power peovider for DCs brings a deal that is bigger than everything they have sold so far I would expect much more….time will tell but for me it is still massively under-priced

@embgaspbaud They already re-rated ahead of earnings off the back of the news a couple weeks ago

I usually front-run Earnings for specific companies. By placing trades during Earnings szn. (Don't advise this for most of Retail). $BE was one example from earlier today: - $ORCL agreement dropped 2 wks ago - up to 2.8 gw total. A deal bigger than most of $BE's prior annual shipments combined - Pre-Earnings, consensus was still pricing in only ~$500-540M rev + $0.09-0.13 EPS (mid-50s % growth) - But...the product backlog had hit a record ~$6B entering 2026, total pipeline ~$20B, and the $ORCL ramp was going to pull massive Q1 product revenue forward Plus: - Q1 is usually their weakest Q. So any strength in this Earnings would force a re-rate. So it was a pretty easy trade to make imo: - huge $ORCL contract + visible backlog conversion + hyperscaler tailwinds; all hitting at the right time pre-Earnings - also, defo could've entered off the back of the $ORCL news initially (would've had a much better cost basis), but didn't really have $BE as a high priority buy back then Note - this is my first ever $BE entry, and I plan to hold + potentially scale up slowly over time.

@ContrarianCurse It’s so big that one wonders if BE can deliver. 1GW= 1 nuclear reactor. It is estimated that it will require 7300 BE boxes more than their total ib. stock should be through the roof

Oh my

@fundmyfund Wasn’t this necessary after such a run? And won’t the market be like this onwards: run up and big pull back on some negative rumors? ….it’s still a hated rally

Going to be a rough morning for everyone's favorite sectors photonics / semiconductors / data center $GLW $AEHR $ALMU $VSH

@Mr_Derivatives You have it today 😅

If I was a bull in this bull market I want to see a healthy 3 or 4 step forward and 1 step backward. A kinda zig zag way up and to the right building solid bases and foundation at each incremental levels. Right now we are seeing a 10 step forward no step backward in the semis space. Well at least as of yesterday.

@Toronto242M @damnang2 Could be, you never know. MRVL has certainly the upper hand

@embgaspbaud @damnang2 $POET likely spent the weekend trying to salvage the $MRVL deal. When that didn’t happen, they disclosed the loss Monday—nothing unusual there. If the $poet tech is real, $MRVL could still return—possibly on better terms and use the NDA breach as leverage.

3:23 PM, analyzing with optics industry insiders whether it makes sense to re enter $POET.

@jaysyoon Yup. Lots of learning with Poet…. AAOI will be choppy but it’s probably way much safer than Poet. Real business, structural deficit for inp lasers, large orders…but let’s stay vigilant on execution

The $POET debacle is exactly the reason why I have sized $AAOI much smaller than my $LITE position. Execution is what ultimately matters, and AAOI has a history of piss poor execution. They seem to have somewhat gotten their sh*t together, so I think the risk/reward is worth it. But would not be surprised if this ends up being a disaster as well.

@Craaazy1231 @POETtech Cut your losses or be ready for a very very long recovery. It’s not about the tech anymore, it’s management malpractice bringing trust to 0. Having the (suposedly) best tech is not enough

I’m sorry for every $POET investor. I’m heavily in Poet myself with more than 4.000 shares, and nobody saw this coming. Even I was so bullish, I believed this could hit at least $30 EOY. I still haven’t sold yet, until a statement from @POETtech themselves. I hope everyone makes back their money.

If Sk and the 2 others start going into heavy capex cycle then indeed memory will stay cyclical and rated like that. Seems like they don’t learn. 3 companies that could cleverly decide to keep price power and go into dividends and buybacks with minimal reinvestments to keep their moat and control….I hope they don’t or else this trade will end abruptly i. The next 12 monthes

The reason why memory trades at low multiples

@FeroceResearch It’s sad but probably true; 9bn of zombies doomscrolling until they die for good….no ceiling

$META is the most misunderstood mag 7 company, while also having the highest ceiling potential as it is reshaping itself with Agentic and AI as a whole The perfect combo when it comes to upside eruptions in share price

@kevinxu It’s pathetic to try to get attention with such idiotic posts. Blocked, you are garbage

too rich to feel broke too poor to feel free net worth: $10,698,729.17

@FeroceResearch Love it but struggling to break out…. Patience

Rare Earths should be on your radar immediately #URANIUM based names seeing traction recently, and $UUUU happens to be my favorite name of the industry