Analysis: AI's Two-Dimensional Exponential Growth Explains Extended Infrastructure Supercycle
Summary
- • AI demand scales across two exponentials — user count and tokens per user — unlike internet's single dimension
- • Enterprise AI spend averages $12/person/month today; analysts see a plausible path to $1,000/month
- • AI user penetration crossed 50% in 2026; per-user token consumption remains in its early growth phase
- • Two multiplying growth curves set a much higher infrastructure ceiling, extending the semiconductor supercycle
Details
Two-dimensional AI growth vs. internet's single dimension
The internet scaled on one exponential: user count with flat-fee pricing. AI scales on two multiplying exponentials — user penetration and tokens-per-user — raising the infrastructure ceiling dramatically higher and sustaining demand far longer than the internet cycle.
AI user penetration crossed 50% threshold in 2026
The first growth dimension (user count) has now passed the midpoint of its S-curve, historically the point of inflection where adoption growth begins slowing — but the second dimension (tokens per user) is now accelerating to compensate.
Enterprise token spend: $12/month now, ~$1,000/month potential
Median US enterprise AI spend is currently $12 per person per month. The analyst argues this could plausibly reach ~$1,000/month long-term (roughly 10% of a white-collar salary), representing nearly two orders of magnitude of remaining headroom.
Relay of sigmoid curves sustains semiconductor buildout
The two growth dimensions are running sequentially: user penetration is decelerating while tokens-per-user accelerates, creating a relay of S-curves that extends compute demand well beyond prior technology cycles.
High per-token marginal cost distinguishes AI from internet era
Unlike the internet, where marginal cost of an additional user request was near zero, AI inference carries significant per-token cost — shifting business models from 'traffic × conversion rate' to gross margin per token, reinforcing the need for sustained infrastructure investment.
Dot-com bubble burst at same 50% penetration milestone
The dot-com bubble burst in March 2000 when US internet penetration hit ~50% — the same inflection point AI user count has now crossed. The key difference: AI's second dimension (token usage growth) prevents the same demand collapse from occurring at this milestone.
Analysis sourced from TechMeme-aggregated analyst piece (trust score: 0.65). Represents analyst opinion, not primary company reporting.
What This Means
This analysis offers a compelling structural explanation for why the current AI infrastructure buildout will outlast the internet bubble: two independently growing exponentials (users and token consumption per user) multiply together to create a demand ceiling far higher than any single-dimension technology cycle. With enterprise token spend still near the bottom of its S-curve — and nearly two orders of magnitude of potential growth ahead — the semiconductor supercycle supporting AI has a much longer runway than skeptics assume. For investors, infrastructure companies, and AI developers, this framing helps explain why hyperscaler capex commitments continue to climb despite already-significant user penetration, and why the current AI hardware cycle is structurally different from prior tech buildouts.