Market & Macro
Nvidia vs Broadcom vs Qualcomm: The AI Revenue Numbers Aren't Comparable
Nvidia reports $75.2B of quarterly Data Center revenue, Broadcom $10.8B of quarterly AI semiconductor revenue, while Qualcomm's >$15B data-center figure is a fiscal-2029 target.
The direct answer: these three AI-infrastructure numbers should not be ranked on one chart. Nvidia's $75.2B and Broadcom's $10.8B are realized revenue for one quarter, but they cover different product boundaries. Qualcomm's more-than-$15B figure is an annual FY2029 company goal, not current reported revenue.
Use the five-gate AI revenue field guide to apply the same period, status, boundary, reconciliation, and next-evidence test to any earnings release.
Thesis
The useful comparison is not headline size. It is evidence maturity: whether revenue has been realized, whether the reporting period is fixed, and whether the product boundary is visible. On that axis, Nvidia provides the strongest current revenue evidence, Broadcom provides large but partially aggregated evidence, and Qualcomm provides a future conversion target.
Source-derived visual: NVIDIA Q1 FY2027 results, Broadcom Q2 FY2026 results, and Qualcomm's 2026 Investor Day release. “Maturity” describes disclosure status, not investment quality or expected return.
This ordering does not say Nvidia is automatically the best investment or Qualcomm the worst. It says the burden of proof is different. Nvidia's discussion begins with a reported business; Broadcom's with a reported business whose internal mix is hidden; Qualcomm's with whether a target becomes a reported business at all.
Source Evidence Snapshot
| Company | Headline figure | Period | Evidence status | Critical boundary |
|---|---|---|---|---|
| Nvidia | $75.2B Data Center | Q1 FY2027 quarter | Realized revenue | Compute $60.4B; networking $14.8B under the previous sub-market lens |
| Broadcom | $10.8B AI semiconductors | Q2 FY2026 quarter | Realized revenue | Custom accelerator versus AI networking split not disclosed |
| Qualcomm | More than $15B data center | FY2029 annual | Management target | Not current reported revenue; target economics remain unobserved |
Nvidia: broad reported revenue, with a changing lens
Nvidia reported $75.2B of Q1 FY2027 Data Center revenue, up 92% year over year. Its previous sub-market lens divided that quarter into $60.4B of compute and $14.8B of networking. That is the most complete realized-revenue proof in this comparison, even though the category includes more than a single type of AI chip.
2026-05-20, captured 2026-07-14. The company says it is moving from the compute/networking sub-market view to Hyperscale and ACIE customer-market categories.The catch is continuity. The new Hyperscale and ACIE lens helps readers see who buys the infrastructure but does not preserve the same “what was sold” series. The dedicated Nvidia evidence review shows why networking attachment matters and why a future product bridge would improve the thesis.
Broadcom: reported AI scale, hidden product mix
Broadcom reported $10.8B of Q2 FY2026 AI semiconductor revenue, up 143% year over year, within $22.187B of total company revenue. The AI figure is therefore about 48.7% of total revenue. It is realized and financially material, but the release does not divide the number between custom accelerators and AI networking.
2026-06-03, captured 2026-07-14. Broadcom identifies custom AI accelerators and AI networking as drivers but reports one combined AI-semiconductor figure.Broadcom also said it expected Q3 FY2026 AI semiconductor revenue of $16.0B. That is company guidance, not yet a realized result. The dedicated Broadcom evidence review separates the reported $10.8B from the still-hidden product and margin mix.
Qualcomm: a large destination, not current revenue
Qualcomm's 2026 Investor Day release targets more than $15B of annual data-center revenue in FY2029 and $40B of annual QCT non-handset revenue in the same year. The numbers define management's destination. They do not establish current data-center sales, the quarterly ramp, customer concentration, or margins.
Source-derived visual: Qualcomm 2026 Investor Day release. The FY2029 figures are forward-looking company targets, not editorial forecasts or current reported revenue.
That distinction is easy to lose when all three numbers appear beside the same AI label. The Qualcomm evidence review treats the target as a conversion test: future filings must turn the destination into dated revenue and economics.
