Zan Digital › Companies
Meta
Every piece of Zan Digital analysis that covers Meta — 65 pieces, each figure traced to a named source.
- 001AI Capex 2026: The $725B Bet With No Payback MathFour hyperscalers plan roughly $725B of capex in 2026, up 77% on 2025. Here is where the money goes, what it earns today, and the three numbers to watch.$725B
- 002AI Circular Deals: 10 Arrows, Only 3 With a Cash FigureWe mapped 10 money flows between Nvidia, OpenAI, Microsoft and AMD. Only 3 carry a cash figure, led by Microsoft's $10.0B of OpenAI revenue in one quarter.$10.0B
- 003AI Market Share: Four Trackers, Three Different LeadersMenlo puts Anthropic at 40% of enterprise LLM spend, Ramp counts 43.8% of firms paying it, OpenRouter's tokens favour DeepSeek. What each tracker counts.40%
- 004AI Revenue Gap: Sequoia's $600B Question Rerun at $1.4TRerun Sequoia's $600B AI revenue gap math on Nvidia's $89.0B data centre quarter and the bill is $1.42T. OpenAI and Anthropic run rates cover about 8% of it.$89.0B
- 005AMD OpenAI Warrant: 160M Shares, $600 and 6 GigawattsAMD gave OpenAI a warrant for 160M shares at $0.01, vesting on 6 GW of GPU buys and a $600 price target. With Meta's twin, 320M shares, or 16.4% diluted.160M
- 006AWS Trainium Revenue: $225B Committed, None DisclosedAmazon has disclosed $225B of Trainium commitments and a $25B chips run rate, but no Trainium-only revenue. Every statement listed, and Broadcom compared.$225B+
- 007Broadcom AI Revenue: $16.7B, Up 221%, vs Nvidia's $89BBroadcom AI revenue hit $16.7B, 18.8% of Nvidia's data center line, and its top 5 end customers reached 55% of sales. Here is what the filings disclose.$16.7B
- 008Circular Financing: How Auditors Test AI Vendor RevenueAMD will book its 160 million share OpenAI warrant as a cut to revenue. How ASC 606 and related-party rules treat an AI vendor's equity stake in its customer.160M
- 009CoreWeave Customer Concentration: 62% From Two BuyersCoreWeave's top two customers paid 62% of Q2 2026 revenue. We tabulate Microsoft, OpenAI and Meta across filings and set them against the $35.6B debt stack.62%
- 010CoreWeave Customer Concentration: Top Client 71% to 36%CoreWeave's top customer fell from 71% to 36% of quarterly revenue in a year. Filing by filing, with Microsoft, OpenAI and Meta set against $35.6B of debt.36%
- 011Data Center Financing: Where the $5.7B Junk Tier SitsMeridian Arc's $5.7B junk bond at 6.25% was the largest AI deal in high yield. Here is where sub-investment-grade money sits in data center financing.$5.7B
- 012DRAM Shortage 2026: Who Pays the AI Memory PremiumDRAM contract prices rose 93 to 98% in one quarter and Micron's gross margin hit 87%. We trace who absorbs the AI memory premium: Nvidia, clouds or PC buyers.87.0%
- 013Firmus IPO: Why Public Investors Refused a 3x MarkupFirmus pulled its ASX float at a US$30.6B valuation, nearly 3x its August round. What public buyers refused: US$30B of debt, leased sites and few customers.US$30.6B
- 014GPU Depreciation: The $176B Accounting Question in AIHyperscalers depreciate AI hardware over 5 to 6 years. One investor puts the resulting overstatement at $176B through 2028. Both sides of the argument, checked.$176B
- 015H100 Rental Price Index: $3.00 and Three Sources ComparedSilicon Data put the H100 rental price at $3.00 on 8 November 2026, up 50% in 11 months. Ornn read $3.37 that day. Both tested against 6-year useful lives.$3.00
- 016Hyperscaler Bond Spreads 2026: Deal-by-Deal TableHyperscaler bond spreads, deal by deal: Meta's 40-year rose from +110bp to +147bp in six months and Oracle paid +145bp at 10 years. Seven deals compared.+37bp
