Meta's 51% Margin Hides a 20% ROIC Decline

Last Updated: February 12, 2026 Data Currency: Meta Platforms FY 2025 10-K filing (January 29, 2026). META SEC Filings

Meta's Family of Apps earns a 51.5% operating margin — extraordinary by any measure. Its return on invested capital is declining at -6.3 percentage points per quarter. Both facts are true simultaneously.

The gap between them tells the whole story about the largest AI infrastructure bet in advertising history. Invested capital is exploding faster than revenue — not because the business is failing, but because Meta is deploying capital at unprecedented scale into AI infrastructure whose returns are unproven. And its accounting policies are designed to make that bet look cheaper than it is.

Return on invested capital (ROIC) measures how efficiently a company generates after-tax operating profits relative to the capital invested in the business. It's the single best metric for evaluating whether Meta's AI spending is creating or destroying shareholder value — and right now, the trend is unambiguous.

This isn't a debate between bulls and bears. It's a decomposition of what the FY 2025 10-K actually reveals about the financial engineering beneath the headline numbers.

---

Meta's Depreciation Policy: Why Amazon Went the Opposite Direction on AI Hardware

Start with the most analytically important finding in the 10-K — the one that reframes everything else.

Effective January 1, 2025, Meta extended the estimated useful life of certain servers and network assets to 5.5 years. The financial impact:

• Reduced FY 2025 depreciation expense by $2.92 billion
• Increased net income by $2.59 billion after tax

In February 2025, Amazon shortened the useful life of a subset of its servers, explicitly citing "the rapid pace of AI and machine-learning innovation." Amazon took a $700 million earnings hit by accelerating depreciation.

Both companies operate data centers filled with identical hardware — Nvidia GPUs, custom networking equipment, high-density servers. They face identical technology obsolescence curves. Yet they reached opposite accounting conclusions about how long these assets generate economic value.

This isn't a trivial accounting footnote. It's two competing hypotheses about AI hardware economics, and the financial stakes are enormous.

What Each Hypothesis Implies

Meta's hypothesis (5.5-year life): AI hardware improves incrementally. Today's GPUs remain productive alongside newer chips for over five years. There's no cliff-edge obsolescence — older hardware handles inference and less demanding workloads while newer hardware handles frontier training.

Amazon's hypothesis (shortened life): AI hardware faces rapid obsolescence. Each generation of chips offers 2-3x efficiency gains per watt. Older GPUs become economically unviable — not because they stop working, but because their power consumption makes them more expensive to operate than replacing them with newer hardware.

Michael Burry, through Scion Asset Management, has sided with Amazon. His specific projections: hyperscalers will collectively understate depreciation by $176 billion between 2026 and 2028. He estimates Meta's reported earnings will be inflated by approximately 20.8% by 2028. His reasoning: A100 GPUs "take 2-3x more power per FLOP" than H100s, meaning their economic useful life is shorter than their physical useful life — the asset still works, but it costs more to run than to replace.

The Extension Timeline Tells Its Own Story

Meta has systematically lengthened server depreciation schedules since 2013. Every extension boosted reported earnings:

Period	Server Useful Life	Direction	Earnings Impact
2013-2018	~3 years	Baseline	—
2019-2021	~4 years	Extended	Positive
2022 Q2	4.5 years	Extended	Positive
2022 Q4	5 years	Extended	Positive
2025+	5.5 years	Extended	+$2.92B
Amazon 2025	Shortened	Opposite	-$700M

The pattern is unidirectional. In twelve years, Meta has never shortened a depreciation schedule. Meanwhile, Amazon absorbed a $700M hit because it concluded the economic reality demanded it.

The Normalized Earnings Test

Strip the $2.92B depreciation benefit from FY 2025 results:

Metric	As Reported	Depreciation-Adjusted	Difference
Net Income	$60.46B	$57.87B	-$2.59B
Net Income YoY	-3.1%	-7.2%	4.1pp hidden
EPS (diluted)	$23.49	$22.48	-$1.01
P/E (at ~$670)	28.5x	29.8x	+1.3x

The headline story is "net income dipped 3% on higher taxes." The depreciation-adjusted story is net income declined 7% — masked by an accounting policy change that management controls. The extension didn't change any economic reality. It changed how the economic reality is reported.

