The Index that flipped 30 points.

I. The thesis, in one sentence.

When the cost of intelligence falls faster than the price of software, margins compress. That sentence is the whole macro view. The software-sector margin structure of the 2010s and early 2020s is not a permanent fact of the business model. It is a function of compute being expensive, inference being scarce, and software pricing being insulated from the cost of producing it. Each of those three is changing. We have been writing into that for a while.

II. The v1 reading was polite. The v2 reading is honest.

The v1 of the AI Margin Compression Index was composed of proxy signals — Nasdaq strength, ISM Services, GitHub trending, wage growth, PPI, industrial production. Six components, all reading the macro around the thesis rather than the thesis itself. The reading came in at 25.2 — thesis weak, equity markets unconcerned, services strong enough to absorb whatever compression pressure exists.

We added four signals to v2 — Microsoft, Alphabet, Meta, and Amazon AWS quarterly AI infrastructure capex, sourced directly from 10-Q filings. The four together carry fifty percent of the composition. The v1 proxies retain the other fifty percent. Nothing was removed. The reading is now 55.4.

Two readings on the same week, separated by thirty index points. One looked at how equity markets were pricing software margins. The other looked at how much capital is being deployed into the supply build that will compress those margins. The first thought markets were calm. The second saw that capex was running hot.

III. The market is reading the wrong half of the equation.

The thirty-point gap is the trade idea. Equity markets are pricing the demand side — the software running on the infrastructure — and they are pricing it as if the supply side does not exist. The supply side, meanwhile, is reporting record quarterly capex through the SEC. Both halves of the equation are public. Only one is being weighed.

Microsoft reported $27B in capex in its most recent quarter. Alphabet reported $28B. Meta and Amazon together added another $57B. Roughly $112B in a single quarter, with most of it attributable to AI infrastructure — GPUs, data centers, networking, the depreciation wave that follows. That is the build of supply that compresses margins. It is accelerating, and it is on filing.

IV. Why we are publishing this.

Palanor builds indices because the world is not in short supply of opinions — it is in short supply of instruments. An opinion that the software margin structure is unstable is a forecast. An index that reads twelve signals through a transparent methodology and produces a number that moved from twenty-five to fifty-five when we added the half of the data the market is ignoring — that is an instrument.

Instruments earn their authority by surviving the comparison. The v1 reading was defensible. The v2 reading is more defensible. The v3 reading, which will replace manual capex entry with an automated SEC EDGAR parser, will be more defensible still. Each version's history stays anchored to the composition it was scored against, so what we said yesterday does not silently revise.

V. What changes if this thesis is right.

If the cost of intelligence keeps falling faster than the price of software, several things follow. Software gross margins compress toward something closer to services margins. Pricing power moves to whoever controls the most differentiated layer — the agent, the data, the workflow, the brand. The Magnificent Seven becomes a more fragmented basket. The valuation multiples that supported the 2020-2024 software bull market come under quiet, structural pressure that does not arrive as a single moment.

Operators who watch the Index now — quarterly, alongside their own KPIs — have a single number that reads whether the compression is materializing in the data. Boards who carry it into the strategy review have a defensible reading to anchor the conversation. Investors who use it as one signal among many in their portfolio framework have an instrument that improves quarter by quarter without rewriting the thesis.

VI. The instrument.

The AI Margin Compression Index is one of twelve Palanor Indices — six house composites reading the broad economy, six thesis composites reading named positions Palanor has been writing into. The full catalog is at /indices; the AMC methodology page sits inside it. Each Index is built on the same signal lattice — the same FRED series, market prices, prediction-market depth, manual-entry data — scored through the same engine. Twelve numbers, one math.

VII. About this article.

This is the first article in The Palanor Index Series— methodology essays accompanying the platform's flagship instruments. Each Index gets its own essay over the coming weeks. The next is Capital Tightness, observed on the credit window for mid-market operators in 2026.

Palanor is the Enterprise Intelligence Platform. We exist to help stewards see clearly across long horizons. If this essay landed, the full platform is at palanor.com.