Additions to property and equipment were $30.9 billion cash this quarter (about $31.9 billion with finance leases), up roughly 85% year-over-year. The CFO named the disconnect outright: capex growing faster than revenue. The defense rests on roughly two-thirds short-lived assets that correlate with revenue and a $633 billion RPO book — commercial RPO up 26% excluding OpenAI, making OpenAI a material part of the backlog. Guided to roughly $190 billion for calendar 2026. Whether the short-lived spend converts fast enough to close the gap is the open timing risk.

The tension stopped being hypothetical in early 2026 when analysts began downgrading seat-priced software names explicitly on per-seat AI risk. The disconnect they describe is the meter at work: seat-billed software compresses when AI removes humans, consumption-billed gets a tailwind. Seat incumbents are quietly adding consumption overlays — assists, flex credits — as defensive conversion. The billed unit predicts durability.

The NVIDIA CEO sells compute-as-utility: his racks are 'one giant GPU' and his efficiency gains are ~50× — directional, unverifiable. The course tags every claim to its speaker's book: the token seller sees 'uncapped' demand, the enrichment CEO calls enrichment 'THE bottleneck' (US <0.1% world supply; turbines sold out years) — trust structural facts, discount the definite article — the buyer preaches 'many winners, no zero-sum.' The clearest signal? Insiders with opposed books describing the same market differently.

FY2025 total revenue was $3,427.2M, up 28% from $2,684.3M. Our telemetry picks up that the top line is expanding while free cash flow remains strong at $914.7M for the year (~27% margin) and $289.1M in the quarter (~29%). GAAP net income stays thin at $107.7M (3.1% margin) because of stock‑based compensation and amortization, so we treat free cash flow as the cleaner read. Non‑GAAP operating margin runs about 22%.

Datadog is winning some hyperscaler business: our CEO said hyperscalers typically build everything themselves, yet they come to us to replace parts of their stack. As they roll out heterogeneous silicon—Trainium, Graviton, TPUs, Maia—traditional monitoring struggles on mixed fleets, making a neutral cross‑fleet observability layer more valuable. We’ve launched GPU Monitoring and are moving beyond inference into training workloads.

I map a feedstock mirror to the capex/revenue disconnect: Google fired its top annotation vendor and internalized raters; Meta's $14.3B Scale AI stake made Google, OpenAI, xAI pull contracts. Commodity-labeling margins hit a hard ceiling. The toll-booth owners — expert evaluation — stay untouched. Value retreats toward what can't be substituted, in-housed, or fair-used.

I see the disconnect they describe as the meter at work: seat-billed software compresses when AI removes humans, consumption-billed gets a tailwind. Early 2026 analyst downgrades of seat-priced names on per-seat AI risk made the stress test explicit. Seat incumbents are quietly adding consumption overlays — assists, flex credits — as defensive conversion. The billed unit, not the moat type, predicts durability.

The tension stopped being hypothetical in early 2026, when analysts began downgrading seat‑priced software names explicitly on per‑seat AI risk. I treat the billed unit as the moat’s stress test: seat‑billed software is structurally exposed—if AI removes or reduces humans, revenue compresses even as usage rises; consumption‑billed software enjoys a tailwind as agents consume more tokens, records or workflow executions. That’s why seat incumbents are quietly building consumption overlays—'assists,' 'flex credits'—as an explicit defensive conversion.

In-housing: Google terminated its major annotation contractor and internalized rater capacity; Meta's $14.3 billion stake in Scale AI prompted Google, OpenAI, and xAI to pull their work — a hard ceiling on commodity-labeling margins, though it never touches the toll-booth owners. The capex/revenue disconnect everyone debates upstream has a mirror down here: the "picks-and-shovels" data layer is mostly rented land, and the landlords are evicting the tenants. Value keeps retreating up the gradient toward what can't be substituted, in-housed, or fair-used.

Our telemetry shows the strongest sequential existing-customer usage growth since Q1 2022 — a $53M QoQ revenue add, a Q1 record. Because we bill on consumption, this reads deployed infrastructure being used directly. Excluding AI-native names, revenue growth re-accelerated to the mid-20s percent and net retention to the low-120s, suggesting the base is broadening. The same meter would show a pullback first.

Microsoft's quarter ending March 2026 (its fiscal Q3): total revenue of $82.9 billion, up 18% year‑over‑year and 15% in constant currency. We reported operating income of $38.4 billion, up 20%, and net income of $31.8 billion, up 23%. Cloud revenue reached $54.5 billion, up 29% at a 66% gross margin – down from 69% and guided to roughly 64% next quarter as AI‑infrastructure scaling compresses margins. Azure grew 40% (39% constant) and demand still exceeds capacity; the margin path under inference load is falling and we guide it lower again next quarter.

Snowflake's product revenue grew 34% with 126% net retention; these revenues scale with agent activity and genuinely corroborate infrastructure demand. The seat-priced leg is thinner: Microsoft's ~$37B AI run-rate is a management construct, Copilot at ~3.3% of 450M seats, and one survey shows 8% prefer it over ChatGPT when both are available. If the seat-to-consumption transition stalls or agent ROI disappoints at renewal, the productivity leg weakens while infrastructure legs hold. Blending them oversells the demand story.

From the best seats, the course confirms the buyer: power — not capital or chips — gates gigawatts, and memory bandwidth, not raw compute, is the wall. It refines with 20-year take-or-pay contracts and optical circuit-switching. It oversells where incentives predict: 'uncapped demand' is a token-seller's claim, the 2027-28 memory-glut counter-case goes unspoken, and the 'advanced packaging' layer has zero packaging content across twelve lectures. Clearest signal: insiders with opposite books describing the same market differently.

Look, you know, our filings show the concentration you're describing — top ten at 46% of revenue last quarter, up from 40% and 41% the two years before. Quite frankly, one customer over 10%, different one than a year ago, and we can't name any of them. It's the component supplier shape: revenue climbing while the buyer list narrows. We'll see how it tracks.