This matters more going forward than backward. The $72.2B in 2025 capex now depreciates over 5.5 years instead of 5. The filing is explicit about what's coming: "We anticipate making capital expenditures of approximately $115 billion to $135 billion in 2026 to support our AI efforts and core business." Note the conflation: "AI efforts and core business" with no breakdown between the two. All of that capex enters the 5.5-year depreciation schedule. The gap between reported and economic earnings compounds every year that the policy remains in place.

---

Meta's ROIC Is Declining 20% — DuPont Decomposition Shows Why

The depreciation paradox sets the stage. Now look at what it does to capital efficiency.

Meta's FY 2025 ROIC is 22.7%, down from approximately 28% the prior year — a 20.8% year-over-year decline. The 8-quarter trend shows compression of -6.3 percentage points per quarter. This is one of the fastest rates of ROIC decline in Meta's history, rivaling the 2022 efficiency crisis.

The question isn't whether ROIC is declining — it clearly is. The question is why, and whether the cause is temporary or structural. DuPont decomposition answers this.

DuPont Decomposition: Three Hyperscalers, Three Structural Paths

ROIC = NOPAT Margin x Invested Capital Turnover

Each hyperscaler achieves returns through fundamentally different structural paths:

Company	ROIC (Q4)	NOPAT Margin	Capital Turnover	How They Generate Returns
META	33.5%	37.1%	0.904x	Pricing power (margin machine)
GOOGL	34.2%	28.0%	1.223x	Operational efficiency (asset velocity)
AMZN	25.0%	9.96%	2.508x	Capital turnover (extreme efficiency)

The non-obvious finding: Meta and Alphabet achieve nearly identical quarterly ROIC through opposite structural mechanisms. Meta generates returns through extraordinary pricing power — a 37.1% NOPAT margin that reflects its dominance in digital advertising. Alphabet generates comparable returns through superior operational efficiency — turning its invested capital 1.22x versus Meta's 0.90x.

This structural difference matters enormously for durability. Pricing power is vulnerable to disruption — if AI commoditizes ad targeting (making Meta's algorithms less differentiated from competitors'), margin compresses. Operational efficiency is more durable because it reflects ingrained processes and infrastructure scale that are harder to compete away.

Amazon is the outlier: razor-thin 10% NOPAT margin but 2.51x capital turnover. Amazon squeezes every dollar of invested capital harder than any other hyperscaler. This is why Amazon can justify massive capex — its infrastructure generates revenue at 2.5x the velocity of Meta's.

The Margin Trajectory Problem

Here's where the structural concern becomes specific:

Company	NOPAT Margin Trend (TREND8)	Direction
META	-99 bps/quarter	Compressing
GOOGL	-5 bps/quarter	Stable
AMZN	+12 bps/quarter	Improving

Meta's NOPAT margin is declining at -99 basis points per quarter. This means the structural advantage that defines Meta — ad pricing power — is beginning to erode even before the AI capex cycle peaks.

Alphabet's margins are essentially flat. Amazon's margins are improving as AWS leverage kicks in.

If Meta's margin advantage compresses while its capital base bloats, ROIC faces a double squeeze: lower numerator, higher denominator.

The Invested Capital Explosion

The denominator tells the rest of the story.

Meta's invested capital jumped $72 billion in one year — from approximately $219.7B to $291.9B. Of that, $53.8 billion hit in Q4 2025 alone as back-loaded capex flowed onto the balance sheet.

This is just the beginning. At $125B annual capex (2026 midpoint guidance), invested capital could reach $380-420 billion by end of 2027. Capital turnover would compress from 0.78x (FY 2025) toward 0.55-0.60x unless revenue growth keeps pace.

The recovery math is unforgiving:

Scenario	Revenue Growth	2026 CapEx	Projected 2028 ROIC	Recovery Timeline
Bull	25%/year sustained	$115B	~26%	3 years (2028-29)
Base	22%/year sustained	$125B	~20%	4-5 years (2029-30)
Bear	18%/year (deceleration)	$135B	~15%	5+ years (if ever)

In the base case, even with 22% revenue growth sustained for years, ROIC doesn't recover to pre-capex levels until 2029-2030. In the bear case — if advertising growth decelerates toward the upper teens — ROIC could remain below 20% for most of the decade. This would represent a structural repricing of Meta from a high-ROIC growth compounder to a capital-intensive infrastructure operator.

