Tõnis

$75B+ in annualized ad revenue now runs through Meta's Advantage+. But what exactly is Advantage+ and why it matters? What Advantage+ is? Advantage+ is part of Meta's suite of AI powered advertising tools that automate many of the decisions which businesses traditionally had to make themselves. How does it work? Imagine that you are clothing brand trying to sell new shoes. You previously had to make bunch of manual decisions: > Who to target? (age, interests, demographics) > Where to advertise? (Instagram feed, Reels, Facebook) > How to distribute your advertising budget? With Advatage+, much of this can now be done automatically. Human advertisers make decisions based on who they think their customers are. But AI can identify patterns in user behavior that humans might never think to look for. As Meta's AI models improve, so does their ability to find the right customers for advertisers. Meta's AI helps to optimize: 1. Audience: Finds people likely to buy, even if they fall outside the advertiser's suggested audience. 2. Placements: Determines where ads are most likely to perform well across their ecosystem of apps. 3. Budget: Dynamically distributes spending across ad sets based on expected performance. 4. Creative: Uses AI tools to generate and optimize different versions of ads. Advertisers still control their objectives, budgets, brand assets and certain restrictions. Does it actually work? During their Q2 earnings call, Meta highlighted an Indian apparel brand that switched from manually configed campaigns to Advantage+. The company reported 13% incremental lift in purchases and 16% increase in add to cart conversions. Why does it matter? Advertising is becoming increasingly automated. Small businesses no longer need same level of expertise to manage every aspect of their campaigns. For Meta, this could mean more businesses advertising on its platforms and existing ones spending more as their results improve. Meta isn't only improving the ads people see, but also making the process of buying advertising easier. $META

Meta's advertising business is simply best in class. $META $AMZN $GOOG

$META shared new details on Andromeda, its new ads retrieval system What is Andromeda? Andromeda is responsible for the first stage of Meta's ad recommendation process. It narrows down massive pool of eligible ads into smaller group of more relevant ads, which are then evalued by other models. What's new? They redesigned Andromeda to make it way more efficient. Compared to 2024, new improvements are: - 300x less computing work needed to evaluate ads - 100x more ads processed per second The new system can now evaluate over 1 billion ads per second on a single GPU server. Interesting improvement is also introduction of 'Mixture of experts' architecture. Instead of relying on a single model to understand user interests, Andromeda uses multiple specialized models that can better capture different interests and behaviors. Meta also worked with $NVDA to optimize how Andromeda runs on its GPUs. Why does it matter for Meta? As AI makes it easier for advertisers to generate new ads, the number of ads Meta needs to evaluate is also growing. Andromeda allows Meta to evaluate more candidates, imrpove ad matching and use its computing resources more efficiently. Meta reported that Andromeda has contibuted to 2.4% increase in Facebook conversions since 2025. Great to see that they aren't simply throwing more compute at their ad models, but always optimizing what they have.

$GOOGL Q3 earnings: - EPS miss due losses in equity (spaceX) - Slight miss of revenue $127.6B, 25% YoY - Google Cloud platform growth 103% YoY🤯with 36.2% operating margins. - TPU system sales were meaningful contributor to Cloud growth, nearly 100 customers already use TPUs. - 3rd party compute came online at the end of Q3 and they will increase reliance on external capacity. - GCP backlog of 561B and expected to realize 55% of the revenue in next 24 months - Customers who use gemini enterprise more than quadrupled their spend YoY - Search revenue grew 12% YoY and management said that query growth was highest in 5 years. - AI mode still less monetized than traditional search and contributed to the slowdown in revenue growth, alongside FX effects. I think there is still ways to go for search growth here. - Gemini has now 1B monthly users, Google's first-party model APIs process nearly 32B tokens per minute. - Introduced Gemini 4 carbon, strongest frontier model atm. - Talked about golden age of productivity, where they can scale gemini 4 carbon models internally, and to customers. We have to find out how good these models actually are outside benchmarks. - Google spent $51.4B on capex in Q3, compared to $42.3B in operating cash flow, which means negative cash flow of $9.1B - They mentioned in the call how capex will increase significantly in 2027 but remain confident that these investments will generate good returns. - Youtube advertising revenue gre 9% YoY to $11.2B, with living room viewing now accounting for 1/3rd of ad revenue. - Youtube premium also is now at 150m subscribers globally, including trials.

Google reports their Q3 finanicial results today. Here is a pre-earnings warmup for you: - Revenue is expected to come in at ~$128B, 25% YoY. EPS of around $3,02. - Q2 free cash flow was negative and this quarter will also be negative. How many negative cash flow quarters will we have? - Capex is estimated to be around $55,4B. Watch for capex raise during the call. Sundar himself mentioned that they are capacity strained. - All eyes will still be on GCP growth, margins and backlog. GCP revenue could come in at wild 100% growth, OP margins at 38% and RPO at $520B+ - Search and other growing at 15% but worth listening for commentary on agents and what does that means for search going forward. - Youtube watch time keeps going up and up. Youtube ads revenue could surprise people. - Subscriptions, prices have been going up lately but are they growing in general? - Could hear commentary on 2027 TPU business and how well that is doing. - Gemini 4 Argon processed nearly 25 trillion tokens in its first 72 hours, around 3x Gemini 3.0s launch. Watch for more Gemini 4 commentary - According to Similarweb, Gemini's share of website traffic has grown from 13,6% to 25% over the past year, although it's down slightly from 26.7% in April. - Implied move after the earnings is around 6% or ±19.07 per share - From technical perspective, Google has been under pressure lately as it has fallen roughly 10% from its October highs. Daily RSI is approaching oversold territory at 34. Watch for $316 area for support and $325-330 for resistance $GOOG $GOOGL

