Nvidia vs AMD vs Broadcom: Which Stock Leads The AI Race?

As of mid-November 2025, the AI build-out runs through three tickers: Nvidia (NVDA), AMD (AMD) and Broadcom (AVGO).

All three are making billions from AI. They are not playing the same game.

• Nvidia sells general-purpose GPUs and a software ecosystem.

• AMD is selling cheaper, competitive GPUs and trying to break that ecosystem.

• Broadcom is selling the high-speed plumbing and custom chips that sit around both.

For context, spot prices are roughly 190 dollars for NVDA, 247 dollars for AMD and 342 dollars for AVGO on 15 November 2025.

The quick view: three different AI bets

The headline: Nvidia is still the AI “platform” stock. AMD is the price-performance challenger. Broadcom is the infrastructure enabler.

Revenue mix and business models

NVIDIA is the pure AI engine

• Around 80 percent of revenue now comes from data centre GPUs and AI infrastructure.

• CUDA, networking and software tools create a form of lock-in for enterprises, cloud providers and AI labs.

• Data centre gross margins above 70 percent are unusually high for hardware.

This makes Nvidia highly geared to AI capex. It also means results are sensitive to any pause in hyperscaler spending.

AMD: The aggressive second mover

• Data centre revenue is growing quickly, helped by the MI300 and MI350 Instinct lines.

• The company still earns a meaningful share of revenue from PC and gaming, so the AI exposure is thinner than Nvidia’s.

• The key swing factor is how much workload hyperscalers are willing to shift from Nvidia to AMD-based clusters.

AMD has narrowed the performance gap at the chip level. The gap in the software ecosystem and developer familiarity remains larger.

Broadcom: The infrastructure backbone

• Revenue is dominated by networking silicon, custom ASICs and enterprise software, including VMware.

• AI exposure comes from Ethernet switches, optical interconnects, and custom accelerators designed with cloud giants.

• Growth is slower than Nvidia and AMD, but the business is more diversified across end markets.

Broadcom is effectively selling shovels into the AI gold rush rather than competing directly to be the GPU standard.

AI moats: where the advantages really sit

NVIDIA has the software and ecosystem moat

Nvidia’s durable edge is not only in chips, but in CUDA and the surrounding ecosystem:

• Fifteen plus years of libraries, tools and frameworks built on CUDA.

• Most AI research labs, enterprises, and hyperscalers first train staff and models on Nvidia.

• Rewriting and reoptimising models away from CUDA involves real cost and time.

This history produces high switching costs. A rival needs more than a slightly better chip. It requires a whole ecosystem migration.

AMD has strong hardware, a weaker moat

On paper, AMD’s latest accelerators are very competitive:

• MI300 and MI350 series close much of the raw performance gap.

• Pricing is often 20 to 30 percent below comparable Nvidia systems, which improves the total cost of ownership.

• ROCm, AMD’s open-source software stack, has improved but still trails CUDA in maturity and adoption.

AMD tends to win in two scenarios:

1. Buyers want “good enough” performance at a lower cost.

2. Hyperscalers want explicit diversification away from Nvidia to gain bargaining power and improve supply security.

Broadcom: custom silicon and networking

Broadcom’s moat is concentrated in two areas:

• Custom AI ASICs and XPUs for large accounts such as Google, Meta and OpenAI. These are chips built for specific workloads, often more power efficient than general-purpose GPUs at scale.

• High-speed networking for AI clusters, including Tomahawk and Jericho switch families, plus co-packaged optics.

Broadcom is not trying to displace Nvidia’s general-purpose accelerators. It is positioning itself as the partner that helps hyperscalers build their own silicon and connect everything.

Numbers that matter: growth, margins and valuation

Growth

Only Nvidia currently combines very high growth with very large absolute revenue.

Margins

Margins illustrate the moat:

• Nvidia looks more like a platform company than a hardware seller.

• AMD looks more like a high-end component supplier.

• Broadcom sits in the middle, helped by software and design work wrapped around its silicon.

Valuation snapshot

Multiples move quickly, but the broad picture is consistent:

High-level interpretation:

• Nvidia is priced as if it can sustain a dominant AI position for many years.

• AMD looks cheaper on some growth-adjusted views, but results depend heavily on future share gains.

• Broadcom trades on the lowest cash-generation expectations, reflecting its more diversified profile.

Who really threatens Nvidia

Short term: AMD

In the next few years, AMD will be the most direct competitive pressure:

• It sells drop-in GPU alternatives for AI clusters.

• It can undercut Nvidia on price while still offering strong performance.

• It is winning visible hyperscaler and OpenAI-related deals that scale from 2026 onwards.

Every percentage point of share that moves from Nvidia GPUs to AMD GPUs has a clear impact on revenue and profits.

Long term: custom accelerators and networking vendors

Over a longer horizon, custom silicon is the more structural threat:

• Cloud providers increasingly design their own accelerators for specific workloads.

• These chips often rely on partners like Broadcom to handle design and manufacturing.

• The more spending that flows into bespoke chips, the less flows into merchant GPUs.

In that scenario, Broadcom benefits from design wins and networking demand, even if Nvidia’s unit share in GPUs stays high.

Traditional general-purpose chipmakers that are not tightly integrated into AI software stacks look less threatening in comparison.

Risk and reward profiles

This is a comparison of profiles, not a recommendation.

Nvidia

• Key sensitivity: dependence on hyperscaler AI capex and the pace of custom ASIC adoption.

• Upside case: AI compute demand continues to grow at a high rate, monetisation remains robust, and Nvidia defends most of its share, allowing very large absolute earnings over time.

• Downside case: AI spending normalises faster than expected, or more budget shifts to in-house silicon, compressing growth and margins from today’s levels.

AMD

• Key sensitivity: ability to close the ecosystem gap with CUDA and to convert design wins into durable, large-scale deployments.

• Upside case: AMD reaches double-digit share of AI accelerators, supports strong compound growth in data centres, and demonstrates sustainable margins.

• Downside case: workloads remain concentrated on Nvidia and custom chips, leaving AMD with a limited share of the highest-value clusters.

Broadcom

• Key sensitivity: concentration in a small group of hyperscaler customers and the timing of large custom programs.

• Upside case: AI networking and custom ASIC revenue steps up sharply as 10-gigawatt-class projects roll out, while the rest of the business remains stable.

• Downside case: project timing slips, or customers slow AI roll outs, creating lumpier growth than the current narrative implies.

What this means for NVDA’s valuation

Pulling the threads together:

• Nvidia currently combines the highest growth, the strongest margins and the most entrenched ecosystem in AI accelerators.

• AMD has established itself as a credible second source in GPUs and is gaining share, but still operates under Nvidia’s software shadow.

• Broadcom has carved out a different layer of the stack, focused on custom silicon and networking that benefits from AI spend regardless of which GPU wins.

None of this, on its own, undermines Nvidia’s premium valuation today. The scenarios that could do that are more macro and structural:

1. AI capex slows more sharply than current forecasts.

2. Custom accelerators designed with partners like Broadcom capture a much larger share of budgets than expected.

3. The economics of AI workloads shift in ways that reduce the willingness to pay for top-end general-purpose GPUs.

For now, Nvidia remains the purest expression of the AI compute trade, AMD provides high-beta exposure to share gains, and Broadcom offers diversified AI infrastructure exposure tied to the scale of the build-out rather than any single GPU winner.