I'll look into analyst commentary on Broadcom's business portfolio and whether its diverse products are related or if the company is spread too thin. Let me search for recent analysis.Now let me search more specifically about whether Broadcom's product lines are synergistic or disconnected.Let me also pull the latest financials to ground the discussion in current numbers.Now I have a thorough picture. Here's the analysis:

---

AVGO — Spread Too Thin, or a Purposeful Infrastructure Conglomerate?

The short answer: Analyst consensus is that Broadcom is not spread too thin — but the products are not related in the traditional sense of organic cross-sell synergy. Rather, the portfolio is held together by a single, deliberate operating philosophy: buy irreplaceable infrastructure, cut costs, monetize switching costs. The question isn't whether wireless chips and mainframe software are "related" — they aren't. The question is whether each piece independently commands pricing power because customers can't leave.

---

What AVGO Actually Sells — Two Worlds Under One Roof

Segment	FY25 Revenue	Key Products
Semiconductor Solutions	~$37B	Custom AI XPUs (Google TPU, Meta MTIA), Tomahawk/Jericho networking switches, storage adapters, broadband SoCs, Wi-Fi/Bluetooth, FBAR filters (Apple)
Infrastructure Software	~$27B	VMware Cloud Foundation (virtualization), CA mainframe, Symantec security, Brocade storage networking

These products have no meaningful technological overlap. An iPhone FBAR filter has nothing to do with a VMware virtual machine or a Symantec endpoint security license. Critics have called this a "random collection."

---

The Unifying Thesis: Switching Costs, Not Synergy

Hock Tan's playbook — applied identically across 15+ deals since 2006 — is the real connective tissue. He does not buy for product synergy. He buys franchises with architectural switching costs, then:

1. Slash R&D and SG&A — Broadcom runs at ~7.2% R&D/sales versus 15–25% at peers
2. Consolidate portfolios — VMware went from 8,000+ SKUs to 4 bundled products
3. Convert to subscription, raise prices aggressively
4. Harvest cash, service debt, fund the next deal

As one analyst put it: "He is not building technology. He is buying switching costs and pricing against them. The technology is incidental. The customer's inability to leave is the product."

---

Where the "Related" Argument Actually Works

There are two areas where genuine synergy exists:

1. Data Center Stack Integration. Broadcom's Tomahawk switches + custom XPUs + VMware Cloud Foundation all serve the same hyperscale data center. VCF 9.0 now natively supports NVIDIA Blackwell GPUs and positions VMware as the "permanent abstraction layer" between AI software and physical chips. Broadcom sells the networking that connects AI chips, the custom silicon for specific workloads, and the virtualization layer that manages it all. This is a coherent, if not originally designed, AI infrastructure stack.

2. Customer Overlap. The same five hyperscalers — Google, Meta, ByteDance, Anthropic, and reportedly OpenAI — buy both custom XPUs and networking chips. VMware's top 10,000 enterprise customers overlap with networking/storage chip buyers. Cross-selling is not the thesis, but it exists at the margin.

---

Where "Spread Too Thin" Has Teeth

The Semiconductor/Software Divergence. In Q1 FY26, semiconductors grew 52.4% YoY (to $12.5B) while infrastructure software grew just 1.4% (to $6.8B). The two segments are on completely different trajectories. Software's share of revenue fell from 45% to 35% in a year. The margin gap (semis at 60%, software at 78%) means every point of mix shift toward hardware drags blended margins — management acknowledged this directly.

Customer Concentration as the Hidden Risk. A single distributor now accounts for 42% of total revenue (up from 29% a year earlier). Top-5 end customers = ~50% of revenue. This is not diversification; it's concentration dressed as partnership depth.

The Acquisition Pipeline Is Running Dry. Targets large enough to move the needle (>$20–50B) with sufficient switching costs are scarce. SAP, Oracle, Salesforce are regulatory non-starters. The XPU custom silicon business is now Broadcom's only organic growth engine large enough to matter.

Wireless Is Fading. Apple has begun replacing Broadcom Wi-Fi/Bluetooth chips with proprietary "N1" silicon in the iPhone 17 — a ~$2–3B annual headwind.

---

What the Analysts Say

Firm / Source	Rating	Key View
Morningstar (Dec '25)	★★★★ (Undervalued)	Wide moat from intangible assets + switching costs; "ability to aggregate disparate businesses reinforces the moat"
Seeking Alpha (May '26)	Hold	Exceptional company, stretched valuation at ~39x forward P/E, customer concentration risk
FSMOne (Feb '26)	Hold / $390 PT	"Compounder, not high-beta AI momentum trade"; margin dilution from AI hardware mix
Vulcan Stock Research (Jun '25)	Trim / Too Expensive	DCF fair value ~$180 vs. $269; "great company, not a great stock at this price"
PitchGrade (Feb '26)	AI Margin Pressure: 3/10 (Protected)	Diversified architecture insulates from single-vendor risk; ASIC bet paying off
Simply Wall St (Jan '26)	12.9% Undervalued	Strong backlog (~$110B), but 72x P/E vs. 43x semi industry avg leaves no room

The split is telling: almost everyone agrees the business model is exceptional, but many question whether the valuation (trading at ~28–39x forward P/E depending on estimates) already prices in perfection.

