BARCLAYS 2026 BENCHMARK

AI Revenue & Unit Economics Waterfall Engine

Simulation Presets
Barclays estimate: $35–$40 per $100 of AI revenue
Jevons' paradox: sequential tool invocation increases compute load
Key Barclays Thesis: API pricing scales dynamically with usage, whereas flat subscriptions face heavy margin drag under agentic loops as customers execute high sequential call volumes for a fixed fee.
Cloud Revenue Pass-Through
$38.0M
Flows to AWS / Azure / GCP
Cloud Operating Profit
$15.2M
Hyperscaler operating earnings
Model Lab Gross Profit
$57.5M
Retained by frontier labs
Cybersecurity ARR Lift
$4.5M
CrowdStrike / telemetry pipeline
Subscription Margin Drag
-$8.4M
Overhead from fixed pricing
Effective Operating Margin
34.2%
Net blended lab conversion
AI Value Chain Waterfall Flow (Per $100M Baseline)
SVG Interactive Topology
Empirical Tech Earnings Grounding (Q3 2026 Actuals)
NVIDIA (NVDA)

$96.2B Total / $89B Data Center

Data Center revenue up 117% YoY. Hardware supplier capturing silicon revenue at ~75% gross margin directly from hyperscaler CapEx pools.

Salesforce Agentforce

$3.9B ARR / 7B Agentic Units

Enterprise ARR jumped 210% YoY. Shift from chat to autonomous sequential tasks where single customer requests trigger multi-system calls.

CrowdStrike (CRWD)

$333M Net New ARR (+51% YoY)

Every autonomous tool-call requires runtime inspection, data leak prevention, and identity authorization, routing a steady take to cybersecurity.

IREN & Nebius (NBIS)

$128.8M Cloud ARR / 1.2GW

Alternative GPU neoclouds showing 8x revenue growth as hyperscaler capacity limits push compute arbitrage to dedicated bare-metal farms.

Under the current 60% API / 40% Subscription allocation with an Agentic Depth multiplier of 5, the model lab transfers $38.0M to hyperscalers, while experiencing -$8.4M in flat-rate subscription margin drag.

Mapping AI Enterprise Value Flow and Subscription Margin Drag

Read the explanation

In the baseline AI enterprise model, every one hundred million dollars in revenue splits between dynamic API usage and fixed subscription contracts. Cloud hyperscalers siphon thirty-eight million dollars for compute, while cybersecurity subtracts four and a half million, carving model lab gross down to fifty-seven point five million. Increasing the agentic call multiplier to five sequential sub-calls expands unmetered inference load, triggering eight point four million dollars in flat subscription margin drag. After deducting subscription drag and overhead, the model lab generates thirty-four point one million dollars in net operating profit, yielding a thirty-four point one percent net operating margin.

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.