- 017Hyperscaler Capex vs Operating Cash Flow: 96% in Q2 2026Hyperscaler capex took 96% of operating cash flow at Microsoft, Alphabet, Amazon and Meta in Q2 2026, up from 61%. The ratio, quarter by quarter, from filings.96%
- 018IMF AI Deals: $950B Raised Against $1.4T PromisedThe IMF counted 14 AI deals: $300B of equity and $650B of debt for $1.4T of promised revenue. We rebuilt the list and found $440B of it in company filings.$1.47
- 019Is the AI Bubble Real? 7 Tests, 2 Already TrippedIs the AI bubble real? We scored 7 falsifiable tests on Q3 2026 filings. 2 tripped: Alphabet capex hit 121% of operating cash flow. Demand tests still pass.121%
- 020Meta Data Center Financing: Who Holds the $27B RiskMeta's Hyperion campus runs on a $27.3B 144A bond issued by a Blue Owl vehicle. Who owns the asset, who carries the lease, and what bondholders can claim.$27.3B
- 021Microsoft Leases Not Yet Commenced: The $329B Beyond CapexMicrosoft's leases not yet commenced hit $329.1B in June 2026, 3.5x a year earlier. Add $140.6B of FY2026 capex and the AI data center bill nears $470B.$329.1B
- 022Oracle Credit Rating: Moody's, S&P and the 4.5x LineS&P cut Oracle to BBB- in July 2026 and named 4.5x leverage as the junk line. Moody's holds Baa2 negative. What each agency said, and how close Oracle sits.BBB-
- 023S&P 500 AI Mentions: 331 Calls Against the Profit LineFactSet counted 331 S&P 500 earnings calls citing AI in Q2 2026, 67% of the total. The count has stalled while profit swung, so the talk predicts little.331 of 493
- 024SaaS Multiples in an AI Capex Slowdown: 3 ScenariosAI capex adds roughly 140 basis points to US growth in 2026. A slowdown would lift SaaS multiples on one side and crush them on the other. Three scenarios.140bp
- 025Server Useful Life: $33B to $41B of Big Tech ProfitMicrosoft, Alphabet, Amazon and Meta depreciate servers over 5 to 6 years. At the 4-year life all four once used, operating income falls $33B to $41B.$2.92B
- 026Server Useful Life: $38B of Hyperscaler Profit at StakeMicrosoft, Alphabet, Amazon and Meta depreciate servers over 5 to 6 years. At a 4-year life, their combined operating income would fall about $38B, or 8.5%.$2.92B
- 027The $662B Data Centre Leases Off Big Tech Balance SheetsMoody's found $969B in hyperscaler lease commitments, $662B of it not yet on any balance sheet. What the rule is, why it matters, and when it arrives.$969B
- 028AI Budgeting: Uber's $1,500 Cap and the Exception RuleUber spent its 2026 AI coding budget in about 4 months, then capped tools at $1,500 per engineer a month. A per-seat AI budgeting method with exception rules.$1,500
- 029AI Pricing Changes: 6 Vendors That Deleted Plans (Ledger)Sembly, DocsBot, Z.ai, Zendesk, Microsoft and Rewind removed AI plans instead of repricing them. A dated ledger of 6 deletions, with 9 to 365 days of notice.14 days
- 030AI Data Retention Policies, Ranked by What They PromiseEvery major AI vendor's data retention policy, scored on 5 criteria. Anthropic keeps covered-model prompts for 30 days even where zero retention was agreed.30 days
- 031LLM Leaderboard Rankings: Why Arena, AA and SEAL DisagreeArena ranks Gemini 4 Argon first, Artificial Analysis ranks Claude Opus 5.5 first, Scale's HLE board ranks GPT-6 Astra first. 3 methods, 3 different winners.3
- 032Open Weights in Production: Qwen, DeepSeek, Llama, GLMChinese models take 61% of OpenRouter tokens and about 1% of enterprise LLM spend. Comparing Qwen, DeepSeek, Llama and GLM on quality, cost and licence.61%
