Axelrod Research — The Week: The Capex Bill Landed

The same quarter produced both the bull and the bear case for AI infrastructure, and the market picked winners by cash flow, not growth. Meta printed record revenue and watched free cash flow collapse 91%; Microsoft posted a milder FCF decline and had its best day since 2008. The through-line of every name we touched this week: the market is now pricing the cost of the AI buildout, not just the demand.

The Book

META — Meta Platforms — HOLD

What changed this week: Q2'26 (period ended 2026-06-30, reported 7/29) delivered the sharpest data point of the week: FCF $784M, −91% YoY, on $31.1B of capex plus ~$2.4B legal charges. Revenue was a record $60.8B (+28% YoY), net income $15.85B, op margin 31% (down 8pp YoY). Reality Labs lost $4.6B on $431M revenue. The market's verdict: the stock fell from $593.87 (7/27 close) to $556.71 (7/31 close, −6.3% over the week), with the bulk of the drop on the print.

Thesis check: The ad engine is intact — that is not the problem. The problem is that ad cash now funds a $31B/quarter buildout and a $4.6B/quarter Reality Labs drag. The El Paso data-center financing (BlackRock raising >$12B via an SPV for a Meta campus) confirms the capex trajectory is not slowing.

Bear case, argued honestly: If Q3 FCF does not recover once the legal charges roll off, this stops being a cash-return story and becomes a capital-intensive buildout with an uncertain payoff — at a high-multiple valuation. The $90B cash pile is the cushion, not the answer.

Trigger to watch: Q3'26 print — ex-legal-charges run-rate FCF. Recovery above ~$5B/quarter holds the call; another sub-$1B quarter forces a downgrade. Horizon: Two quarters.

GOOGL — Alphabet — ACCUMULATE

What changed this week: Alphabet crossed the threshold our macro memo flagged as the group's defining event: Q2'26 OCF $39.1B vs capex $44.9B → FCF −$5.9B, the first negative quarter since the 2004 IPO. FY capex guide raised to $195–205B (from $180–190B). Cloud grew ~+82%. The stock absorbed the news and rallied: from $326.56 (7/27 close) to $356.13 (7/31 close, +9.1% over the week), with most of the move on 7/31.

Thesis check: TTM FCF is still ~+$53B — this is a capex spike, not an earnings problem, and the market correctly read it that way. One caveat from this week's claim-checking: the circulating claim that Alphabet's Anthropic stake is worth $94B is unsubstantiated — no primary source, and it implies an implausible Anthropic valuation. Don't underwrite the stock on that number.

Trigger to watch: Q3'26 — whether negative FCF is a one-quarter spike or the new run-rate. Horizon: 12 months.

AMD — Advanced Micro Devices — HOLD

What changed this week: Our daily made the call and the week tested it: after the Anthropic partnership announcement (up to $5B equity investment, up to 2 GW of MI450 GPUs, first GW in H1 2027), the stock sold off from $521.95 (7/24 close) to a 7/29 low close of $429.56, then recovered to $476.15 (7/31 close, −8.8% over the week). The selloff tested the thesis; the thesis held — but note what it cost.

Thesis check: The deal is real but both figures are ceilings ("up to"), not commitments — the newsletter framing overstated certainty. The harder problem is the balance sheet: $5B is nearly all of AMD's $5.6B cash (Q1 FY26, period ended 2026-03-28). This requires phasing or debt. ROCm still trails CUDA; Claude optimization narrows, not closes, the gap.

Trigger to watch: Q2 FY26 earnings (imminent) — capex/funding plan for the $5B, and any MI450 ramp detail. Horizon: 12–18 months.

LMT — Lockheed Martin — BUY

What changed this week: The cleanest verified event of the week: a U.S. Army award worth up to $58.6B over ~7 years to triple Patriot PAC-3 MSE production, driven by stockpile depletion from the Iran war and Ukraine. The stock did almost nothing on it — $582.60 (7/24 close) to $582.74 (7/31 close, +0.02%). A multi-year revenue underpinning for the Missiles & Fire Control segment priced at zero.

Thesis check: The award is confirmed by multiple outlets; the contract-structure detail ($53.86B modification on a $4.7B April contract) is plausible but not itemized. Execution risk on tripling production is real, but this is a demand-certain, multi-year backlog addition.

Trigger to watch: Q2 earnings commentary — whether the award was already embedded in guidance, and margin terms (fixed-price ceilings can compress on cost overrun). Horizon: Multi-year.

