Meta Platforms

The print confirmed the mechanism we underwrote and the multiple survived it: revenue ACCELERATED to +28% YoY, price per ad held +12%, and forward P/E is 17.33 at the $556.71 Jul 31 close. The cash engine accelerated too — operating cash flow $31,862M, +25% YoY — with the $784M quarterly FCF an outcome of capex timing that our own bear case pre-specified. On top of the read holding, two legs we never modelled appeared at once: a named enterprise/compute revenue pillar, and a JV structure that funds gigawatt-scale capacity without consolidating it.

• The multiple is intact and verified: forward P/E 17.33 at the $556.71 Jul 31 close, 17.6x on consensus FY2026 EPS of $31.72 (both ESTIMATE) — against revenue that ACCELERATED to $60.80B in Q2 FY26, +28% YoY reported / +27% cc (advertising +27% / +26% cc; H1 FY26 +30%), with Q3 guided $61-64B. One caveat: the 'cheapest mega-cap' peer-relative framing is NOT verified — META's own 17.33x is a figure, the peer ranking is an assumption
• The composite ad output accelerated while the mix normalized: price per ad held at +12% YoY (flat vs Q1) and total advertising revenue accelerated to +27%, even as impressions decelerated from +19% to +14%. A genuinely broken volume leg would decelerate the composite; it did the reverse. Correction in place: we published this signal as unqualified confirmation that AI capex converts to ad yield — it restates as 'price per ad holding at +12% while impression growth normalizes toward lower-monetizing surfaces and regions' (management's own attribution), with total ad revenue growth as the composite check
• The cash engine accelerated even as reported FCF compressed: operating cash flow $31,862M, +25% YoY. The $784M Q2 FCF is a capex-timing figure — capex of $31,078M incl. finance-lease principal was +57% sequentially off Q1's $19.8B and consumed nearly all of it. On the half-year, the shortest window that survives that lumpiness, FCF is $13,170M vs $18,883M
• The FCF compression landed inside the band our own model pre-specified, which is confirmation of the read rather than new information: we projected ~50-60% capex intensity, and observed intensity is 51% of Q2 revenue, 43% for H1, and ~51-57% for the full year against the $130-145B guide and ~$253B consensus revenue (ESTIMATE). The verifier's materiality on this field is LOW — 'our projection was right; it can now be restated as fact'
• A second revenue stream is named rather than hypothetical: enterprise APIs, business agents (>1M businesses using them weekly on WhatsApp/Messenger, self-reported; rolling out on Instagram; Meta Business Agent Platform launched July), and 'potentially selling compute directly' — Zuckerberg says Meta is 'getting a lot of offers for compute at a significant premium over what we paid for it' (self-reported, unquantified). Monetization is described as subscriptions, volume pricing, and eventually a results-based auction 'similar to how we do that for our advertisers today'
• Non-ad revenue shows in the P&L at scale for the first time: Family of Apps 'other revenue' $1,007M in Q2 FY26, +73% YoY — the first $1B+ quarter — on WhatsApp paid messaging and subscriptions, with a payments founder (Kunal Shah) newly running WhatsApp
• The buildout is partly financeable off balance sheet: the $14B BlackRock El Paso JV (BlackRock-managed funds 80% / Meta 20%, with Meta keeping full operational control and sole initial occupancy; Meta contributes ~$2.3B of land and construction assets and receives a ~$1B one-time distribution; BlackRock contributes ~$4.9B cash; $12.5B of debt financing behind the venture; online 2028). Management framed it as 'partnerships that we can structure to complement our approach to building infrastructure capacity,' signalling more to come — announced Jul 28, one day before the $784M FCF print
• The GAAP EPS shortfall is charge-driven, not demand-driven: ex $2.40B of legal-proceeding charges and $1.18B of May-2026 severance, operating income rose ~9% YoY (company-stated) on a ~36.8% adjusted margin (derived), and FY2026 operating income is still guided above 2025
• The owner-operator expense lever is demonstrated, not asserted: the ~8,000-person May 2026 reduction took headcount to 75,472 (-1% YoY, -3% QoQ) at a $1.18B Q2 severance cost — $1.18B against $31,862M of quarterly operating cash flow. Its direction is cost discipline, which supports the 'owner-operator can pull the lever' point rather than undermining it
• Meta AI engagement lifted after the Muse Spark rebuild — daily interacting users +60% and still growing week-over-week (self-reported, relative only) — alongside a paid public model API at a competitive price, Muse Spark on OpenRouter for US developers, and a 'Meta One' subscription portfolio. Meta disclosed no absolute user figure this quarter, so only the relative lift and the named products stand

The cost of the buildout is arriving roughly on schedule — but the FUNDING MIX has become the live question. Zero buybacks in H1, net cash down to ~$6.6B after $24,910M of net debt issued in a single quarter, capex intensity now observed rather than projected, and a regulatory line that turned into a $2.40B quantified P&L item with management warning of possible further material loss — while 2027 capex visibility was explicitly withdrawn.

