Broadcom (AVGO): The AI Chip Giant Quietly Outpacing Nvidia

The Question Nobody's Asking About Nvidia

Nvidia (NVDA) has delivered a 1,300% return over the past five years. Its market cap sits at $4.45 trillion, with a 55% profit margin, 70% revenue growth, and 73% earnings growth. By any measure, these are extraordinary numbers.

But here's the mathematical reality check: for Nvidia to deliver another 1,300% from here, its market cap would need to reach $62 trillion — roughly twice the size of the entire US GDP. Is it possible? Technically. Is it the most compelling risk-adjusted opportunity in AI semiconductors right now?

In my analysis, the answer is no. And the company I keep coming back to is Broadcom (AVGO).

---

Core Analysis: The Numbers Broadcom Is Quietly Building

A 10x Story That's Still Unfolding

Broadcom has grown from a $163 billion market cap in 2020 to $1.58 trillion today — approximately a 10x return, or 900% gain. But what makes this more interesting than the raw number is the recent trajectory.

Over the past 12 months:

• Broadcom (AVGO): +73%
• Nvidia (NVDA): +56%

Broadcom outperformed Nvidia over the last year. That's not a fluke — it's the result of a structural shift in how AI infrastructure is being built.

The Custom Chip Advantage

Most investors think of the AI chip market as Nvidia's GPU monopoly. But AI infrastructure actually runs on two parallel tracks:

1. General-purpose GPUs (Nvidia's H100/H200/Blackwell) — flexible, off-the-shelf compute for diverse workloads
2. Custom AI accelerators (ASICs/TPUs) — purpose-built chips optimized for specific use cases at massive scale

Broadcom dominates the second category. Google's TPU chips are designed and manufactured with Broadcom's help. Broadcom has also partnered with Meta on custom AI silicon. And critically, OpenAI and Anthropic are now commissioning Broadcom to build custom chips for their next-generation AI systems.

Why does this matter? Because the hyperscalers — the companies spending $50-100 billion annually on AI infrastructure — are actively reducing their Nvidia dependency. They want chips optimized for their exact workloads. Broadcom is the go-to partner for that.

The Infrastructure Scale Trajectory

Here's the data point I find most compelling when thinking about Broadcom's long-term opportunity:

• 2022: ~4,000 accelerators per AI cluster
• 2024: ~30,000 accelerators per cluster
• 2027 (projected): up to 1,000,000 accelerators per cluster

That's a 250x increase in cluster scale within five years. If these projections materialize even partially, the demand for AI chips — including Broadcom's custom accelerators — will dwarf anything we've seen so far.

CEO Hock Tan has set a specific target reflecting this: $100 billion in AI chip revenue. Broadcom generated roughly $20 billion in AI chip revenue last year. Reaching $100B means 5x growth from current levels.

Full-Stack Positioning

Broadcom's competitive moat goes beyond chip design. The company covers the entire AI infrastructure stack:

• Custom AI accelerators: ASIC design and development for hyperscalers
• Networking: Ethernet solutions connecting massive GPU/ASIC clusters
• Connectivity: High-speed data transfer optimization
• Storage: Data management solutions for AI workloads
• Enterprise software: VMware acquisition (2023) added a critical software layer

The 2023 VMware acquisition was particularly strategic. It transformed Broadcom from a pure-play semiconductor company into an end-to-end AI infrastructure provider, combining hardware and software capabilities that few competitors can match.

Financial Comparison: The Profitability Lens

Here's how the major AI semiconductor players stack up on the metrics I care most about:

Metric	Broadcom (AVGO)	Nvidia (NVDA)	AMD	SMCI
Revenue Growth (this year)	64%	70%	—	88%
EPS Growth (this year)	64%	73%	—	8.7%
Profit Margin	40%+	55%	12%	3%
Market Cap	$1.58T	$4.45T	—	—

The number that stands out most is SMCI's EPS growth of just 8.7% despite 88% revenue growth. When a company is growing revenue by 88% but only growing earnings by 8.7%, the economics are broken — the growth isn't converting to shareholder value.

Broadcom tells the opposite story: 64% revenue growth matching 64% EPS growth. This is what profitable, scalable growth looks like.