What the Street Is Pricing
This comparison does not use a current share-price multiple, private consensus data, or price targets. It therefore does not claim to reverse-engineer an exact market probability. The observable underwriting burden is narrower and more defensible:
Explanation visual: Nex checks the denominator before comparing the headline. The character is not evidence; the official figures and boundaries are listed in the adjacent table.
| Company | What investors can underwrite now | What remains an assumption |
|---|---|---|
| Nvidia | A very large, realized Data Center business with a recent product split | Product-mix continuity, margins by component, and durability of networking attachment |
| Broadcom | A large, fast-growing realized AI-semiconductor business | Accelerator/networking mix, AI-specific profitability, and how much guidance becomes revenue |
| Qualcomm | A dated strategic target and addressable product plan | Revenue conversion, timing, margins, customers, and competitive position at scale |
Only after this evidence reconciliation should a reader compare growth, margins, customer concentration, capital intensity, and valuation. A cleaner disclosure is not automatically a better security. A higher target is not automatically a larger present business.
Risks to the Thesis
The strongest countercase is that this framework can over-reward disclosure neatness. Nvidia, Broadcom, and Qualcomm participate at different layers of the stack, with different accounting boundaries and business models. A compact category can still be economically superior to a broad one, and a future entrant can create value before its revenue becomes material.
| Risk | Why it matters | Guardrail |
|---|---|---|
| Scope mismatch | Data Center, AI semiconductors, and data-center targets are not identical markets | Never total the three figures or call them market share |
| Period mismatch | Two quarterly results sit beside one annual FY2029 target | Preserve the stated period in every chart and caption |
| Disclosure change | Nvidia's product lens is changing; Broadcom's mix remains combined | Require a comparable bridge before claiming a trend |
| Target conversion | Qualcomm's destination may arrive later, smaller, or with weaker economics | Use reported revenue and margins when they appear |
| Missing valuation | Evidence quality alone does not determine expected return | Add market price and valuation only in a separately sourced analysis |
The framework also omits product-level margins because the cited releases do not provide a comparable set. It does not normalize customer concentration or capital requirements. Those omissions limit the conclusion deliberately: this is a map of what is observable, not a complete security ranking.
What Flips the Call
The evidence changes when later filings close the current gaps. No arbitrary revenue threshold is needed. The decisive events are category continuity, realized results, and disclosed economics.
Source-derived visual: the same three official issuer releases. The gates are an editorial monitoring framework. Only Broadcom's stated $16.0B Q3 FY2026 AI-semiconductor figure is company guidance.
| Company | Evidence that strengthens the comparison | Evidence that weakens it |
|---|---|---|
| Nvidia | Future results preserve a product bridge or another reproducible way to track networking attachment | Total Data Center growth remains visible while product attachment becomes impossible to reconcile |
| Broadcom | Reported Q3 AI revenue can be reconciled with the $16.0B guide and product or margin detail improves | The combined number grows but accelerator, networking, and profitability remain inseparable |
| Qualcomm | Filings begin reporting data-center revenue, timing, customers, and economics against the FY2029 target | The target shifts without a dated bridge to realized revenue |
The conclusion flips only when the evidence stage changes. Until then, Nvidia is the realized broad-platform case, Broadcom the realized but aggregated AI-semiconductor case, and Qualcomm the target-conversion case. Those are three different research questions, not one leaderboard.
Methodology, Sources & Disclosure
This article compares only official issuer disclosures and preserves each company's stated scope and period. It does not annualize quarterly figures, sum unlike categories, estimate undisclosed product splits, or treat a target as revenue. Broadcom's 48.7% share is the reproducible calculation $10.8B / $22.187B = 48.68%; the numerator is rounded. Facts were rechecked as of 2026-07-14.
- NVIDIA Q1 FY2027 financial results
- Broadcom Q2 FY2026 financial results
- Qualcomm 2026 Investor Day data-center strategy
AI assisted with structure and consistency checks. Official sources, figures, captions, EN/KO parity, and final wording require human review before production deployment. No sponsorship or affiliate relationship with the companies is disclosed. This is general information, not individualized investment advice; it does not issue a rating or share-price objective.