META Q1 Earnings Prep: Pulling Cash out of the Walls
Meta is -- obviously -- going big on AI. META’s AI capex differs from peers because it has the cleanest ROI math: it feeds right back into a 52% margin advertising business. Their AI capex must eventually turn into higher ad pricing through better targeting, more user time through better engagement, or lower cost per inference at scale. In contrast, Microsoft’s AI capex underwrites Azure capacity for OpenAI and Copilot attach that’s still proving itself, Google’s feeds Search which is being cannibalized by AI Overviews, and Amazon’s feeds AWS, which has been losing AI share to Azure and GCP.
The questions for Wednesday are whether Meta’s ad machine is still going strong enough to absorb a $145B infrastructure build, what’s left in the tank if it stalls, and most optimistically, are AI investments bearing fruit for the ad business already?
Start with the ad engine. It has been consistently accelerating, not slowing. In aggregate, daily users have continually marched upward from 2.82B (2020) → 3.58B (2025). ARPU more than doubled from $33.68 → $57.03. In FY25, ad volume was up 12% and price per ad up 9%. Revenue per employee is $2.55M (vs MSFT’s $1.24M). This is an extraordinarily profitable platform business, and it’s what makes the rest of the story even possible. Meanwhile, the drag on it has been Reality Labs, and that looks as if it is being systematically dismantled. It has lost $88B since 2019, but several rounds of layoffs and a shift in strategic mission are winding down its impact on the balance sheet. Listen for: ad price decel (4Q25 was +6%, BofA models +14% in 1Q26), any Family of Apps segment margin compression, or more aggressive cuts at Reality Labs.
What size of investment do ads need to fund? Capex doubled in 2025 — $37.3B → $69.7B, but the off-balance-sheet picture is much bigger. Multi-year cloud commitments hit $40B as Meta, the company that built its own data centers for 20 years, is now signing big multi-year deals with CoreWeave, Nebius, and others. Leases not yet commenced are at $103.8B. And the Hyperion joint venture with Blue Owl Capital carries a $46B max exposure that Meta keeps off the balance sheet. Total contractual commitments went $33B → $131B in a single year.
That’s a 4x jump, too big for the ad engine alone, and so Meta is already drawing fuel from the tank in three ways:
Accounting. This follows the same playbook as peers like MSFT. Server useful life: 4 yrs (2021) → 4.5 → 5 → 5.5 (Jan 2025). The 2025 reassessment alone added $2.92B in depreciation savings, $2.59B in net income, and $1.00 to diluted EPS. But this is a temporary cushion: D&A is still set to ramp from $18.6B (FY25) to $46.7B (FY27E), so the lever gets overwhelmed by sheer asset base growth fast. Listen for any further useful-life extension as a sign they need more cushion, or, eventually, an impairment.
Debt, for the first time in a decade. Cash + marketable securities went $77.8B (FY24) → $44.5B (Q3 2025) — down $33B in 9 months. That’s why Meta issued $30B in senior notes in November. Long-term debt nearly doubled, and net debt-to-equity went from 3% to 22% in one year. This is a company that ran on essentially zero debt for the first 18 years of its public life. They couldn’t fund this build from cash flow alone, and they didn’t pretend they could. Listen for: cash position refilled, or another debt raise.
Headcount. Last Thursday’s layoff announcement is a strategic shift. Just last quarter they were guiding FY26 expenses up 38-44% y/y. Now 8,000 employees gone, 6,000 open roles eliminated, with the memo’s stated reason being to offset AI investments. Microsoft announced 7% buyouts the same day. Listen for: dollar value of FY26 opex savings, and any change to the FY26 expense guide of $162-$169B.
If you’re looking for risk to materialize, look to Hyperion first. Hyperion is the codename for Meta’s largest single data center project in Richland Parish, Louisiana. The off-balance-sheet treatment of the Blue Owl JV makes Meta’s reported leverage look better than it is, and if the SEC or Ernst & Young ever forces consolidation, $46B+ of liabilities flow onto the balance sheet. Meta’s own 10-K doesn’t even mention the transaction, per the auditor’s working notes.
One footnote that will trip up coverage: net margin “dropped” from 38% to 30% in FY25, but that’s because the OBBBA tax law triggered a $15.93B one-time charge to write down deferred tax assets, taking the effective tax rate from 12% to 30% in a single year. Cash taxes barely moved. We expect this to be pure accounting noise and the FY26 tax rate normalizes back to ~14% per BofA.
In the end, Meta has grown operating income while doubling capex, doubling debt, absorbing a $16B tax hit, and carrying a $19B/yr losing segment all in the same year. That doesn’t happen without an ad engine that is genuinely accelerating, and without aggressive use of every available financial lever. If the ad engine wobbles Wednesday — pricing decel, FoA margin compression, EU-related softness — none of the other levers can carry the build. If the engine holds and they show a credible RL or AI monetization milestone, the bet gets cheaper to defend by the quarter. Listen for ad price acceleration, capex 2026 guide (does it cross $100B?), RL loss trajectory, layoff opex savings, and whether cash gets refilled or they raise more debt.
All numbers are from Revelata’s deepKPI, which gets you 1-click auditing on every data point and hooks into Excel, Claude, ChatGPT, OpenClaw, and your API. Check it out -- deepKPI unlocks unbelievably powerful exploration and data for financial analysis.
 | A guest post by
|