---

Meta's $131 Billion in Non-Cancelable Commitments: What the Footnotes Reveal

Buried in the commitments footnote, Meta discloses:

These are primarily for third-party cloud capacity, servers, and data center infrastructure. They are not owned assets — they're contractual obligations to pay third parties regardless of how AI monetization evolves.

What $131 Billion Means in Context

Comparison	Ratio
$131B vs. FY 2025 Revenue ($201B)	65% of annual revenue
$131B vs. FoA Operating Income ($102.5B)	1.28x annual FoA profit
$131B vs. Consolidated Operating Income ($83.3B)	1.57x annual profit
$131B vs. FY 2025 Free Cash Flow ($46.1B)	2.84 years of FCF

Even if Family of Apps margins hold at their extraordinary 51.5%, over a year's worth of FoA operating income is pre-committed to infrastructure that hasn't yet demonstrated AI-driven returns.

The commitment pipeline includes identifiable deals: a $6B fiber contract with Corning, a $27B joint venture with Blue Owl for a gigawatt-scale data center campus, $14.2B with CoreWeave, $10B+ with Google Cloud, and reported talks for $20B with Oracle.

The Flexibility Asymmetry

This is the analytical point that most commentary misses.

Amazon's data center capex flows into AWS — infrastructure with proven 2.5x capital turnover that monetizes immediately through cloud computing revenue. Every dollar Amazon commits to infrastructure has a demonstrable, measurable return profile.

Meta's data center capex must be justified entirely through indirect advertising improvements — making AI models better at targeting, measurement, and content recommendation. There is no direct revenue line for "AI infrastructure." The payoff is diffuse, delayed, and unobservable at the infrastructure-spending level.

If AI monetization disappoints, Amazon can flex its capex program because AWS customers provide immediate price signals. Meta cannot flex — it has $131 billion in contractual obligations that must be paid regardless of whether Advantage+ advertising efficiency improvements plateau.

The option value is asymmetric: upside is substantial if AI capex generates proportional advertising returns. But the downside floor is $131 billion in committed spending with no exit ramp.

---

Meta's Free Cash Flow: Why $115.8B in OCF Isn't What It Seems

Meta's FY 2025 operating cash flow of $115.8 billion looks powerful. Decompose it and the picture becomes more nuanced.

The filing itself reveals the composition:

Three of those four components are non-cash add-backs. Here's the full decomposition:

How OCF Is Constructed

Component	Amount	Nature
Net Income	$60.5B	Cash earnings
Depreciation & Amortization	$18.6B	Non-cash (understated per depreciation analysis)
Stock-Based Compensation	$20.4B	Non-cash (shareholder dilution)
Deferred Income Taxes	$18.8B	Timing (driven partly by "One Big Beautiful Bill Act")
Working Capital & Other	-$2.5B	Operating timing
Total OCF	$115.8B

$57.8 billion — exactly 50% of OCF — consists of non-cash add-backs (D&A + SBC + deferred taxes). These are legitimate components of operating cash flow, but they warrant scrutiny.

The depreciation add-back ($18.6B) is understated relative to economic reality, as established in the depreciation analysis above. The SBC add-back ($20.4B) represents real dilution to shareholders, not free money. The deferred tax benefit ($18.8B) is a one-time timing adjustment that could reverse as tax law stabilizes.

If deferred tax benefits normalize and depreciation is adjusted to economic levels, sustainable OCF is closer to $90-95 billion — still robust, but meaningfully lower than the headline $115.8B.

OCF-to-FCF Conversion: Where Meta Trails

The ratio of operating cash flow that survives capex tells a capital intensity story:

Company	OCF (TTM)	Capex	FCF	OCF-to-FCF Conversion
META	$115.8B	$72.2B	$46.1B	39.8%
GOOGL	$164.7B	$91.4B	$73.3B	44.5%
AMZN	$139.5B	$131.8B	$7.7B	5.5%

Meta converts less than 40 cents of every operating dollar into free cash flow. Alphabet converts 44.5 cents. Amazon is in a different category — its massive retail/logistics capex consumes nearly all operating cash flow, a strategic choice that funds market share.

The trend matters more than the absolute level. Meta's FCF declined 14.7% year-over-year despite 22% revenue growth. This is the first sustained divergence between revenue growth and FCF growth in Meta's history as a public company.