- 033OpenAI Dots Review: Plans, Admin Controls, EU GapOpenAI dots run on GPT-6 Astra, but Pro users in the EEA, UK and Switzerland are shut out. A $125 Business Premium seat is the route in. Our buyer review.1 dot
- 034Agent Design Patterns: What 3 Vendor Guides Agree OnAnthropic, OpenAI and Google name 21 agent design patterns between them. Mapped side by side, 5 survive in all three guides, and each has a shipped case study.5 of 9
- 035Agentic AI Projects Canceled: The Public Ledger So FarGartner said 40% of agentic AI projects would be canceled by 2027. 16 months on, we found 3 named public cancellations, and not one of them blames cost.40%+
- 036AI Generated Unit Tests: Coverage vs Mutation ScoreOne AI test suite reached 100% line coverage and caught 4% of injected bugs. What mutation testing studies show, and how QA teams should score AI tests.4%
- 037AI Vendor Lock In: What Switching Models Really CostsAnthropic retired 9 models in 10 months, one on 61 days notice. The API call is the cheap part. Here is what model switching actually costs to redo.40%
- 038Meta Project OT: Code Up 220%, Features Up 36%Meta Project OT figures: code changes up 220%, user-facing changes up 36%, major incidents up 40%. What the numbers support and where causation stops.220%
- 039AI Copyright Lawsuits: The $1.5B Repricing of Training DataAnthropic paid $1.5 billion, about $3,000 a book, to settle. Here is what the live AI copyright cases could do to model economics and licensing prices.$1.5B
- 040AI Election Safeguards: What 4 Labs Published by Nov 3Anthropic found 0.51% of 250,000 Claude chats were about elections. We compare its rules with OpenAI, Google and Meta, cell by cell, each sourced to the lab.0.51%
- 041AI Licensing Deals Tallied: 14 Deals, 5 With Real NumbersReddit filed $203M of AI licensing contracts. We tallied 14 AI content licensing deals, graded each by evidence, and compared them with litigation payouts.$203.0M
- 042Data Loss Prevention for AI Prompts: What Purview BlocksNetskope counts 223 genAI data policy violations a month at the average firm. Here is what Purview, Netskope and Zscaler DLP each catch in AI prompts.223
- 043EU AI Act GPAI Enforcement: What Downstream SaaS OwesSince 2 August 2026 the AI Office can fine model providers 3% of turnover or €15M. What falls on OpenAI and Anthropic, what falls on you, article by article.3% or €15M
- 044EU AI Act Transparency Rules: The 2 August ChecklistArticle 50 transparency duties applied from 2 August 2026, with fines to €15M or 3% of turnover. Only 9 of 27 member states have both enforcers named.€15M or 3%
- 045FTC Personalized Pricing Draft: What Pricing Vendors OweThe FTC's 2-0 personalized pricing draft expects 3 disclosures. A clause-level read, the industry objections, and what AI pricing vendors will have to show.2-0
- 046Getty v Stability AI Ruling: Model Weights Are Not CopiesAt paragraph 600 the UK High Court held Stable Diffusion's weights never stored Getty's images. What was decided, what was left open, and the US contrast.5.85B
- 047Open Source AI Licence: The 4 Clauses Nobody ReadsMost open weight models are not open source. 4 clause types decide commercial use, from Llama's 700 million user cap to terms you must pass downstream.700M
- 048Prompt Injection: 12 Defences Broken Above 90%Researchers broke 12 published prompt injection defences with attack success above 90%. Why the SQL injection fix has no equivalent, and what actually works.Above 90%
- 049Texas AI Law TRAIGA: Nine Months, Zero Public CasesNine months in, Texas has announced zero public TRAIGA cases. Its AI cases ran on older laws, and the statute still sets a 60-day cure and $200,000 penalties.0
