Editorial review ·
AI demand still has to become profitable delivery.
A return to two published investment cases: the VRT correction recorded September 15 and the NVDA evidence reviewed August 27. This issue revisits what those records establish and the next results to watch.
What changed · VRT
Our earlier thesis said:
Power and thermal capacity is the binding constraint on data-center buildout, which makes the infrastructure supplier less cyclical than the compute it cools -- backlog converts even if chip demand pauses. The thesis fails if backlog turns out to track orders rather than lead them.
The revised thesis says:
Vertiv's power and cooling business can benefit from AI data-center constraints, provided orders become revenue at healthy margins. The next test is whether third-quarter sales and adjusted operating margin meet the company's July guidance; material cancellations or project delays would weaken the thesis.
The earlier wording overstated protection from a spending slowdown. Vertiv discloses that customers can defer or reduce orders, so backlog is not assured revenue. The revised thesis keeps the power-and-cooling opportunity but tests delivery against stated sales and margin guidance. This correction follows a review of existing disclosures; it does not signal a newly observed downturn.
This is TickerTrac's published reasoning. The explanation and source selection were recorded later. They do not establish what caused the change or describe an investor's past belief.
Still unresolved
For VRT, the open question is whether orders become sales at the margins the company expects. Its annual report says customers can reduce or defer some firm orders; expected backlog revenue may therefore go unrealized.
The published NVDA assessment was strengthened by Q2 results: revenue exceeded the low end of guidance, GAAP gross margin stayed within its guided range, and the Q3 outlook cleared the forward checkpoint. Supply availability, product mix, transition costs and export restrictions remain risks to delivery. Its outlook assumes no Data Center compute revenue from China.
Across companies · compare the next tests
The two cases ask different questions within AI infrastructure. NVIDIA’s test is revenue growth with gross-margin resilience. Vertiv’s revised test is converting power-and-cooling orders into revenue at healthy operating margins.
Read each result against that company’s own guidance. NVIDIA’s GAAP gross margin and Vertiv’s adjusted operating margin measure different things; their percentages are not a ranking of which business has the stronger thesis.
Next checkpoints
VRT · Third-quarter 2026 results: sales against $3.65–$3.85 billion guidance and adjusted operating margin against 24.0–25.0%. Disclosed cancellations or deferrals that materially reduce expected backlog revenue would weaken the thesis.
NVDA · Third-quarter fiscal 2027 results: revenue against $108.0 billion, plus or minus 2%, and GAAP gross margin against 74.0%, plus or minus 0.5 percentage points. Guidance remains a target; the results are still to be tested.
General market commentary, not investment advice.