# Meta Makes $16 Billion a Year From SCAMS... ON PURPOSE > Internal documents show Meta earns $16 billion a year from scam ads and shields that revenue by explicit policy. Not incompetence, but design. - Canonical: https://siliconvalleyconfidential.com/en/dossier/meta-makes-16-billion-a-year-from-scams-on-purpose/ - Site: Silicon Valley Confidential (https://siliconvalleyconfidential.com) — weekly executive intelligence on Silicon Valley and global tech - Author: Jose Luis Cases (https://es.linkedin.com/in/jose-luis-cases-lozano) - Language: en - Published: 2025-11-08 (original LinkedIn edition: https://www.linkedin.com/pulse/meta-gana-16-mil-millones-anuales-con-scam-adrede-jose-luis-cases-klllf/) --- THE BOMBSHELL OF THE WEEK --------------------- On November 6, Reuters revealed internal Meta documents exposing a very dirty business model: approximately 10% of the company's annual revenue ($16 billion in 2024) comes from fraudulent ads, illegal casinos and banned products. The numbers don't lie: * 15 BILLION "high-risk scam ads" shown DAILY * $7B annualized from ads classified as "high-risk" alone * Meta only bans advertisers when it's 95%+ certain (a deliberately high threshold to protect revenue) * Internal "revenue-protective" policy: the anti-fraud team CANNOT take actions that cost more than $135M in lost revenue * UK Financial Conduct Authority: Meta involved in 54% of ALL payment fraud losses in 2023 This isn't incompetence. The leaked internal documents reveal an EXPLICIT POLICY to protect the revenue that comes from scams. The Trust & Safety team raised the alarm multiple times, but the product side decided not to implement aggressive filters because it "would hurt revenue in emerging markets." POWER MOVES ------------------------ ### Robin AI: from top 10 British startup to forced sale in 9 months I didn't see this one coming — legal AI looked like a safe bet. Anatomy of the fall: * April 2025: Ranked 10th in the Sunday Times list of the UK's 100 best tech startups * October 2025: Put up for urgent sale after failing to raise $50 million in funding * Deadline for offers: November 13, 2025 Robin AI was one of the "safe" bets in AI — it built tools for lawyers with clear use cases, well-paying enterprise clients, no consumer products with dubious margins. If a company with real clients and proven sales can't raise funding, it means AI investors are starting to close their wallets. The AI bubble could be showing serious cracks. MONEY TALKS --------------- ### The US blocks Nvidia chips in China. Again. Nvidia had designed the H20 chip to "comply" with the previous restrictions. The government changes the rules yet again and blocks it too. Nvidia had already absorbed $5.5B in losses. Analysts are now talking about $14–18B less per year in China alone. And China isn't sitting still: by 2027, all public data centers will use 100% domestic chips. With this, Nvidia loses. AMD and Intel lose. The United States loses the market. The only winners: Huawei, SMIC and the new Chinese startups, which now have a protected $30B market to grow in without competition. ### The unicorn bubble is deflating. New data from EquityZen shows that almost 1 in 3 unicorns is already worth less than $1B in secondary markets. In other words: the private valuations of 2021–2022 were fantasy. In the real market, most are worth less. Much less. Funding rounds are no longer "up only": * Down rounds are multiplying * And almost half of late-stage rounds are flat or negative Who's exposed? The funds that marked their portfolios at the peak. They have enormous losses... but unrealized. For now. PRODUCT SECRETS -------------------- ### Apple cancels the Vision Pro 2 and halts production According to The Information, Apple halted Vision Pro production in the last quarter and has officially canceled development of the Vision Pro 2. They're now planning a cheaper model for 2026 ($1,500−2,000 vs the current $3,499). The failure numbers Apple doesn't want you to know: * Fewer than 500,000 units sold since February 2024 (original target: more than 1 million in the first year) * Return rate: ~20% (abnormally high for Apple) * Active use after 6 months: less than 30% of buyers Apple is admitting that premium virtual reality is dead. Either they make a cheap model or they kill the product. Meta Quest has beaten them. ### Tesla FSD v14 is still hallucinating Beta users in November 2025 (Reddit and YouTube) report that version 14 of Tesla's autopilot STILL has dangerous failures: * It sees vehicles and obstacles that don't exist (visual AI hallucinations) * It brakes abruptly for no reason on highways * Aggressive lane changes that scare passengers * "Phantom braking" persists after 8+ updates MY TAKE --------------- The Meta story is complicated and ugly. Very ugly. If Meta loses the $16B lawsuit, it will set a global precedent for platforms' direct liability for fraudulent ad content, and that will translate into a regulatory risk that I expect the markets will have to price in, in $$$. It's a perfect example of KPI misuse. A KPI can never be crowned king, because then it rules you. And on top of that, this KPI has ruined the lives of countless people who were scammed with Meta's blessing. Moving on to legal AI — the Robin AI story surprises me. I still genuinely believe AI applied to the legal field has enormous potential. We'll grab some popcorn and wait to see what happens. On unicorn valuations: this is just the beginning. Many of the pre-AI-era unicorns can (and should) be replicated with a model that carries only a fraction of the costs. That will enable lower prices and trigger a competitive wipeout that eliminates many of the incumbent players. Secondary markets don't lie... The NVIDIA story is a textbook mistake. It's not the first time the US has paved the way for China, and it's doing it again. Unless there's some precisely timed play with a very narrow window of opportunity, I don't understand the move. The Tesla story is entering an interesting phase.... They haven't publicly acknowledged the problem. The testers appear to have signed NDAs, but they're leaking information through anonymous YouTube accounts. This could bring a lot of volatility to the stock in the coming months. And that's it. I'm approaching the 12 issues I set myself as the deadline to apply a survival test to this "product." If it adds value for you and you don't want it discontinued, I need ONE SIGNAL. Commenting is best, sharing on WhatsApp, reposting. And if not... there's always the "like," which is the bare minimum. But if you truly value it, I'm asking you not to be indifferent. Thanks for reading.