Actionable insights — NVIDIA Gone Parabolic

1. Use gross-margin direction as the demand/competition lie-detector

1. Take gross margin (revenue minus the direct cost of the product, as a %) and track its direction quarter over quarter, not just its level.
2. Stress-test it: list the things that should have pressured it this period (a product ramp, costlier inputs, a mix shift to lower-margin systems). If margin held anyway, pricing power and demand are intact.
3. Treat the first crack as the earliest tell: if competition were biting or demand softening, margin slips before unit growth or revenue does.

• Any sequential gross-margin slip at a market-leader despite "sold out" commentary — that's the leading edge of a demand/competition turn.

2. Strip equity-investment gains to find the real operating profit

1. Don't take the headline net-profit number at face value — read the income statement for "gains on equity investments" / "other income."
2. Subtract those mark-to-market gains (paper gains on stakes the company holds in other firms) to get profit from the actual business.
3. Check whether the gains are concentrated in a few holdings — a single name moving can swing the reported figure.

• Any quarter where "other income" / investment gains are a large share of net profit — normalize before comparing earnings growth.

3. Watch new segment disclosures to test whether a buildout is broadening

1. When a company re-cuts its reporting segments, ask what concern the new split is designed to answer — disclosure changes are rarely neutral.
2. Compare growth rates across the new buckets and their relative sizes to test the bear thesis directly.
3. Watch the faster-growing, less-concentrated bucket sequentially — if it keeps pace, the demand base is widening, not narrowing.

• ACIE growing faster than Hyperscale sequentially — confirmation the AI buildout is broadening beyond hyperscaler capex.

4. Treat banned/zeroed markets as a call option valued at zero

1. Identify any large market the company has deliberately modeled at zero (export bans, regulatory holds, geopolitics).
2. Confirm the base business is growing without it — that's what makes the zeroed market pure optionality rather than a hole in the forecast.
3. Value any restart as incremental upside on top of guidance, not as part of it — a "free" call option that only pays if policy turns.

• Beijing approval / first H200 shipments — the trigger that converts the zeroed China line into realized revenue.

5. Judge a growth story by cash conversion, not just the top line

1. Pair revenue growth with free-cash-flow growth — does the business actually convert growth into cash, or fund it with capex/working capital?
2. Sanity-check the balance sheet (cash vs debt) to gauge whether that cash buys strategic flexibility.
3. Be skeptical of fast revenue growth where FCF lags badly (the opposite case — see the prior post on Amazon's FCF collapsing into capex).

• A widening gap between revenue growth and FCF growth — the first sign a "growth" story is really a capex story.

6. Track where the bottleneck has moved — own the new scarce input

1. For any supply-constrained boom, ask what the binding constraint is now — it migrates as one bottleneck is relieved.
2. Watch the partnerships/contracts the leader signs: they reveal what it can't get enough of (here, the move from "GPU access" to power/land/cooling).
3. Note how the leader takes exposure — an outright stake vs a warrant struck above the current price tells you whether it's a conviction buy or a cheap upside kicker.

• New power / data-center / energy partnerships from chipmakers — signposts that capital is chasing the new scarce input.

7. Make the product-cadence handoff the key risk to monitor

1. For a company on a one-product-per-year cadence, map the handoff dates between the current product and the next.
2. A clean handoff sustains the growth outlook; a delay creates a "demand air pocket" — the gap bears wait for.
3. Tie the multi-year revenue outlook to that cadence so a slip is immediately legible as a downgrade.

• Any slip in Rubin production/ship dates, or supplier commentary implying a gap between Blackwell wind-down and Rubin ramp.