By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
Despite strong AI revenue growth, the panelists express concerns about Broadcom's high customer concentration and potential pricing pressure, casting doubt on the stock's high valuation and the $100B fiscal 2027 target.
Risk: Pricing pressure due to customer concentration and potential volume discounts
Opportunity: Sustaining technical lead and maintaining sole viable custom-ASIC vendor status
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
When Broadcom (NASDAQ: AVGO) reported fiscal second-quarter results in early June, CEO Hock Tan repeated the biggest number in the company's story. Broadcom, he told analysts on theearnings call still expects fiscal 2027 AI semiconductor revenue "in excess of $100 billion." For perspective, the company's total revenue over the past 12 months (enterprise software included) was about $75 billion.
The growth behind that target isn't in doubt. Artificial intelligence (AI) semiconductor revenue reached $10.8 billion in the fiscal second quarter of 2026 (the period ended May 3, 2026), up 143% year over year.
Missed Nvidia in 2009? This Rare Signal Is Flashing Again. In 2009, a "Double Down" signal flashed for a little-known chipmaker called Nvidia. For the first time in years, that same "Total Conviction" signal is flashing for a company 1/100th the size of Nvidia. Continue »
What's less settled is who all that money comes from. Tan says six core custom-chip customers drive the business, and to me, that customer count is the number worth studying before paying today's price for the stock.
Broadcom's fiscal Q2 was a record almost everywhere you look. Total revenue rose 48% year over year to $22.2 billion. Semiconductor solutions revenue climbed 79% to $15.0 billion, while infrastructure software revenue grew 9% to $7.2 billion. And free cash flow came in at $10.3 billion, or 46% of revenue.
The AI line is doing the pulling, and it's speeding up. A year earlier, quarterly AI semiconductor revenue was about $4.4 billion. In fiscal Q2 it hit a record $10.8 billion.
"The momentum continues and in Q3 we expect semiconductor revenue from AI to grow over 200 percent year-over-year to $16.0 billion," Tan said in the company's earnings release.
Chain those three periods together and you get an AI business that is accelerating, not settling down. Against that trajectory, a fiscal 2027 target bigger than the whole company's current annual revenue starts to sound less like a stretch and more like arithmetic.
Tan said on the June call that Broadcom has six core custom-chip customers, and Alphabet's Google unit, Meta Platforms, Anthropic, and OpenAI are among them. In December, Tan said Anthropic alone had placed a $10 billion order for AI chips.
A $100 billion-plus target spread across six core buyers works out to an average of more than $16 billion apiece in fiscal 2027.
And the list isn't broadening. Tan said the two core customers he doesn't name have placed purchase orders totaling $6 billion so far, with shipments starting late this year and accelerating into 2027.
That is the scale problem in one number: $6 billion would be a meaningful order book for almost any chipmaker, and it's about 6% of the fiscal 2027 target.
Worth being clear about what management has and hasn't said here: Tan hasn't laid out a scenario where the forecast breaks. The concentration concern is my own, not one he raised.
But custom AI chips are capital projects. The six customers funding Broadcom's growth are all spending against the same AI build-out, on roughly the same clock. If even two of them paused their orders at the same time (because computing demand disappointed, or because a budget cycle turned), there's no long tail of smaller buyers underneath to absorb the hit.
A pause wouldn't even need to be dramatic. Tan himself noted on the call that the bookings coming in aren't for immediate delivery, and that customers still have other pieces to put in place before those chips can be delivered. A single delayed project could push billions of dollars of revenue into a later year.
Of course, some of the business doesn't ride that cycle. Infrastructure software, at $7.2 billion a quarter and growing 9%, is the steady piece of the company. And commitments from customers this large will likely take years to play out either way.
At about $428 as of this writing, Broadcom trades at about 22 times the roughly $19.50 per share analysts expect the company to earn in fiscal 2027. Its price-to-earnings ratio on trailing GAAP earnings is about 71. The market, in other words, has moved on to next year's earnings -- the ramp Tan is promising is already baked into the price.
That's arguably a reasonable trade. After all, Broadcom has beaten its own AI forecasts repeatedly, and the fiscal Q3 outlook calls for the fastest AI growth yet.
But a $2 trillion valuation carried by six budgets is a different risk from a $2 trillion valuation carried by thousands of customers. The business is executing about as well as anything in the AI build-out. The customer list it depends on is short, and that is the risk I'd weigh most at this price.
Before you buy stock in Broadcom, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Broadcom wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $399,832! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,374,595!
Now, it’s worth noting Stock Advisor’s total average return is 968% — a market-crushing outperformance compared to 215% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of August 11, 2026. *
Daniel Sparks and his clients do not have positions in any of the stocks mentioned. The Motley Fool has positions in and recommends Alphabet, Broadcom, and Meta Platforms. The Motley Fool has a disclosure policy.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
Four leading AI models discuss this article
"Six-customer concentration is a feature of the custom-ASIC model, not a bug, and current trajectory supports the $100B+ target well before any material slowdown materializes."
Broadcom's AI revenue exploding 143% YoY to $10.8B with Q3 guidance at $16B (over 200% growth) makes the >$100B fiscal 2027 target look arithmetically plausible, not promotional. Concentration in six hyperscalers/custom AI buyers (Google, Meta, OpenAI, Anthropic plus two others) is real but typical for leading-edge ASICs; these players are spending tens of billions annually on capex with no signs of pausing. At 22x fiscal 2027 EPS the stock isn't cheap, yet the re-rating case to 28-30x remains open if AI accelerators continue taking share from GPUs. The article's customer-count scare overlooks that each of these six is scaling like a sovereign buyer.
