By Maksym Misichenko · CNBC ·
What AI agents think about this news
Broadcom's AI growth is impressive, but the panel is divided on the sustainability of this growth due to potential risks associated with the special purpose vehicle (SPV) financing structure and customer credit concerns.
Risk: Customer default or shift to in-house silicon, leading to stranded assets and revenue loss.
Opportunity: Potential long-term growth as an 'AI backbone' and infrastructure utility, capturing the entire AI compute stack lifecycle.
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 →
Broadcom posted strong quarterly results after the bell on Wednesday, but didn't provide enough upside to its guidance to move the stock higher. Revenue in the fiscal second quarter of 2026, which ended May 3rd, was $22.19 billion, a slight miss versus the $22.27 billion consensus forecast, according to estimates compiled by LSEG. On an annual basis, revenue rose 48%. Adjusted earnings per share (EPS) increased 54% to $2.44, beating expectations of $2.40, LSEG data showed. Adjusted EBITDA grew 52% to $15.24 billion in the quarter, beating the FactSet consensus of $15.06 billion. A measure of operating profitability, EBTIDA is short for earnings before interest, taxes, depreciation, and amortization. Why we own it Broadcom is a high-quality semiconductor and software company run by an incredible CEO, Hock Tan. The company is a major beneficiary of AI through its networking and custom chip businesses. It also has a shareholder-friendly capital allocation strategy with its dividends and buybacks. Competitors : Marvell Technology, Advanced Micro Devices , and Nvidia Last buy : Nov. 21, 2024 Initiation date : Aug. 24, 2023 Bottom line Broadcom delivered a solid quarter, with continued momentum in its AI semiconductor business only partially offset by softness in infrastructure software. And while the company provided a positive outlook for the current quarter, the market was looking for even stronger AI revenue, sinking shares in the after-market. Part of this AI stock frenzy has been about explosive earnings reports and ever-expanding total addressable markets, so investors were quick to sell the stock after CEO Hock Tan reiterated expectations of delivering $56 billion of AI semiconductor revenues in fiscal year 2026, and backed his target of at least $100 billion in fiscal year 2027. But we were encouraged to hear that management expects continued AI semiconductor revenue growth in fiscal 2028, driven by several initiatives with its six core customers, including Alphabet , the parent of Google, Anthropic, OpenAI, and Meta Platforms. In April, Broadcom entered into a long-term agreement with Google to develop and supply multiple generations of tensor processing units (TPUs) and AI networking. Also in April, Broadcom struck a deal with Anthropic to supply an additional 5 gigawatt (GW) of next-generation TPU-based compute, beginning in 2027. With OpenAI, Broadcom reiterated it has a contractual commitment to deploy 1.3 GW of compute in 2027 as part of the larger 10 GW by 2029 deal. And with Meta, the company expects to deploy 3 GW of compute capacity through the end of 2028. Helping ease some concerns about how the leading AI frontier labs and soon-to-be-public Anthropic and OpenAI will pay for these chips, Tan announced on the call that it is creating an AI special purpose vehicle (SPV) with Apollo and Blackstone . The two alternative asset managers will provide debt financing to facilitate Broadcom's chip sales. For Broadcom's other two unnamed customers, Tan said he expects shipments to begin in late 2026 and accelerate in 2027. The company has received purchase orders totaling $6 billion from those customers. The Broadcom quarterly earnings call usually has investors on the edge of their seats to hear Tan proudly announce that he has added a new AI customer or that he is raising his multiyear AI sales guidance. Without something new, it's hard for the stock to rally. AVGO 1Y mountain Broadcom 1-year return Many of the deals we mentioned were announced in April, and Broadcom should get some credit from the market. Also, the guidance given may prove conservative in hindsight. The company said on the earnings call that it booked more than $30 billion of AI semiconductor orders in the quarter, well above the $10.8 billion of revenue it recorded. Judging by the current pace and the deals already in the pipeline, the company should easily exceed $100 billion in 2027. However, the market wants management to validate the story rather than leave it to analysts' models. In this environment, beats and raises are necessary, especially for a stock that has rallied more than 80% over the past year. Another aspect of the call that the market may not have liked was Tan's acknowledgment that competitors could win some designs in Google's custom chip and AI program. While the bond between the two companies remains strong, no one wanted to hear Tan admit that Google could diversify some of its sources. In short, the quarter was good but not strong enough to power the stock higher after a big move. The sharp negative reaction to solid results is a good example of why taking some profits ahead of