# AI prices vs commitments: spenders held, the basket still fell
Latent Equity · published Nov 5, 2026 · AI capital ledger
Post: https://x.com/Latent_Equity/status/2118365925879451655

The big AI spenders' commitments mostly held or grew after their Q3 reports, and prices kept falling anyway.
Our 13-name AI basket closed Nov. 4 17.2% below its Oct. 6 peak, down from 11.3% below on Oct. 29. Only 2 of the 13 are above their 50-day average ($MSFT and $AMZN), and 4 of 13 are above their 200-day, down from 9 (our calc).

The split since Oct. 29 runs along the chain. The spenders held up: $AMZN +9.1%, $MSFT +4.8%, $GOOGL +1.8%, $META +0.9%. The chip and hardware names fell: $ARM -16.2%, $MU -14.9%, $SMCI -14.0%, $AMD -13.7%, $ORCL -11.3%.

Commitments, Q3 updates:
$AMZN: about $220B of 2026 cash capex, in its earnings release.
$META: 2026 range narrowed to $135-145B, and it said in its release it will "continue investing aggressively" next year.
$GOOGL: $195-205B for 2026, and 2027 to "increase significantly", the company said on its call.
$MSFT: calendar-2026 capex trimmed to about $170B from about $175B because outside data-center capacity arrived late, the company said on its call.

Backlog, the revenue customers have signed for but that hasn't been booked yet: $MSFT $683B, down $1B from last quarter (flat) and up 72% from a year ago. $GOOGL $567.7B, up $48.2B. $AMZN $578B, up $82B.

One layer down, a supplier rather than a big spender: $CLS spent 66% of its operating cash flow on capex in the first 9 months of 2026, vs 26% a year earlier, and 2% on buybacks vs 29%. It sold $3.45B of new stock on Aug. 7, and its share count is up 9.7% from a year ago. Inventory went from $2.19B at the end of 2025 to $4.51B. Its 2027 revenue outlook is $40B, up 90%.

Our tests, straight:
- Two or more big spenders cut next year's capex: not tripped. None did.
- Any backlog falls from one quarter to the next: tripped on the letter by $MSFT's -$1B (-0.1%). That is flat in substance, but the test said any.
- Basket breadth still falling 4 weeks after guides held: the clock started Oct. 29. One week in, breadth has not improved. It can be judged around Nov. 25.

What would prove the read wrong: two or more of the big spenders cut next year's capex plans, or backlogs actually shrink rather than hold flat.

General research, the same for every reader. Not investment advice.