Broadcom Booked $10.8B in AI Chips Last Quarter — and It's Guiding to $16B Next

If "AI chips" only makes you think of Nvidia, look at Broadcom's numbers

Here's the deal: on June 3, 2026, Broadcom (AVGO) reported its fiscal Q2 results, and the headline is heavy — $10.8 billion in AI semiconductor revenue alone, up 143% year-over-year. Total quarterly revenue rose 48% to a record $22.2 billion. If you had the impression that the AI boom flows only to Nvidia, Broadcom is the company showing that picture is only half the story.

The guidance is even punchier. Broadcom said next quarter (Q3) AI semiconductor revenue would grow over 200% YoY to $16 billion. It expects total Q3 revenue up 84% to $29.4 billion, with non-GAAP operating margin holding around 67%. The very fact that AI revenue is projected to jump from $10.8B to $16B in a single quarter is proof this company sits at the center of the AI cycle.

Then the long view. Broadcom sees full-year FY2026 AI semiconductor revenue at roughly $56 billion (+180% YoY) and reiterated its target of $100 billion+ in AI revenue by FY2027. Earning $100B a year from AI chips alone is a bet that this is a structural super-cycle, not a one- or two-quarter flash. The core engine is twofold — custom AI accelerators (ASICs) and AI networking.

The players — Broadcom, ASICs, and the hyperscalers

The first protagonist is Broadcom itself. Unfamiliar to consumers, but a giant in data-center, networking, and communications chips. Its stature shifted in the AI era. Where Nvidia sells "general-purpose AI GPUs," Broadcom is the company that helps Big Tech design its own AI chips and supplies the critical components. It's laying the skeleton of AI infrastructure out of sight.

The second protagonist is the concept of custom ASICs (application-specific chips). Google's TPU and the in-house AI chips from hyperscalers like Meta, Amazon, and Microsoft are all ASICs. They're optimized for specific tasks, so at sufficient scale they're cheaper and more efficient than general GPUs. Big Tech keeps investing in its own chips to reduce Nvidia dependence — and Broadcom backs much of that design and production. As the AI chip market grows, so does Broadcom, the company "building Nvidia's alternative."

The third protagonist is the hyperscalers — Google, Meta, Amazon, Microsoft. Their astronomical capex on AI data centers is the source of Broadcom's revenue. Buying lots of AI chips isn't the end of it; you also need enormous "AI networking" gear to connect those chips. Broadcom holds both chokepoints — ASICs and networking — and collects tolls on both.

The substance — Q2 and guidance by the numbers

Item	Detail
Reported	June 3, 2026 (FY2026 Q2)
Q2 AI semi revenue	$10.8B (+143% YoY)
Q2 total revenue	$22.2B (+48% YoY)
Q3 AI revenue guidance	$16B (+200% YoY)
Q3 total revenue guidance	$29.4B (+84% YoY)
FY2026 AI revenue outlook	~$56B (+180%)
FY2027 AI target	$100B+
Non-GAAP operating margin	~67%

The thing to note in this table is the accelerating AI-revenue curve. The growth rate itself is steepening, 143% → 200%. Usually growth slows as revenue scale grows, but Broadcom's rate is climbing — a sign hyperscaler AI-chip orders are exploding. The 67% operating margin is striking too: this isn't just selling a lot, it's a high-margin, high-value business.

There was a shadow in the same report, though. The software segment came in slightly below market expectations, and the stock wobbled briefly right after. Broadcom is also a major infrastructure-software player since the VMware acquisition, and that side didn't grow as hot as AI semis. So "AI chips are exploding, but the whole company isn't all AI" is worth noting.

Still, the big picture is clear. The company reiterating its $100B FY2027 target signals that management sees this demand as a multi-year structural trend, not a transient bubble. As long as hyperscaler capex plans hold, Broadcom's AI revenue curve likely keeps pointing up. Of course, the "what if capex rolls over?" question always tags along.

What's in it for whom

Broadcom is cementing its role as the "hidden toll collector of AI infrastructure." Less flashy in the headlines than Nvidia, but the more Big Tech tries to cut Nvidia dependence with in-house chips, the more Broadcom earns. It sits in the curious position of winning "whichever way the AI-chip fight goes" — whether Nvidia wins or Big Tech's own chips do, both are Broadcom customers.

Hyperscalers partner with Broadcom to secure a "de-Nvidia" option. Depend only on Nvidia GPUs and your pricing leverage weakens and supply risk grows. Co-designing your own ASIC with Broadcom lets you lower chip unit costs long-term and optimize for your workloads. Pricey short-term, but the math says it pays off at sufficient scale.

For investors, Broadcom reads as a "card to diversify AI exposure." For those uneasy about all the AI bet concentrating in one stock (Nvidia), Broadcom is the "ride the same AI boom from another angle" option. But weigh the risks too — software-segment swings, dependence on hyperscaler capex. A company whose revenue leans heavily on a few giant customers is a double-edged sword.