• Capital return is gone and the balance-sheet cushion is thinning: share repurchases were $0 in Q1 and Q2 FY26 versus $10,167M in Q2 FY25 and $22,921M in H1 FY25, leaving only the ~$1,353M quarterly dividend; net cash fell to ~$6.6B ($90,260M cash+securities vs $83,664M long-term debt at Jun 30, 2026) from ~$22.8B at Dec 31, 2025, after $24,910M of net long-term debt issued in Q2 alone, while property and equipment net grew from $176.4B to $225.7B in six months. On this trajectory the balance sheet flips to net debt within roughly a quarter
• Capex intensity is now observed, not projected: 51% of revenue in Q2 FY26 ($31,078M incl. finance-lease principal), 43% for H1 ($50,918M), and ~51-57% for FY2026 against the $130-145B guide and ~$253B consensus revenue (ESTIMATE). Half-year FCF is $13,170M vs $18,883M; the single-quarter figure is $784M (1.3% of revenue). One caveat on the comparison base: the FY2025 figures the deterioration is measured against ($46.1B FCF, $69.7B capex, 41.4% operating margin) are NOT verified against the 10-K, so every 'versus FY2025' comparison rests on an unconfirmed denominator
• GAAP operating margin fell to 31% from 43% a year earlier, with income from operations of $18,775M DOWN 8% YoY on 28% revenue growth; adjusted for the $2.40B legal and $1.18B severance charges it is ~36.8% (derived). The Reality Labs drag on consolidated margin is ~7.6pp on current figures ($4,619M loss against a $60,801M revenue base) — a '~19pp drag' computed off the FY2025 base overstates it
• Regulatory cost became a quantified P&L item: $2.40B of legal-proceeding charges in Q2 G&A (roughly 4pp of the operating-margin compression), with an explicit forward warning that 'a number of youth-related trials scheduled for this year in the US … may ultimately result in a material loss.' Two limits on how far this can be read: Meta did NOT name the underlying matter, so the youth-litigation attribution circulating in secondary reporting is unverified and recurrence is currently unsizable. Separately, the EU DMA matter (€200M scale) is a far smaller item on materiality, and carries no current update — which is not the same as resolved
• Forward visibility was withdrawn while the spend floor rose: FY2026 capex incl. finance-lease principal was NARROWED UP to $130-145B, FY2026 total expenses are now quantified at $165-169B, and the remaining-quarter tax rate was raised to 15-17% from 13-16% — with 'we aren't providing a specific outlook for 2027 CapEx at this time.' Read strictly against our own pre-registered trigger ('bearish if raised again'), the top end was unchanged and the bear trigger did NOT fire; what did happen is a $5B rise in the guaranteed floor and the loss of the forward visibility this signal depends on
• True economic commitment exceeds on-balance-sheet capex: $83,664M of long-term debt, plus a residual-value guarantee around a ~$13B threshold (decreasing over time) and a lease running 4 years initial with four renewal options up to a potential 20-year span on the El Paso JV. The same structure that eases the funding question makes the spend contractual rather than discretionary — which cuts directly against the 'the spend is discretionary, not contractual debt' bull point
• The ad mix is normalizing and the pricing tailwind is already reversing in the guide: impressions +14% YoY (from +19% in Q1), Family DAP 3.60B (Jun 2026 avg) +3% YoY and decelerating 1pp, and management named better macro and FX tailwinds as part of the +12% price per ad — while the Q3 guide assumes FX is a ~1% headwind to YoY growth
• Reality Labs remains a structural drain with no disclosed profit path: -$4,619M operating loss on $431M of revenue in Q2 FY26; H1 FY26 loss $8,647M versus $8,739M in H1 FY25 — roughly on par, i.e. not improving. Two wider figures are NOT verified — a '>$83B cumulative loss' and a '~$19B/yr cash drain' — and the segment figure is an operating loss, not a cash-flow line, so the reported H1 comparison is the basis here