The 2026 Stress Test

Scenario	OCF Estimate	CapEx	FCF	FCF Margin
Low capex	~$141B	$115B	$26B	~10.6%
Base	~$141B	$125B	$16B	~6.5%
High capex	~$141B	$135B	$6B	~2.4%

At the high end of capex guidance, Meta's FCF margin collapses from 22.9% to approximately 2.4%. At the base case, $16 billion in FCF must cover dividends ($5.3B run rate), buybacks (historically $26B/year), and any debt service — a mathematical impossibility without cutting shareholder returns or tapping reserves.

---

Meta's $148 Billion Buyback Program: Where 82% Actually Went

Meta's capital return program is one of the most misunderstood aspects of the investment case.

Over the past 8 years, Meta has spent $148 billion on share repurchases. The result: net share count declined 3.9%. Approximately 82% of cumulative buybacks went to offset dilution from stock-based compensation.

The Math in FY 2025

Flow	Amount
Gross buybacks	$26.3B
Less: SBC dilution offset	-$20.4B
Net buyback benefit	$5.9B
Plus: Dividends	$5.3B
Actual shareholder return	$11.2B
Headline shareholder return	$31.6B
Sterilization rate	64% (FY 2025)

The headline figure — $31.6B returned to shareholders — overstates the economic reality by nearly 3x. The actual incremental capital return to shareholders was $11.2 billion. That's 5.6% of revenue, not the 15.7% implied by the gross figure.

The SBC Divergence Problem

Meta's SBC expense is growing at 22% year-over-year — the same rate as revenue. This means dilution is not improving with scale.

Company	SBC / Revenue	YoY Trend	Direction
META	10.2%	+22% YoY	Not improving
GOOGL	6.2%	-4.8% YoY	Improving
AMZN	2.1%	-21.3% YoY	Rapidly improving

Amazon compensates employees at one-fifth Meta's rate relative to revenue — and the gap is widening. Alphabet sits in between and is trending in the right direction. Meta is the only hyperscaler where SBC growth matches or exceeds revenue growth.

Q4 2025: The Signal

Q4 was the first quarter in recent memory where Meta executed zero net buybacks. The capital return ratio fell from a median of 64% to just 9%.

This is management signaling — whether intentionally or by mathematical necessity — that the AI capex program has consumed the capital that would otherwise flow to shareholders. When FCF compresses further in 2026, Meta faces a binary choice: allow dilution to accelerate (as SBC continues while buybacks decline) or draw down the $84.8B cash position during the most capital-intensive period in company history.

---

Reality Labs: $80 Billion in Losses and a Warning for the AI Bet

Reality Labs is more than a $19 billion annual drag on earnings. It's an analytical framework for evaluating Meta's AI capex bet.

The Numbers Everyone Knows

Year	Revenue	Operating Loss	Loss per $1 Revenue
FY 2022	~$2.2B	~$13.7B	-$6.23
FY 2023	~$1.9B	~$16.1B	-$8.47
FY 2024	~$2.1B	~$17.7B	-$8.43
FY 2025	$2.21B	$19.19B	-$8.69
Cumulative		~$80B

FY 2025 operating margin: -870%. At current cost structure, Reality Labs would need 64x its current revenue just to break even.

The framing is revealing: management presents a $19 billion annual loss as a plateau, not a concern — "remain similar" normalizes the spend rather than questioning it.

The Analytical Framework

Here's what the bulls and bears both miss about Reality Labs.

What RL reveals about Meta's capital allocation process: Management committed $80+ billion over five years to a bet where the filing itself acknowledges: "our ability to support our RL efforts is dependent on generating sufficient profits from other areas of our business." This is a tacit admission that Reality Labs is a permanent subsidy funded by Family of Apps profitability.

The language is identical year after year — "We believe this will become the next computing platform." No milestones. No timeline. No measurable criteria for success or failure.

Now apply this framework to the AI capex bet:

The AI infrastructure commitment ($115-135B in 2026 alone) is structurally similar to the Reality Labs bet but 6x larger in annual scale. Both represent management conviction deployed at massive scale into unproven returns.

The critical difference: AI has demonstrable near-term ROI. Meta's Advantage+ advertising platform reached a $60 billion annual revenue run rate by late 2025, delivering $4.52 in advertiser returns per dollar spent. This is real, measurable evidence that AI investment improves the core advertising business.