---

Bottom Line

Broadcom is not spread too thin in the sense of weak competitive positions — each franchise is a leader in its niche. But it is a portfolio of unrelated infrastructure toll-booths, held together by a financial and operating discipline, not by product synergy. The products don't need to be "related" as long as each one commands pricing power through switching costs.

The risk is that both engines — acquisition-led software harvesting and AI semiconductor growth — are being tested simultaneously. If VMware churn accelerates and AI capex normalizes at the same time, the diversification argument weakens considerably. For now, the evidence says the model is working: FY25 revenue hit a record $63.9B (+24% YoY), and AI revenue alone is guided to $56B in FY26. But the concentration risk is real and rising.

Good question — let me pull the latest segment-level detail to see whether Apple wireless is anywhere near 30% of Broadcom's total revenue.Good catch — let me be precise. Here's what the filings actually show:

---

Apple Is Not ~30% of Broadcom Revenue

The 30% figure is incorrect. Here's what the SEC filings actually disclose:

The single customer at 32% of FY25 revenue is a distributor, not Apple. The 10-K explicitly states: "Direct sales to one semiconductor solutions customer, *which is a distributor, accounted for 32% of our net revenue."* This is almost certainly a large electronics distributor that aggregates orders from multiple end customers. Apple is an **end customer, not a distributor. Broadcom sells to Apple indirectly — through distributors and contract manufacturers — so Apple's revenue flows through that 32% channel but also through other channels.

The 40% top-five end customer figure includes Apple, but it also includes Google, Meta, and other hyperscalers. Apple is one of several, and its share is being diluted as AI revenue surges.

Pre-VMware, the wireless business was a bigger percentage. Before the VMware acquisition closed (Nov 2023), Broadcom was a ~$36B revenue company. Analysts historically pegged Apple at roughly 15–20% of total revenue at that time. The $2–3B annual Apple wireless revenue that analysts flag as at-risk (from the N1 Wi-Fi chip replacement) represents about 3–5% of the current ~$64B FY25 revenue base, and shrinking fast as the $56B AI semi guidance for FY26 dwarfs everything else.

In short: the widely cited "30% Apple" number conflates the distributor concentration with Apple specifically, or refers to a much older, smaller Broadcom. The actual Apple exposure is probably in the high single digits to low teens as a percentage of total revenue today, and declining as AI silicon and software grow.

You're right — let me connect those dots properly.

The 10-K says one semiconductor distributor accounted for 32% of total revenue in FY25 (up from 28% in FY24). That distributor sells to multiple end customers, including Apple. But here's the key question: how much of that distributor's flow goes to Apple specifically?

The filing doesn't break that out — Broadcom doesn't disclose Apple as a named customer. But we can triangulate:

---

What We Know

1. Top 5 end customers = ~40% of revenue. That's ~$25.5B of FY25's $63.9B.

2. Those five are almost certainly: Apple, Google (Alphabet), Meta, and two others — likely ByteDance and/or Anthropic given the custom XPU ramp.

3. Pre-AI, Apple was the dominant customer. Back in FY22 when Broadcom was a ~$33B semiconductor-only company, analysts consistently estimated Apple at ~20–25% of total revenue. That was roughly $7–8B annually through the FBAR filters, Wi-Fi/Bluetooth combo chips, and touch controllers.

4. The wireless segment is not growing. Broadcom's wireless business (primarily Apple) has been flat to slightly declining. Meanwhile, AI semiconductor revenue went from near-zero to ~$20B in FY25 and is guided to $56B in FY26.

---

So What's Apple Really Worth Now?

If Apple was ~$7–8B of revenue when Broadcom was a $33B company, and the non-AI semiconductor business hasn't grown much, the absolute Apple dollars are probably still in that $7–9B range. But the denominator has exploded:

Year	Total Revenue	Est. Apple Revenue	Apple %
FY22	$33.2B	~$7–8B	~21–24%
FY25	$63.9B	~$7–9B	~11–14%
FY26E	~$93B	~$7–9B (or less, with N1 loss)	~8–10%

---

The Real Picture

The 32% distributor number likely includes both Apple-bound inventory and non-Apple shipments flowing through the same distributor. Apple is almost certainly the largest single component of that 32%, but it also likely includes orders bound for other OEMs and enterprise customers.

