Amphenol posted 52% revenue growth, 450bps of margin expansion, and closed its largest acquisition in history. The stock dropped. DuPont decomposition explains why: 86% of the balance sheet growth came from a single quarter, and incremental returns fell below the cost of debt.
86% of Amazon's ROIC decline is capital turnover compression — not margin deterioration. The 10-K reveals why: $59B in new servers, a depreciation reversal on AI hardware, and $440B in irrevocable commitments. At $200B capex, the margin of safety is compressing.
Meta's Family of Apps earns a 51.5% operating margin. Its ROIC is declining at -6.3 points per quarter. Both facts are true simultaneously. The gap between them reveals everything about the largest AI infrastructure bet in advertising history — and the accounting policies designed to make it look cheaper than it is.
FIX has 40.8% ROIC. PWR has 9.7%—despite 4x the revenue. The 'data center exposure' explanation is incomplete. The full story: FIX earns 2.5x higher margins AND turns assets 30% faster. Here's the DuPont analysis of six infrastructure contractors building America's data centers.
Both Jabil (29% ROIC) and Celestica (39% ROIC) rank among the best capital allocators in EMS. But DuPont decomposition reveals they get there through opposite mechanisms — JBL wins on turnover (13x), CLS wins on margin (10.2%). The counter-intuitive finding: CLS has better metrics AND faster growth, yet trades at an apparent discount. Here's what the data reveals.
Oracle's -$10B Q2 free cash flow triggered alarm bells. Decomposition reveals the collapse is 70% working capital timing, not structural deterioration. The real story: $248B in off-balance sheet lease commitments dwarfs their $100B balance sheet debt. This is the largest infrastructure bet in enterprise software history.
Three fuel cell companies target AI data center power, but execution quality varies dramatically. Bloom Energy's +4.5% ROIC vs Plug Power's -104% isn't a small difference—it's a fundamental divide between a functioning business and a value destroyer. Our SEC filing analysis reveals warning signs even for the winner.
GE Vernova's Electrification segment grew +32% YoY, but the company trades at 98.9x earnings with a 3.57% ROIC. Eaton's Electrical Americas grew +15% with 30.3% operating margins—twice GEV's margin at half the P/E. MetricDuck data shows why quality beats hype in AI infrastructure.
At $25B market cap and 41x revenue, Rocket Lab prices in Neutron success. Our 5-pass filing analysis reveals what the narrative misses: Launch Services growth inflated 32% by catch-up accounting, backlog concentration (69% top 5) exceeds revenue concentration, and gross margins compressed from 55% to 37%. Here's what to watch before the rocket flies.
AI infrastructure spending is approaching $500B annually across hyperscalers alone. Our research library covers capex efficiency metrics, depreciation manipulation detection, concentration risk screening, nuclear power for data centers, and supply chain analysis—everything you need to evaluate AI infrastructure investments.
The conventional wisdom says NuScale (SMR) is safer because it has NRC approval. But filing data reveals SMR paid $495M—$6.9M per reactor—to trigger a non-binding customer agreement. OKLO has regulatory uncertainty but uncommitted cash.
VST's +321% appreciation prices in nuclear optionality, but ignores a $510M battery fire disaster and antitrust lawsuit. CEG's 'capacity price boom' masks 970bps margin spread between regions. TLN's 0.9x interest coverage means debt service consumes nearly ALL operating income. We analyzed 9 filings to surface what the market overlooks.
While Caterpillar is up +60% YTD, most see it as construction/mining. But CAT's Energy & Transportation segment—which powers AI data centers—is the real growth story at +16.8% YoY with stable 20% margins.
EQIX's capex intensity jumped +8.7pp in 8 quarters. CCI's ROIC collapsed to -12%. ANET's customer concentration is HIGH (Microsoft + Meta). Standard screeners show none of this. Here's our 4-signal framework for screening AI data center infrastructure stocks.
Learn how to monitor $370B in AI infrastructure spending quarterly with a 3-metric framework. Track capex/revenue trends, depreciation manipulation signals, and growth alignment across Google, Microsoft, Amazon & Meta. Updated December 2025 with Meta's $600B commitment and Michael Burry's depreciation thesis.
Michael Burry warns Big Tech will understate depreciation by $176B through 2028. Amazon is taking a contrarian approach—shortening useful lives and accepting a $700M profit hit while peers extend to boost earnings. Learn how to screen for earnings quality using the capex/depreciation ratio.