The critical risk: Advantage+ monetizes existing ad inventory more efficiently — it's unclear how much represents genuinely new revenue versus better optimization of impressions that would have been sold anyway. If AI capex is primarily about efficiency gains (not revenue creation), there's a ceiling on how much advertising revenue growth it can drive. And if that ceiling is reached while $131B in commitments remain outstanding, Meta faces the same structural problem as Reality Labs: sunk costs in an initiative that management can't exit.

The filing provides no framework for distinguishing between these scenarios. That's the unresolvable uncertainty at the center of the investment case.

---

The Bull Case: What the Filing Actually Says About AI Revenue

Any honest analysis of Meta's ROIC compression must grapple with the evidence that AI capex is generating returns — and the filing provides specific language worth examining.

This is the strongest evidence for the AI investment thesis. Ad prices rose 9% in 2025, which Meta attributes to AI-driven targeting improvements. Advantage+ — Meta's AI advertising platform — reached a $60 billion annual revenue run rate by late 2025, delivering an estimated $4.52 in advertiser returns per dollar spent. These are not hypothetical gains; they're measurable improvements in Meta's core advertising product.

But parse the language carefully. "Which we believe is mostly due to" is hedged attribution, not definitive. Management is claiming credit for ad price improvements without providing the counterfactual — how much would ad prices have risen without AI improvements? In a strong digital advertising market (Google's ad revenue also grew), isolating Meta's AI contribution from overall market tailwinds is inherently difficult.

The filing also discloses the permanent headwind on the other side:

"For the foreseeable future" is the critical phrase. This disclosure has appeared in every Meta 10-K since Apple's ATT changes in 2021. It's not new — but its persistence means regulatory drag on ad targeting is structural, not transitional. AI investment partially offsets this drag, which reframes the investment case: some portion of the $115-135B in capex isn't generating new revenue — it's defending existing revenue against ongoing regulatory erosion.

The investment question becomes: does AI capex generate enough incremental revenue beyond the regulatory offset to justify the ROIC compression? The filing provides evidence for both sides but answers definitively on neither.

---

Peer Comparison: Three Business Models, Three Risk Profiles

Three Observations That Require Analytical Work

1. Meta's ROIC Advantage Is a Margin Story — and Margins Are Peaking

Meta's Q4 ROIC (33.5%) and Alphabet's (34.2%) are nearly identical. But they arrive through fundamentally different paths. Meta's 37% NOPAT margin is 32% higher than Alphabet's — offset by Alphabet's 35% capital turnover advantage.

This means Meta's returns depend on maintaining extraordinary pricing power in digital advertising. If AI makes ad targeting more commoditized (the same models available to smaller competitors), Meta's margin advantage erodes. Alphabet's returns depend on operational efficiency — processes, infrastructure scale, and cost discipline — which are structurally harder to disrupt.

The TREND8 data confirms the concern: Meta's margin is compressing at 99 bps per quarter. Alphabet's is flat. Amazon's is improving. The margin machine is showing the first signs of structural stress.

2. Alphabet's Capex-to-Depreciation Is Actually Worse Than Meta's

This surprises most analysts. Alphabet's capex-to-depreciation ratio of 4.33x exceeds Meta's 3.95x. Both companies are investing more than four dollars for every dollar they depreciate — meaning both have depreciation expense that is significantly detached from actual investment pace.

The difference: Alphabet generates direct cloud revenue from its data center infrastructure ($110.8B Google Cloud revenue in 2025, growing 34% YoY). Meta does not. Alphabet's under-depreciation is partially offset by an immediate revenue stream. Meta's under-depreciation has no direct offset — the payoff comes only if advertising efficiency improvements materialize.

3. Amazon Is Playing a Different Game

Amazon's 2.0x capex-to-depreciation ratio and depreciation shortening tell a fundamentally different story. Amazon is the only hyperscaler acknowledging the economic reality of AI hardware obsolescence in its accounting.

Amazon's $65.8B in annual depreciation absorbs infrastructure costs far more quickly than Meta's $18.6B. When the AI capex cycle eventually moderates, Amazon's P&L will benefit immediately from lower depreciation expense. Meta will still be amortizing prior years' spending at artificially slow rates.