Historical echoes — the light and shadow of semiconductor super-cycles

The semiconductor industry always oscillates between "super-cycle" and "winter." A recent example is the 2020–2021 pandemic-era memory and chip boom. Demand exploded and everyone said "this time is different," but oversupply and softening demand combined to bring a brutal 2022–2023 correction. Hence the old chip-investing lesson: "the steeper the growth rate, the more you should suspect the cycle peak."

There's a successful ride too: Nvidia's transformation from a gaming-GPU company into the center of AI computing. By catching the structural shift in data-center demand early, it turned a transient boom into multi-year growth. Broadcom's bet runs on the same logic — wagering the whole company that AI data-center buildout is a "multi-year infrastructure transition," not a "single-season fad."

But there's a difference. Nvidia stood on broad demand for "general-purpose GPUs," while Broadcom's AI revenue is more concentrated in a handful of hyperscaler ASIC orders. Few, large customers means one customer's plan change moves revenue a lot. So Broadcom's $100B target is tightly bound to the premise that "hyperscaler capex keeps rising." Shake that premise and the curve shakes too.

How rivals counter-play — Nvidia and other chip camps

Nvidia counters with its "integrated ecosystem." Its moat isn't just chips — it's CUDA software, the developer ecosystem, and networking (InfiniBand) bundled together. Even as hyperscalers move some workloads to their own ASICs, the fastest, most-proven general-purpose training often still runs on Nvidia GPUs. Nvidia will hold the line with "Broadcom ASICs are for specific workloads; we cover the whole field."

A rival like AMD is a wildcard too. AMD is rapidly gaining share with its MI-series AI accelerators and data-center CPUs. If hyperscalers adopt AMD general-purpose accelerators more as the "Nvidia alternative," some of the Broadcom-style "custom ASIC" demand could flow there instead. Big Tech ultimately builds a portfolio, weighing Nvidia/AMD general-purpose chips against Broadcom-collab ASICs.

Networking is a battleground too. In AI data centers, the "interconnect" that shuttles data fast between chips is an increasingly important front. As Broadcom pushes Ethernet-based AI networking, Nvidia counters with InfiniBand and its own switches. Who controls "the roads between chips" matters as much as the chips themselves in the next round. Broadcom holding both ASICs and networking is strategic precisely here.

So what actually changes — by who you are

If you're an investor, it's time to update the simple equation "AI boom = Nvidia." Broadcom's numbers prove AI-infrastructure spend is spreading beyond one GPU stock into ASICs, networking, and infrastructure broadly. But before getting excited by hot guidance, recall the customer concentration and the history of semiconductor cycles for a balanced view.

If you're an IT/infrastructure decision-maker, this signals the "AI infrastructure cost structure" is being reshaped. As hyperscalers optimize costs with their own ASICs, it affects the unit economics of cloud AI services long-term. Which chips a given cloud runs on could become a variable that separates future price and performance.

If you're a general reader, you won't feel it directly, but the big picture is worth knowing. Behind the chatbots and AI services you use, enormous fleets of chips and data centers are running, and one pillar of that infrastructure race is an "invisible giant" like Broadcom. In the AI era, the real money often moves in the infrastructure behind the screen.

🥄 Three Things You're Probably Wondering

— So should I buy Broadcom stock? That's your call. The numbers are strong, but a lot of expectation may already be priced in, and customer-concentration and chip-cycle risks are real. Rather than diving in on a hot growth rate, first ask yourself whether this demand is a structure that lasts years.

— Is this a threat to Nvidia? Less a direct threat than "a different lane." Nvidia is general-purpose GPUs; Broadcom is custom ASICs and networking — overlapping yet distinct. More in-house chips at Big Tech trims Nvidia dependence, but general-purpose training demand is still largely Nvidia's. Both ride the same AI boom from different angles.

— Can I trust the $100B target? View it cautiously. The company reiterating it gives it grounding, but the number is tightly bound to the premise that "hyperscaler capex keeps rising." If AI-investment fervor cools or Big Tech tightens spending, the curve bends. A target is just a target — too early to call before actual quarterly results confirm it.

Sources

• Broadcom Inc. Form 8-K, FY2026 Q2 — SEC
• Broadcom Earnings: Record AI Revenue Hits $10.8B — TechTimes
• Broadcom (AVGO) Q2 2026 Earnings Transcript — The Motley Fool
• Broadcom (AVGO) Q2 FY2026 Earnings: Record AI Revenue — HeyGoTrade
• Broadcom (AVGO) Q2 2026 Earnings — MLQ.ai

Numbers and criteria are as of announcement and may change. Investment calls are yours to make!