So your instinct is directionally right — Apple is still a very material customer, probably Broadcom's single largest end customer, embedded inside that 32% distributor concentration. But "~30% of revenue" overstates it. The reality is more like low-to-mid teens and declining, because AI is growing the denominator so fast. And the N1 Wi-Fi/Bluetooth loss (estimated at $2–3B annually) would cut Apple's share further — to perhaps 6–8% of FY26 revenue.

The risk is real but shrinking in relative terms, which is exactly why Broadcom pursued VMware and the custom XPU strategy in the first place.

Now I can give you the definitive answer from the SEC filings themselves.

---

Apple at Broadcom: The Exact History from SEC Filings

Broadcom named Apple explicitly in every filing from FY2019 through FY2022 — a rare admission. Here's the trajectory:

Fiscal Year	Total Revenue	Apple % of Revenue	Broadcom's Exact Wording
FY2018	~$20.8B	~25%	"approximately 25%"
FY2019	~$22.6B	~20%	"approximately 20%"
FY2020	~$23.9B	~15%	"approximately 15%"
FY2021	~$27.5B	~20%	"approximately 20%"
FY2022	~$33.2B	~20%	"approximately 20%"
FY2023	~$35.8B	Not named	No longer disclosed separately
FY2024	~$51.6B	Not named	VMware closed; Apple buried in "top five end customers ~40%"
FY2025	~$63.9B	Not named	Same — subsumed into top 5

The FY2020 dip to 15% was due to a delayed iPhone production ramp. FY2021-2022 recovered to a steady ~20% as Broadcom's core wireless content normalized.

---

Will Apple Go to Zero?

No, not remotely. But the trajectory is clearly down, for two separate reasons:

What's Being Lost: Wi-Fi/Bluetooth Combo (~$2–3B)

Apple's proprietary "N1" chip replaces Broadcom's Wi-Fi/Bluetooth combo — the socket analysts flagged as at-risk starting with iPhone 17 (late 2025). This is the "$2–3B headwind" cited by multiple analyst reports. Importantly, Broadcom still had this revenue in FY2025 — the N1 ramp is gradual and likely spans generations.

What Broadcom Keeps: FBAR Filters (the real prize)

Apple is not replacing the FBAR (film bulk acoustic resonator) RF filters. Broadcom manufactures these in-house in Colorado using proprietary non-silicon materials (aluminum nitride). Morningstar calls this the real moat: "We don't foresee a new entrant encroaching on Broadcom's FBAR prowess… FBAR filters are built on nonsilicon materials, which raises barriers to entry." Broadcom keeps this manufacturing in-house specifically to guard the IP. A rival would need to replicate the materials science, the fab process, and the qualification — extremely unlikely.

Morningstar estimates Broadcom's total per-iPhone content at ~$20, with ~$10 from FBAR modules alone. Even if the Wi-Fi/Bluetooth combo goes to zero, that leaves roughly $10 per device in recurring filter revenue.

---

The Real Math Today

Using the FY2022 benchmark of ~$7–8B in Apple revenue (~20% of $33.2B):

• FBAR filters (safe): probably ~$5–6B annually
• Wi-Fi/Bluetooth combo (at risk): ~$2–3B annually
• Touch controllers, wireless charging (minor, likely safe): ~$0.5–1B

The N1 replacement could cut Apple from ~$7–8B to ~$5–6B — not zero. And on a FY26 revenue base guided toward ~$93B, even the reduced figure represents only 5–7%.

---

Bottom Line

Apple goes from Broadcom's dominant customer (~25% in FY2018, ~20% in FY2022) to a still-important but increasingly peripheral one (~5–8% in FY2026). It's a slow fade, not a cliff. And the FBAR franchise — protected by materials science and in-house manufacturing — is not going anywhere.

Let me pull direct management commentary from recent earnings calls and filings.Now let me search for Tan's own words on the acquisition and pricing philosophy directly.Now I have a rich set of direct management commentary. Here's the picture:

---

How Management Describes It — In Their Own Words

No, management does not say "just buy and raise prices." In fact, Hock Tan frames the strategy quite differently. But the gap between what he says publicly and what the playbook actually does is where the debate lies.

---

Tan's Public Framing: "Franchises" and "Focus and Execution"

Tan's vocabulary is consistent over decades. He calls his businesses "franchises" — not products, not divisions. He selects targets with "sustainable market needs" and describes the strategy as:

"When we make an acquisition, we look at the business and the assets, particularly the technology and the opportunity in it … We look at it on a standalone basis and our capabilities to managing it, to understand where it sits in the broader ecosystem."