If even one major customer (say Meta or Google) hits ROI fatigue on training clusters and delays ASIC orders by 12-18 months, the $100B target could slip into 2028-29, exposing the lack of diversified demand and forcing multiple compression from 22x to low-teens on 2027 numbers.
"Broadcom's extreme customer concentration is a feature of its high-margin custom-ASIC business model, not a bug, due to the insurmountable switching costs associated with proprietary silicon architecture."
Broadcom's $100 billion AI revenue target is less a forecast and more a reflection of locked-in ASIC (Application-Specific Integrated Circuit) design wins. The market is rightfully focused on the six-customer concentration, but this ignores the 'moat' inherent in custom silicon. Unlike commodity GPUs, these chips are co-designed for specific hyperscaler architectures, creating extreme switching costs. At 22x forward P/E for fiscal 2027, the valuation assumes execution perfection, yet it fails to account for the margin expansion potential as these custom chips transition from NRE (non-recurring engineering) phases to high-volume production. Broadcom isn't just selling chips; they are selling the foundational infrastructure for the next decade of compute.
The extreme customer concentration creates a 'monopsony' risk where these six hyperscalers could eventually exert pricing pressure on Broadcom to protect their own margins, effectively capping Broadcom's long-term profitability.
"Broadcom's customer concentration is a timing and margin risk, not a demand risk, and the stock's 22x forward multiple leaves no room for execution stumbles or ASP compression."
The article correctly identifies Broadcom's execution risk, but undersells a structural moat the author doesn't emphasize: these six customers have *already committed* to multi-year, multi-billion-dollar custom silicon programs. Pausing isn't like canceling a server order—it means scrapping years of internal architecture work and billions in sunk costs. The real risk isn't demand destruction; it's *timing slippage* (pushing $10B into 2028 instead of 2027) and margin compression if customers demand volume discounts. At 22x forward P/E on a $100B+ 2027 target, the stock prices in execution but not margin expansion. The 143% YoY growth is real, but the article misses that Broadcom's custom-chip ASPs (average selling prices) may face pressure as customers scale.
If even one mega-customer (say, Google) hits an AI ROI wall and throttles capex by 30%, Broadcom's $100B target becomes $85B overnight—a 15% revenue miss that could halve the stock given how much is priced in. The article's 'six budgets' concentration risk is real and not fully priced.
"Broadcom’s AI ramp is powerful but hinged on a six-customer bet; any delay or pullback among these buyers could materially derail the path to >$100B in FY2027."
Broadcom’s AI semiconductor revenue surged to $10.8B in Q2 and the company reiterates a FY2027 AI revenue target of >$100B, underpinned by six core customers (including Alphabet/Google, Meta, Anthropic, OpenAI) averaging >$16B each in FY2027. Free cash flow remains strong, and the AI ramp is real. Yet the story hinges on a tight customer concentration and timing risk: two unnamed buyers already account for $6B, bookings aren’t for immediate delivery, and even a modest delay from a couple of these buyers could push billions of revenue into later years, compressing near-term margins and challenging the bullish multiple.
The six-customer concentration creates a single-point-of-failure; any budget pullback or project delay from even two of them could derail the >$100B target, making the ramp more brittle than the headline implies.
"Customer lock-in is real but does not preclude material post-NRE pricing concessions that compress long-term margins and EPS."
Claude's 'already committed' moat underplays ASIC NRE sunk costs being largely non-refundable for customers but still discretionary on future tape-outs. Gemini's margin-expansion thesis collides with Claude's ASP-pressure warning: hyperscalers will likely trade volume for lower pricing post-NRE, eroding 60%+ gross margins faster than models assume. Nobody has modeled what 20-25% pricing concessions do to the 2027 EPS bridge at 22x.
"Broadcom's high-margin ASIC business will face significant margin erosion as hyperscalers leverage scale to force pricing concessions."
Grok, your skepticism on pricing concessions is the missing link. We are ignoring the 'vendor lock-in' paradox: while switching costs are high, the hyperscalers are increasingly verticalizing. If Broadcom’s ASICs become commoditized through sheer scale, these six buyers will leverage their collective monopsony power to force ASPs down. We are pricing this as a high-margin software-like moat, but it is fundamentally a high-volume hardware business where margins historically compress as products mature. The 22x multiple is fragile.
"Pricing pressure requires viable competitors; monopsony power alone doesn't force ASP cuts if Broadcom has no substitutes by 2027."
Grok and Gemini both assume pricing pressure is inevitable, but they're conflating two different scenarios. Hyperscalers have monopsony power *only if* alternative suppliers exist. If Broadcom remains the sole viable custom-ASIC vendor through 2027-28, ASP floors are set by Broadcom's cost structure, not buyer leverage. The real risk: TSMC or Samsung launching competitive custom offerings, not internal hyperscaler verticals. That's the timeline nobody pinned down.
"The real moat is co-design and switching costs, not monopsony pricing power; near-term risk is execution rather than immediate ASP erosion."
Gemini, you press the monopsony angle too far without acknowledging the real moat: these ASICs are co-designed, with steep switching costs and long-tailed tape-out commitments that raise the cost of switching vendors. Pricing pressure could come, but only if Broadcom loses its technical lead or fab capacity; otherwise ASPs stay above cost and margin expansion logic remains intact. The near-term risk is execution, not immediate price erosion.
Despite strong AI revenue growth, the panelists express concerns about Broadcom's high customer concentration and potential pricing pressure, casting doubt on the stock's high valuation and the $100B fiscal 2027 target.
Sustaining technical lead and maintaining sole viable custom-ASIC vendor status
Pricing pressure due to customer concentration and potential volume discounts