I think commerce could become nice opportunity for Meta. TikTok shop generated estimated $50B in GMV during H1 2026, growing 92% YoY Now imagine something like that inside of Instagram and Facebook reels. If all the Facebook mums and doomscrollers could buy stuff they discover in Reels without ever leaving the app. Meta already has infrastructure in place for that like, Meta pay for payments, reels for discovery. Only thing missing is to complete buys directly inside IG or Facebook. Keeping transactions inside its apps, Meta could gain better purchasing data, improve their ads and collect commission on sales. TikTok shop shows that the model can already work in scale. Thoughts on MetaShop? $META

Thought I'd share some thoughts after the new company Sundays, so here goes nothing: - Zeta is AI powered marketing company. Perfect because my portfolio is already filled with ad tech companies. - Founder-led by David Steinberg, who has around 30 years of experience building companies. - Management made decision years ago to rebuild its platform around AI and intentionally pushing profitability to near 0. - 20 quarters of beat and raise. - Platform is delivering 6x return on ad spend. Expecting to push this to 10x with Athena OS. - They have surprising data moat, as they have proprietary data platform, which is almost as big as Meta, Google and Amazon. - They are working semiconductor partner to develop AI inference chips, wtf? Overall it's quite technical company, but I'm quite interested in learning more. Specially with Steinberg in charge. $ZETA

$META is acquiring AI analytics startup Sundial What is Sundial? Sundial was founded by 2 former Meta employees, Julie Zhuo and Chandra Narayanan. Sundial builds AI powered analytics tools that help businesses make sense of their data. Sundial has worked with companies like OpenAI and Gamma. In its openAI case study, Sundial reported that its platform saved 100s of engineering hours monthly. Why does it matter for Meta? Meta mentioned some interesting details about what they are building. Their goal is to use AI to automate analytics work happening inside. Including identifying important trends in business data, running experiments and analyzing their results. They are also working on AI powered monitoring that helps teams focus on the signals that actually matter. Meta is improving internal workflow efficiency with AI and Sundial is something that helps with it.

$META Q3 2026 earnings: - Revenue of $64.6B (+26% YoY), above the $63.3B guide. Q4 guidance is $73 - 76B - GAAP EPS $3.83, Adjusted for one time legal charge it would be $7.40. - Meta joins the gang of negative FCF with Google and Amazon. - New class C shares take effect Jan 19 next year. No financing plans yet, but they are likely raising capital through it at some point. - Core business is still good. Impressions are up 10%, price per ad is up 14%. Revenue per user $17.72 (+23% YoY). Other revenue is $1.1B (+59% YoY), mostly WhatsApp paid messaging revenue. - Daily active people at 3.64B, up 3% YoY - Threads passed X in monthly actives. - Headcount at 66 775, down 15% YoY. - Bottom end of capex raised by $5B - Reality labs lost $4.13B on $443M of revenue. The 2026 loss is guided at $18 - 19B, then $15B in 2027 and trending towards profitability. - About Muse, 10M+ people in the US and Canada have setup personal agents with 25% daily use in ~6 weeks - Not global because compute strained. They're building compute to serve 1B+ Muse users over the next few years and expect 10x+ cost efficiency within next years. - Meta One has 15M+ subscriptions and trials.

$META reports Wednesday, some pre-earnings info: - Revenue is estimated to be roughly $63,3B, near the top of their own $61 - 64B estimate. - They will realize $10B legal charge this quarter, which will impact reported eps. - Is the core business getting even better? Watching ad impressions and average price per ad. - Daily active people. They reached 3.6B in Q2, roughly 60% of the world's internet users. How much room is left for user growth? - Early Muse commentary, can they already talk about plans to monetize Muse? - Muse wonder and other models. Could they start relying more on their own models instead of paying Anthropic and OpenAI? - Usage costs, Muse is wildly popular and currently free. All that compute is costly so might see expenses go up more. - Meta one subscriptions. How much are people adapting subscriptions and some numbers around that. - 2027 capex. Could potentially hear about that. Current guidance is $130 - 145B and free cash flow is $784M. I wouldn't be surprised if 2027 capex is $200B+ and cash flow negative. - EssilorLuxottica earnings on Tuesday also worth a follow as we might hear some Meta glasses commentary there. - At $709, I don't think they are priced for perfection. Will depend on the call tho - Implied move by options after earnings is about 7.3%.

How markets felt this week

Learning more about ads $META, interesting thing I found is that even micro improvements to ad efficiency can translate into hundreds of millions of dollars in additional revenue at this scale.

Got a demo of new $META glasses at a best buy. I didn’t hate them (biased ofc but ye) photos looked decent but audio was apparently leaking quite a bit. Also shoutout to Meta sales assistant. She gave me her lanyard aswell. (No IG unfort :/)

It seems like my ytd gains were so good that T212 decided to nerf me 😂