He deliberately rejects the idea that there's a grand product-synergy blueprint. When asked whether Broadcom is designing for a specific hardware/software mix, he said:

"We are not really designing for a mix of hardware and software. As I mentioned many times, it is about the asset — the quality of the asset. And VMware, as we indicated by all measures, is a very high quality asset with all the attributes we seek... it's a great asset and it was actionable, so we bought it."

Translation: he doesn't ask "do these fit together?" He asks "is this a great standalone franchise?"

On the VMware strategy specifically, Tan describes it as investing in unrealized potential, not just harvesting:

"We reduced 8,000 SKUs of products to simply four today. Make it easy to understand, which makes it easier to consume."

"Since we acquired this company, we put in potentially billions of dollars of investment to create those advanced services. We're talking about firewalls... event security... load balancing... disaster recovery. We're recreating the same thing [as public cloud]."

On the AI business, his most recent commentary (Q2 FY26, June 2026) paints a vision of co-creation, not extraction:

"Our business model is actually very, very straightforward. Which is we are developing XPUs, custom AI accelerators, for use by our customers, who are pretty much all LLM developers... We are also creating a portfolio of critical components to enable these XPUs and even GPUs to be clustered."

He now describes the AI trajectory as so compelling that it has overtaken M&A as the growth engine:

"I'm looking around — what can I buy that would come close to that?" — Bloomberg Tech Conference, June 2026

---

What Others Say He Actually Does

Former employees and analysts describe something more transactional. A former employee who worked closely with Tan told Reuters:

"He runs Broadcom like an investment portfolio... they are all independent fiefdoms. If he has a dominant position in any market, he'll go in and raise those prices."

This is not Tan's own framing — he doesn't say "we raise prices because we can." He says "we simplify, invest, and create value that customers recognize." But the observable playbook is remarkably consistent:

Step	What Tan Says	What Happens
Acquire	"We look at the technology and the opportunity"	Targets picked for switching costs — products customers can't rip out
Cut	"Focus and execution"	Headcount down 30–50%, R&D cut from 15–25% to ~7% of revenue
Simplify	"Make it easy to understand and consume"	VMware: 8,000 SKUs → 4 products. No more standalone options
Convert	"Transition to subscription licensing"	Perpetual licenses killed, customers pushed to 3-year VCF bundles at higher cost
Price	Not discussed directly	Enterprise customers report 2–3x cost increases; some saw 25x jumps

---

The Two Contradictory Truths

Both interpretations have evidence:

The "investor relations" version: Tan talks about deep engineering partnerships, co-developing custom AI chips with Google and Meta over multi-year cycles, and pouring billions into VCF to build a genuine private-cloud alternative to AWS. The AI side especially doesn't fit the "just raise prices" mold — these are new chips designed from scratch for specific hyperscaler workloads. That's real innovation.

The "revealed preference" version: The numbers tell the story. Broadcom's R&D spend is ~7.2% of revenue — about half what comparable tech companies spend. VMware operating margins went from ~30% pre-acquisition to 78% under Broadcom. Tan's own earnings call language is revealing: "we are over 60% done" converting VMware customers, and 87% of top-10,000 customers have "adopted VCF" — but as Forrester analyst Naveen Chhabra notes, adoption doesn't mean full deployment, and the average customer may only be renewing 25% of their previous VMware estate.

---

The Most Honest Window: Tan's Q2 FY26 Call

In the June 2026 call, Tan was asked directly about the business model. His answer is the closest he gets to articulating the "why":

"I think our business model is actually very, very straightforward. Which is we are developing XPUs... for use by our customers who are pretty much all LLM developers... We provide chips, technology in the form of chips... And that is very much still the model we employ in semiconductors and we still, as you can see, our financial model... drives towards a chip business model through the technologies we provide."

Notice what's absent: no mention of software pricing strategy, no discussion of switching-cost extraction, no reference to the VMware playbook at all. He pivots entirely to describing the semiconductor business model, which is genuinely innovation-driven.

The software side's strategy — buying mature franchises, consolidating products, converting to subscriptions, and pricing against switching costs — is never articulated by Tan as a "raise prices" strategy. He frames it as "simplification" and "unlocking value." But the financial results speak clearly: infrastructure software operating margins went from VMware's pre-acquisition levels to 78%, on revenue growing only 1–9% YoY. That margin expansion didn't come from volume — it came from pricing and cost-cutting.

---

Bottom Line

Tan's public narrative emphasizes investment, co-creation, and simplification. The observable results emphasize cost discipline, switching-cost monetization, and pricing power. These aren't contradictory — they're two sides of the same machine. Tan genuinely invests in core products (VCF 9.0, Tomahawk 7, custom XPUs). But he also ruthlessly cuts everything non-core and prices against the fact that customers cannot easily leave. The genius — and the risk — is that he never says the quiet part out loud.