Amazon's ROIC structure — thin margins (10%) but extreme capital turnover (2.51x) — means it can justify massive capex because every dollar of infrastructure generates $2.51 in revenue. Meta generates $0.90. This isn't a quality judgment — it reflects different business models — but it explains why Amazon can absorb $131.8B in annual capex while maintaining positive ROIC, while Meta's ROIC compresses under $72.2B.

---

What Could Make This Analysis Wrong

Analytical honesty requires specifying the conditions that would invalidate the thesis. These are the specific data points that would signal the ROIC compression is temporary, not structural:

1. Q1 2026 advertising revenue growth accelerates to 25%+. If AI-driven targeting improvements generate genuinely incremental ad revenue (not just efficiency on existing inventory), the capex payoff timeline shortens dramatically. The FY 2025 run rate — ad impressions +12%, average price +9% — needs to inflect upward, not just sustain.

2. Meta's depreciation policy proves correct. If servers and GPUs genuinely remain productive for 5.5 years — if the economic useful life matches the accounting assumption — then Amazon is being overly conservative and Meta's reported earnings are accurate. Track hardware refresh rates and power cost disclosures in future filings.

3. Reality Labs losses decline materially. The filing says 2025 is the peak and 2026 losses will be "similar." If 2026 losses actually decline by 15-20%, it would demonstrate management's ability to scale back unproductive capital deployment — increasing confidence that the AI capex program can also be modulated.

4. Invested capital turnover stabilizes above 0.70x. If Meta can sustain capital turnover despite the capex surge, it means revenue growth is keeping pace with asset growth. The current 0.78x would need to hold or improve against the headwind of $125B in annual investment.

5. SBC as a percentage of revenue declines below 9%. This would signal that Meta is improving dilution efficiency as it scales, and that the buyback sterilization problem is transitional, not permanent.

None of these conditions are inherently unlikely. Several may materialize. But until they do, the 10-K data supports a thesis of structural ROIC compression driven by capital deployment at unprecedented scale, understated through accounting policy choices, and locked in by $131 billion in non-cancelable commitments.

---

Methodology

Data Sources

Filing	Company	Form	Period	Filed	CIK
Meta FY 2025 10-K	Meta Platforms	10-K	FY 2025	Jan 29, 2026	0001326801
Alphabet FY 2025 10-K	Alphabet	10-K	FY 2025	Feb 5, 2026	0001652044
Amazon FY 2025 10-K	Amazon	10-K	FY 2025	Feb 6, 2026	0001018724

All financial data extracted via MetricDuck's SEC filing analysis pipeline and quarterly 10-Q filings.

ROIC is calculated as NOPAT (Net Operating Profit After Tax) divided by Invested Capital. DuPont decomposition breaks ROIC into NOPAT Margin x Invested Capital Turnover. TREND8 represents the linear regression slope of the metric over the trailing 8 quarters, measuring acceleration or deceleration. Q.MED8 is the 8-quarter median, used as a stable baseline. Capex-to-depreciation ratio compares capital expenditures to depreciation and amortization expense, measuring how aggressively a company recognizes infrastructure costs.

Earnings quality assessed using MetricDuck's 5-pass filing intelligence system, which analyzes accounting policies, revenue recognition, hidden liabilities, risk factors, and segment performance from SEC filing footnotes. Depreciation policy analysis sourced from Note 1 (Summary of Significant Accounting Policies) of Meta's FY 2025 10-K. Non-cancelable commitment figures from the Commitments and Contingencies footnote. Segment financial data from the Segment Information footnote.

Limitations

• This analysis is based on public SEC filings and cannot assess non-public information, management intentions, or forward-looking outcomes.
• ROIC calculations use reported figures; if depreciation is understated (as argued), true ROIC may differ from both the reported and adjusted estimates presented here.
• Peer comparisons use company-wide metrics; segment-level capex and depreciation data is not available for all companies, which limits the precision of infrastructure-specific comparisons.
• The 2026 stress test assumes proportional OCF growth, which may not materialize if tax law, working capital dynamics, or revenue mix shift materially.

Disclaimer: This analysis is for informational purposes only and does not constitute investment advice, a recommendation to buy or sell securities, or an offer or solicitation of any transaction. Past performance is not indicative of future results. Always conduct your own due diligence before making investment decisions.

MetricDuck Research does not hold positions in META, GOOGL, or AMZN. This analysis is based solely on publicly available SEC filings and is provided for educational purposes.