Methodology
Take a look at how the Bubble Pressure Index is built, and what it does and does not measure.
How to read the Bubble Pressure Index
Think of the Bubble Pressure Index like a stress gauge on a dashboard warning light, not a speedometer. A speedometer (or a stock chart) tells you where something is headed. A stress gauge just tells you how much pressure is building up in a system right now.
Here, the "pressure" is how much of a company's AI spending (data centers, chips, leased compute) is being fueled by debt-like obligations (leases, purchase commitments) relative to the actual cash the company is generating. A higher score means more of that spending is riding on financing rather than free cash flow.
This measures how reliant a company's AI spending is on debt/leases relative to its cash flow. It is not a valuation, price target, or buy/sell indicator.
Full Methodology
Capex/FCF = capital expenditure ÷ free cash flow (operating cash flow − capex), from the most recent fiscal year's 10-K. Shown as N/A when FCF is negative, since the ratio's sign flips and stops representing "burden" in that case.
Lease burden = (operating lease liability + finance lease liability + disclosed purchase commitments) ÷ market cap. Excluded (not treated as zero) for fiscal years before lease-liability disclosure was required (pre-ASC 842, ~fiscal 2019).
Cohort share = company market cap ÷ sum of market caps of all companies tracked on this dashboard. This is a proxy for "AI-sector share of the S&P 500" scoped to this tracked cohort, not the literal S&P 500 figure. That requires a paid data source we don't use here.
Bubble Pressure Index (0 to 100) = a weighted composite of the three components above, each normalized on a common 0 to 1 scale relative to the other tracked companies for that fiscal year, then combined into a single 0 to 100 score. Capex/FCF carries the heaviest weight, followed by lease burden, then cohort share. Missing components are excluded and remaining weights renormalized rather than treated as zero. Exact weighting is a proprietary MVP default, refined as the model matures, and is intentionally not published in full detail.
Gauge zones: 0 to 25 Conservative, 25 to 50 Moderate, 50 to 75 Stretched, 75 to 100 Highly Leveraged. Even split of the 0 to 100 range, an MVP default, not a statistically calibrated risk threshold.
Percentile = the share of the other tracked companies (same shared fiscal year) whose score is strictly lower than this company's. With 29 companies tracked, each step is about 1 in 28 peers, so values land in increments of roughly 3.6 percentage points rather than every whole number.
Companies are compared on the latest fiscal year shared by all of them, not each company's own newest filing. Fiscal year-ends differ (e.g. MSFT: June, NVDA: ~January, ORCL: May, AVGO: ~October/November), so comparing a company's newest filing against another's several-months-old one would be misleading.