Skip to content
Slow Variables
Ask

Compute & physical

Chips' share of the stack's gross profit

The chip layer's share of the AI stack's gross profit per calendar quarter: NVIDIA and AMD whole-company gross profit (filed) over that plus Microsoft Intelligent Cloud's gross profit (filed) and the frontier labs' estimated gross profit (grade C). Derived metric gross_profit_share_by_layer, chips slice.

Reason through this

The reading

The site's main measure of who keeps the money, and so far the chip makers are keeping more of it. The total includes an estimate for OpenAI and Anthropic, so a revision to that estimate can move the share even when no company files anything new; a second chart, built only from filed operating income, is the check.

Direction over the last 5 readingseach dot links to its record
filed compute semis · 63.9% · 2025-03-31 · estimatefiled compute semis · 65.7% · 2025-06-30 · estimatefiled compute semis · 69.6% · 2025-09-30 · estimatefiled compute semis · 71.8% · 2025-12-31 · estimatefiled compute semis · 73.6% · 2026-03-31 · estimatefiled compute semis · 72.6% · 2026-06-30 · estimate
in the rule's windowearliermoves within ±2.0 pts read as stable, and a move counts only if more than half the steps go its way; this window moved +7.0 pts and 3 of 4 steps rose; higher reads concentrating
estimate

6 readings drawn from the derived metric gross_profit_share_by_layer; 43 observations in the evidence.

Derived (gross_profit_share_by_layer) from: SEC EDGAR XBRL company facts; Epoch AI data hub — AI companies; SEC EDGAR 10-Q/10-K XBRL instances (segment data). SEC EDGAR, XBRL company facts API. Epoch AI, 'Data on AI Companies' (CC BY 4.0), epoch.ai/data. SEC EDGAR, inline XBRL filings (segment disclosures).

The numbers
as ofreadingvaluerecord
2026-06-30filed compute semis72.6%estimate
2026-03-31filed compute semis73.6%estimate
2025-12-31filed compute semis71.8%estimate
2025-09-30filed compute semis69.6%estimate
2025-06-30filed compute semis65.7%estimate
2025-03-31filed compute semis63.9%estimate

Why it matters

Gross profit is what is left after paying to deliver the product, so a layer's share of it is the plainest answer to who keeps the money. If rents migrate up the stack as models commoditise, this share falls; if the scarce part stays scarce, it holds or rises.

What cuts against it

NVIDIA and AMD are whole-company figures, so gaming and PC chips count as AI; only one cloud segment files its cost of revenue; the lab slice is an estimate that overstates true gross margin; applications publish nothing usable and are left out, which flatters every measured layer's share.

How it is measured

Four calendar quarters, the same rule as the operating-income stack; a move under two points is fiscal-calendar noise. Rising = profit concentrating at the chip layer; falling = migrating up the stack.

Applied to metric:gross_profit_share_by_layer.

Derived rows (6)
as ofdimsvalueinputs
2026-06-30filed compute_semis72.6%sources123+13 more
2026-03-31filed compute_semis73.6%sources123+7 more
2025-12-31filed compute_semis71.8%sources123+13 more
2025-09-30filed compute_semis69.6%sources123+7 more
2025-06-30filed compute_semis65.7%sources123+13 more
2025-03-31filed compute_semis63.9%sources123+7 more
Latest readings
  • 2026-06-30 · 72.6%as of 2026-06-30(16 obs)
  • 2026-03-31 · 73.6%as of 2026-03-31(10 obs)
  • 2025-12-31 · 71.8%as of 2025-12-31(16 obs)
  • 2025-09-30 · 69.6%as of 2025-09-30(10 obs)
  • 2025-06-30 · 65.7%as of 2025-06-30(16 obs)
  • 2025-03-31 · 63.9%as of 2025-03-31(10 obs)
Why this status, and its history

Evaluator: 69.5% (Epoch AI data hub — AI companies, 2026-06-30) reads concentrating over 4 periods with a dead band of 0.02. Auto-reason; rule rationale: Four calendar quarters, the same rule as the operating-income stack; a move under two points is fiscal-calendar noise. Rising = profit concentrating at the chip layer; falling = migrating up the stack.

  1. 2026-09-14unmeasured to concentratingconf — → 60 · the evaluator (automatic)

    Evaluator: 69.5% (Epoch AI data hub — AI companies, 2026-06-30) reads concentrating over 4 periods with a dead band of 0.02. Auto-reason; rule rationale: Four calendar quarters, the same rule as the operating-income stack; a move under two points is fiscal-calendar noise. Rising = profit concentrating at the chip layer; falling = migrating up the stack.

    sources123456+10 more
Why this confidence

Rests on 43 observations from Epoch AI data hub — AI companies, SEC EDGAR 10-Q/10-K XBRL instances (segment data), SEC EDGAR XBRL company facts; best evidence tier 4, grade A; current.

Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; 0–49 limited or vague evidence.

Predictions that lean on this