AI Capex Transition & Value-Capture Matrix

Institutional modeling of AI capital re-allocation from upfront compute buildout to downstream software monetization

Scenario Presets:
Macro & Stack Levers Interactive Cascade
$220B
14.5%
45.0%
32.0%
18.0%
4 yrs
7 yrs
Enterprise Inflection
Year 4 (2028)
Cumulative Stack Capex
$2,184B
Terminal Software Share
44.8%
Terminal Stack ROIC Spread
+19.2%
AI Value Chain Revenue Pool Migration ($B/yr) D3 Dynamic Stream
Semiconductor & Hardware
Energy & Power Infrastructure
Cloud Platform & Models
Vertical Enterprise Software
Layer Return on Invested Capital (ROIC %) Trajectory Capital Efficiency Shift
AI Technology Stack Long-Term Returns Matrix Sector Economics & Payback
Value Stack Layer Y1 Revenue ($B) Terminal Revenue ($B) Avg Op Margin (%) Peak ROIC (%) Terminal ROIC (%) CapEx Payback (Yrs)

Value-capture matrix: growth, decay and model assumptions

Read the explanation

The matrix begins in twenty twenty five with two hundred twenty billion dollars of assumed capex. Hardware receives point five two times that value, one hundred fourteen point four billion revenue. Enterprise revenue begins at twenty five plus thirty five times one point three five, seventy two point two five billion. Bars encode both at two pixels per billion dollars. Four tier revenues sum to three hundred one point two zero two billion, exceeding capex because cloud and enterprise revenues are separately constructed downstream activity. This is not a conserved capital allocation or measured industry total. The native JSON export records seven annual samples through twenty thirty one. Original D3 absence prevents charts and table after metrics computation, while registered slider and download handlers still work. At thirty two percent adoption, the cloud inference-volume multiplier is one plus adoption times one point two five, giving one point four per year. Forty five percent annual price deflation supplies a point five five multiplier. Their product is point seven seven, so the token component shrinks from thirty five billion to twenty six point nine five after one step even as assumed volume grows. Bars show these token revenues at eight pixels per billion. Enterprise base revenue grows independently at one point three two per year and also adds one point three five times the shrinking cloud token revenue. Neither multiplier is estimated from observations. The first sampled enterprise revenue above hardware is labeled an inflection year; at defaults it is year seven, twenty thirty one, rather than an independently validated forecast. Return on invested capital divides assumed after-tax operating profit by capital defined as a fixed multiple of the same revenue. That revenue cancels. Initial hardware return is fifty eight times point seven nine divided by one point eight, about twenty five point four six percent. Enterprise uses twenty eight times point seven nine divided by point seven five, about twenty nine point four nine percent. Bars encode them at eight pixels per percentage point. Consequently revenue growth does not itself determine these returns; margin schedules and fixed capital multipliers do. Depreciation lifetime affects only the table payback caption, a fixed tier number times lifetime over four, and does not enter the computed annual trajectory. Native lifetime four to six therefore leaves trajectory unchanged, documented by actual JSON and CSV downloads. This source audit preserves original geometry and failures without investment guidance or public completion.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.