Skip to content
brianletort.aiD. Brian Letort, Ph.D.

Cross-stack flywheel

AI Stack Weekly

September 5, 2026No. 20

For officers tracking AI market movement.

Complete issue archive.

Software, hardware, networking, capital, and power signals translated into a weekly decision read.

Subscribe via RSS

Published issues

20 issues · newest first

  1. No.20

    Week 36 of 2026 · September 5, 2026

    AI's scarce asset is now a governed path to service, not an announced model or megawatt

  2. No.19

    Week 35 of 2026 · August 29, 2026

    Navitas priced grid-to-xPU power delivery at 6.65¢ per SAM dollar — while Hot Chips argued about the next rack

  3. No.18

    Week 34 of 2026 · August 22, 2026

    The duration-mismatch thesis just moved into the GPU vendor's own 8-K — NVIDIA is now the residual-value underwriter of a 20-year OpenAI lease, and Marvell is teaching Google how to buy silicon on the revenue that has not yet arrived

  4. No.17

    Week 33 of 2026 · August 15, 2026

    Everything financed this week has a longer duration than the contract underneath it — and that mismatch now runs through all three lenses at once

  5. No.16

    Week 32 of 2026 · August 8, 2026

    Three frontier labs disclosed cyber incidents this week and all three trace to the same outside vendor — the measurement layer is now the concentration risk

  6. No.15

    Week 31 of 2026 · August 1, 2026

    The cheapest capability and the cheapest capacity both came from contracts and harnesses, not from chips or weights

  7. No.14

    Week 30 of 2026 · July 25, 2026

    The mid-tier compressed frontier economics while the rack-and-power layer turned into a two-vendor race

  8. No.13

    Week 29 of 2026 · July 18, 2026

    Model pricing is converging from both directions onto one frontier band — closed labs cutting down into $3/$15 while Chinese open-weight labs raise up into it — and the differentiation is shifting from price to who actually ships weights.

  9. No.12

    Week 28 of 2026 · July 11, 2026

    The harness became the product: OpenAI and Anthropic shipped rival work runtimes 48 hours apart — while memory and power marked the scarcity trade to market with $26.5B of real money.

  10. No.11

    Week 27 of 2026 · July 4, 2026

    The sell-side of AI compute gained two hyperscale balance sheets in 48 hours — just as the model layer started cutting prices.

  11. No.10

    Week 26 of 2026 · June 27, 2026

    The week shifted from model announcements to the physical bottlenecks that decide who can serve agentic demand.

  12. No.09

    Week 25 of 2026 · June 20, 2026

    Power policy and open weights became the week's leverage while the closed frontier stalled.

  13. No.08

    Week 24 of 2026 · June 13, 2026

    The frontier became a sovereign and capital-markets object, not just a benchmark race.

  14. No.07

    Week 23 of 2026 · June 6, 2026

    The AI factory became a power-and-fabric problem, not a model-release problem.

  15. No.06

    Week 22 of 2026 · May 30, 2026

    A quiet release week resolved last week's two biggest open questions — and moved the contested ground to optics and the agent control plane.

  16. No.05

    Week 21 of 2026 · May 23, 2026

    The bull thesis got every confirmation it needed. The strategic battleground moved to runtimes.

  17. No.04

    Week 20 of 2026 · May 17, 2026

    Frontier text took a breath. The constraint stack moved to supply, regulation, and pricing structure.

  18. No.03

    Week 19 of 2026 · May 8, 2026

    Capability stops being about parameters. The constraint stack flipped to power, fiber, and packaging.

  19. No.02

    Week 18 of 2026 · May 1, 2026

    Q1 broke the bear thesis. The marginal dollar started earning more — and the network still compounds.

  20. No.01

    Week 17 of 2026 · April 25, 2026

    The marginal dollar at every layer earns less. Except the network.

Methodology

What survives into every issue

  • Every quantitative claim is graded 1–5 on source quality. Anything 2 or below is flagged as noise.
  • Every prediction is falsifiable, time-bounded, and tied to a specific signal.
  • Tracked levers carry their current value, prior value, direction, threshold, and decision detail.
  • All sources are public. Independent analysis only.