JPM — JPMorgan Chase — ACCUMULATE

What changed this week: Our daily flagged the asterisk and the week confirmed it: Q2'26 reported net income $21.2B includes $5.6B of one-time gains ($4.6B Visa exchange + $1.0B equity gains). The operating quarter is $16.9B net income / $6.14 EPS — still genuinely strong: CIB revenue +27%, IB fees +30% (best since 2021), AWM AUM $5.1T. The stock held roughly flat: $356.20 (7/27 close) to $351.79 (7/31 close, −1.2% over the week).

Thesis check: Dimon's own warning — geopolitical "tectonic plates," sticky inflation, elevated asset prices — is the honest bear case, plus card charge-offs at 3.34%. Equity Markets +86% will not repeat. But 14.1% CET1 and TBVPS +10% say the franchise compounds through it.

Trigger to watch: Q3'26 trading revenue normalization and card credit trends. Horizon: 12 months.

MSFT — Microsoft — HOLD

What changed this week: The counterweight. FCF fell ~23% but stayed positive — the only mega-cap that managed it — and the market paid for the distinction: +15.5% on 7/30 to $451.10 (7/30 close; fresh quote unavailable at publication, data-provider rate limit). The open item from our memo: the FY26 10-K (filed 7/29) reportedly shows uncommenced lease commitments leaping from $196.6B to $329.1B — specific, checkable, and not yet independently verified. Treat as unconfirmed until the 10-K commitments note is read.

Trigger to watch: Confirmation of the lease figure in the 10-K; whether the +15.5% re-rating holds or mean-reverts. Horizon: 12 months.

AMZN — Amazon — HOLD (needs confirmation)

The newsletter's two claims both failed verification: the "negative Q2 FCF" claim could not be confirmed from the 10-Q this session, and the "rose 15.44% today" claim is contradicted by the only verifiable session (+3.9% on 7/30, close $235.50). The hyperscaler capex story applies to AWS, but we won't underwrite on unconfirmed numbers. Trigger: extract Q2'26 OCF and capex from the 10-Q (filed 7/31). Horizon: pending data.

The rest of the week, briefly

• CBZ — AVOID chasing: Grant Thornton acquiring CBIZ at $55.00 cash/share, $5B EV (verified). The newsletter's "17.8% premium" is mis-stated — widely reported as ~54% over 30-day VWAP. With a hard cash ceiling, the equity is now an arb spread, not a thesis.
• TEM — WATCH: Tempus-Personalis deal verified ($1.5B, $16.25/sh), but ~67x revenue for a $22.4M-revenue quarter company with ~15–20% dilution. Wait for the integration plan and Q2 print.
• BLK — HOLD: $12B bond sale for the Meta El Paso campus is verified, but it's a GIP/SPV financing — off BlackRock's balance sheet. The newsletter's "BlackRock tapped the bond market" framing materially misleads.
• MS — HOLD: ~2,000 job cuts verified (~2.5% of ~80,000 staff); the "AI replacing roles" framing is overstated. The separate Epic Energy pipeline acquisition is an IM fund deal, immaterial to MS equity, and unverified.
• STLA — WATCH: New Jeep/Ram CEOs verified, but this is crisis management after a ~40% YTD decline and a CEO ouster — not routine succession. Turnaround will take quarters to evidence.
• WMT — HOLD: The U.S. COO transition is verified but immaterial — a division-level personnel change, not a P&L event.
• WFC / WBD — no call: Dividend raise (WFC) and Paramount-deal delay (WBD) both unverified against primary sources this session.

Changed my mind

Nothing flipped this week. AMD was stress-tested by the post-deal selloff and held; META's FCF print was worse than feared but the thesis was already cautious. A week of holding is a position.

What I'm watching next week

• AMD Q2 FY26 earnings (early August) — funding plan for the Anthropic commitment.
• MSFT FY26 10-K commitments note — the $329.1B lease figure, confirm or kill.
• AMZN Q2'26 10-Q cash-flow statement — the negative-FCF claim.
• WMT Q2 FY27 earnings (mid-August) — margin and automation returns.
• Whether MSFT's +15.5% re-rating holds through a full trading week.

Sources

• Alphabet Q2'26 10-Q
• Alphabet negative FCF coverage
• Meta Q2'26 results
• Meta FCF $784M / capex $31.1B
• AMD–Anthropic press release
• AMD 10-Q (Q1 FY26)
• JPMorgan Q2'26 earnings release
• MSFT FY26 10-K
• AMZN Q2'26 10-Q
• Lockheed $58.6B Patriot award
• BlackRock $12B data-center bond
• CBIZ take-private
• Tempus–Personalis deal
• Stellantis Jeep/Ram appointments
• Price data: Polygon/Massive daily aggregates, 7/24–7/31/2026, accessed 2026-08-01

Axelrod Research is independent equity analysis, for information only, not investment advice.