Declining cash flow generation due to rising Capex

MM: Largest risk esp. if they cannot deliver on the revenue side!

<aside> 🖼️

Overview

Management guides 2025 CapEx of $66–72 billion (more than $30 billion higher than in 2024), with significant growth expected in 2026 (Q2 2025 Earnings Press Release).
Management guides 2025 OpEx of $114-118 billion (around $20 billion higher than in 2024), with significant growth expected in 2026 driven by depreciation costs and stock-based compensation (associated with new hires) (Q2 2025 Earnings Press Release). </aside>

<aside> 💡

Investor Takeaway

Meta’s AI infrastructure spending now rivals hyperscalers’, creating potential cash-flow headwinds if core earnings soften, and while underinvestment risk is lower than overspending, Zuckerberg’s history of pursuing unproven projects like the Metaverse adds execution risk. </aside>

<aside> 🔮

Prediction (Base)

Meta Platforms Valuation Model (Google Sheets)

Key Item	2025 E	Y/Y	2026 E	Y/Y	2027 E	Y/Y	2028 E	Y/Y	2029 E	Y/Y	2030 E	Y/Y
Capex	$71 billion	81%	$120 billion	69%	$140 billion	17%	$130 billion	7%	$120 billion	-8%	$100 billion	-17%
Opex	$116.4 billion	22.4%	$154.8 billion	33.0%	$185.1 billion	19.6%	$211.5 billion	14.0%	$237.2 billion	12.0%	$251.0 billion	6.0%
Depreciation	$19.6 billion	26.6%	$33.9 billion	73.0%	$51.2 billion	50.7%	$67.9 billion	32.6%	$83.4 billion	22.7%	$92.3 billion	10.7%

<aside> 📈

Key Catalysts

The risk of underinvesting is higher than the risk of overinvesting since Meta AI is a disruptive technology: https://forum.investmentwiki.org/t/meta-platforms-ai-developments/136/164?u=aron
Gen AI infrastructure capacity can be used in core AI, minimizing downside risks if GenAI benefits don’t materialize. </aside>

<aside> ⚠️

Key Downsides

Ongoing aggressive capex and opex spend will lower profitability. Also, heavy capex and opex spend may fuel speculative behavior among investors like what happened in 2022 i.e sell now and buy later (min 21:54): https://www.youtube.com/watch?v=z6PHZJLo2Sk
Rising capex and opex yet Llama models have underperformed that of rivals: Meta AI performance and competitiveness (Superintelligence) was slow at first
Higher CapEx reduces Meta’s near-term shareholder returns and could raise debt levels.
CEO Zuckerberg tends to get excited on new bets like he did with Metaverse, making him bet excessively on unprofitable or unproven ventures like his current pursuit for superintelligence.
Rising cloud-compute contracts which are recorded as expenses understates the true investment level of Meta and present significant risk to earnings.
- Meta entered into a $40B in multi-year cloud-compute contracts in Q3 2025 </aside>