- 050Thomson Reuters v. Ross: 2,243 Headnotes, No Fair UseOn 29 September 2026 the Third Circuit held that copying 2,243 Westlaw headnotes to train ROSS's AI was not fair use. The record and what is next for vendors.2,243
- 051Agentic Commerce Protocol: 15 B2B Checks, 5 MissingWe checked 15 B2B purchase needs against the Agentic Commerce Protocol. 6 are supported, 4 partial, 5 missing, including recurring billing and non-card pay.4 to 11
- 052AI Content Labels: LinkedIn, YouTube and Meta RulesAI content labels compared: what LinkedIn, YouTube and Meta make B2B marketers disclose in ads and organic posts, checked against 3 help centres, October 2026.11 Jan 2024
- 053LinkedIn Revenue Growth: 12%, Then Microsoft StoppedLinkedIn revenue grew 12% in Microsoft's last two reports before FQ1 2027 folded it into search ads. The 8-quarter table and what it says about B2B ad spend.12%
- 054Marketing Mix Modeling: Google Meridian vs Meta RobynGoogle Meridian and Meta Robyn compared on method, data needs and documented limits. Robyn wants 10 observations per variable, a bar most B2B teams miss.1 : 10
- 055AI Engineer Salary 2026: Compression Is the Wrong WordPwC measures a 62% wage premium on AI job ads while payroll data shows base pay barely moved. Here is what AI engineer salary spread really looks like.62%
- 056AI Ethics Team Cuts: 15 Dated Events at 6 CompaniesWe logged 15 dated events on AI ethics teams at 6 companies since 2020. 13 were cuts or reshuffles. The only growth on record is Microsoft's 350 to 400.13 of 15
- 057AI Labelling Pay and Lawsuits: Scale, Surge, Mercor LedgerScale AI agreed a $12.5M settlement with California contractors. A ledger of every labelling lawsuit, pay rate and complaint at Scale, Surge and Mercor.$12.5M
- 058AI Layoffs vs Overhiring: What the Hiring Data SaysAI was named in 112,713 US job cuts through July 2026, up from 54,836 in all of 2025, while total announced layoffs fell 41%. The hiring histories disagree.112,713
- 059Amazon Manager Ratio: The 15% Mandate vs the FilingsAmazon told every org to lift its IC-to-manager ratio 15%. Its headcount still rose 1.3% in 2025. Six delayering programmes checked against their filings.15%
- 060Big Tech Headcount 2026: Capex Up 243%, Staff Up 4.9%Alphabet, Meta, Microsoft and Amazon spent $201.9B on capex in Q3 2026, up 243% in two years, while combined headcount rose 4.9%. Capex per employee, computed.$201.9B
- 061Engineering Career Ladder: Redefining the Junior RungWe read the public engineering career ladders of GitLab, Dropbox and CircleCI. 0 of 3 mention AI at the entry level. Here is a first rung written for agents.0 of 3
- 062Generative UI vs Designers: The Craft Moves to ConstraintsGoogle's generative UI scored 1736 Elo against 1800 for human-designed sites. What runtime interfaces change about design systems, craft and hiring.1736.2
- 063Remote Work Monitoring: The Trust Cost Nobody PricesA meta-analysis of 94 studies covering 23,461 workers found no evidence that monitoring lifts performance. What it reliably changes is stress. Four fixes.23,461
- 064Revenue Per Employee: The $2.7M Team Tripled HeadcountLovable runs about $2.7M of revenue per employee against a $141,125 private SaaS median. What AI-native team structure changes, and what it does not.$2.7M
- 065Technical Interviews When Everyone Ships With AI71% of engineering leaders say AI has made technical skills harder to assess. Portfolios and take-homes stopped signalling. Here is what replaces them.71%