Broadcom's specific projections for this year and next:

• Revenue this year: ~$104B (64% growth)
• EPS this year: $11.18/share (64% growth)
• Revenue growth next year: 44%
• EPS next year: $17.20/share (54% growth)

---

Implications: What the Valuation Math Tells Us

The P/S Scenario

Applying the sector's average Price-to-Sales multiple of 22x to Broadcom's projected revenue:

• Next year projected revenue: ~$151B
• 22x P/S = $3.3 trillion market cap
• Current market cap: $1.58 trillion

That math suggests roughly 2x upside from current levels if Broadcom trades at its historical sector average. This isn't a guaranteed outcome, but given the growth rate and profitability profile, the valuation isn't unreasonable.

The Profitability Filter

In my stock analysis framework, I apply a two-step filter:

Step 1: Is the company growing at a high rate? For Broadcom: yes, at 64% revenue growth.

Step 2: Is that growth converting to earnings? For Broadcom: yes, with 64% EPS growth and 40%+ margins.

A company that passes both filters — high growth AND high profitability — deserves a premium valuation. Broadcom qualifies. SMCI, despite higher revenue growth, fails Step 2. AMD, while solid, falls well short on margin (12%).

The profitability gap matters more than most investors recognize. At $11.18/share in earnings growing at 54% to $17.20 next year, Broadcom's earnings power compounds rapidly in a way that SMCI's thin-margin business simply cannot match.

---

Risks and Counterarguments

A fair analysis requires engaging with the bear case.

1. Nvidia's ecosystem advantage remains formidable

Nvidia's CUDA ecosystem — the software layer that makes GPU programming accessible — represents a decade of developer investment and network effects. This moat doesn't evaporate quickly. For AI research, startups, and diverse workloads, Nvidia's general-purpose GPUs will remain the default choice. Broadcom's custom chips are optimized for hyperscale workloads; they're not a drop-in replacement for the broader market.

2. The valuation already prices in a lot

Broadcom's stock isn't cheap by any traditional metric. A significant portion of the $100B AI chip revenue target and the multi-year growth story is already reflected in the share price. Any miss on guidance — or any slowdown in hyperscaler AI spending — could cause meaningful multiple compression.

3. Competition is intensifying in custom silicon

Marvell Technology (MRVL) and Intel (INTC) are actively pursuing the custom AI chip market. ARM-based designs are proliferating. While Broadcom has deep relationships and years of expertise, the market it's targeting is attractive enough to draw serious competition over the next three to five years.

4. Customer concentration risk

Google represents a significant portion of Broadcom's AI chip business through the TPU relationship. If Google were to internalize more of its chip design capabilities or diversify to other foundry partners, the revenue impact on Broadcom could be material.

---

FAQ

Q1. Should I own both Nvidia and Broadcom, or pick one?

They occupy meaningfully different positions in the AI infrastructure stack. Nvidia dominates general-purpose GPU compute; Broadcom leads in custom ASICs and end-to-end infrastructure. Owning both provides diversified exposure to AI hardware spending. If you're looking for the better incremental opportunity based on valuation-to-growth ratio, Broadcom's smaller market cap relative to its growth trajectory makes it the more asymmetric bet today.

Q2. Is Hock Tan's $100B AI chip target realistic?

Going from $20B to $100B requires 5x growth. Aggressive, but not implausible given the AI cluster scale trajectory (4,000 accelerators per cluster in 2022 to potentially 1 million by 2027). The OpenAI and Anthropic partnerships could meaningfully accelerate the timeline. I'd treat it as a directional signal rather than a precise forecast — even reaching $60-70B would represent a step-change in Broadcom's AI revenue profile.

Q3. Is it too late to buy Broadcom after its 10x run?

The "too late" feeling after a 10x gain is psychologically powerful but often analytically misleading. The relevant question isn't where the stock has been — it's what the forward earnings power looks like relative to the current price. With $17.20/share in projected 2027 earnings and a growth rate still in the 40-54% range, there's a credible path to significant further appreciation. That said, position sizing matters: at a $1.58T market cap, this isn't a small-cap moonshot. Size accordingly.**