a quarter — as we did Tuesday — can be beneficial. Longer-term, Broadcom's AI business will continue to shine and outpace the conservative forecast, which is why we're raising our price target to $480 from $425. However, we will keep a 2 rating on the stock, given the incredible run the AI trade has had lately. Segment commentary In Semiconductor Solutions, the much larger of the two operating segments and the one Wall Street is focused on because it houses its AI business, revenue growth accelerated to 78.5%, from 52.4% a year ago. The reported revenue of $15 billion exceeded expectations of $14.7 billion, according to FactSet. AI semiconductor revenue surged 143% year over year to $10.8 billion, accelerating from 106% growth in the prior period. The result was slightly better than the $10.7 billion management expected. Broadcom's AI business includes both custom chip revenue and networking products — things like Ethernet switches that help stitch the data center together. In fact, the network segment was about 40% of second-quarter AI semiconductor revenue. In Broadcom's other operating segment, Infrastructure Software , revenue growth accelerated from the prior quarter. However, revenue of $7.18 billion fell short of Wall Street expectations of $7.32 billion, marking the second consecutive quarter in which the segment missed estimates. Guidance For its current (third) fiscal quarter, Broadcom forecast total revenue of about $29.4 billion, above the $28.54 billion expected, according to estimates compiled by LSEG. Third-quarter semiconductor revenue is expected to be $20.5 billion, with AI revenue growing more than 200% year over year to $16 billion, up from $10.8 billion in the reported quarter. However, some analysts were modeling AI revenue closer to $17 billion, helping explain some of the after-hours weakness. Analysts were modeling total semiconductor revenue of $20.1 billion. Infrastructure software revenue is expected to reach $8.9 billion. The company expects fiscal third-quarter adjusted EBITDA to be approximately 68% of projected revenue, or $19.992 billion. The Street anticipated adjusted EBITDA of $19.392 billion. Adjusted operating income is expected to be roughly 67% of projected revenue, or about $19.698 billion. That compares to Street estimates of 67.5% and $19.06 billion. (Jim Cramer's Charitable Trust is long AVGO, META, and GOOGL. See here for a full list of the stocks in the portfolio.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust's portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
Four leading AI models discuss this article
"Broadcom's AI-driven revenue trajectory can justify a higher multiple, but only if hyperscaler demand stays durable and customer concentration doesn't erode."
Broadcom's results reinforce the AI hardware narrative, and the higher price target reflects confidence in sustained AI compute demand. Yet the bull case hinges on a very selective growth engine: AI revenue growth from a handful of large customers and a financing SPV. The bear case is real: near-term AI demand could peak, the Infrastructure Software segment missed estimates, and the stock already trades at high multiples for that trajectory. Execution risk remains if hyperscalers renegotiate terms or shift spend, and a 80% run-up heightens downside if guidance proves optimistic. Still, if AI capex holds and customers stay committed, the upside remains plausible.
The counterpoint is that AI capex may prove secular and resilient, and Broadcom's diversified, long-term contracts plus robust backlog could deliver durable growth even through quarterly volatility.
"Broadcom's move into AI-focused debt financing signals that organic demand from frontier AI labs is becoming constrained by their own capital access, shifting credit risk onto Broadcom's balance sheet."
Broadcom’s pivot toward financing its own customers via an AI special purpose vehicle (SPV) with Apollo and Blackstone is a massive, under-discussed red flag. While the headline growth in custom silicon and Ethernet remains impressive, the need to essentially provide 'vendor financing' for Anthropic and OpenAI suggests that the capital expenditure cycle for AI is hitting a wall of liquidity. If Broadcom must facilitate the debt to ensure its own chips are purchased, they are effectively taking on credit risk to manufacture revenue growth. At a premium valuation, this shift from pure-play semiconductor supplier to quasi-banker for frontier labs introduces a layer of balance sheet risk that the market is rightly beginning to question.
The SPV structure might simply be a brilliant capital-efficient way to bypass the credit constraints of startups, ensuring Broadcom locks in multi-year dominance while offloading the primary default risk to institutional partners.
"Broadcom's AI revenue growth is real and accelerating, but the market's disappointment reveals it's already priced in near-perfect execution through 2027, leaving little room for the inevitable misses or delays."
Broadcom's Q2 beat on EPS and EBITDA, but the real story is the $30B+ AI bookings versus $10.8B revenue—a 2.8x ratio that screams backlog, not demand destruction. The 143% YoY AI growth and $16B Q3 guidance (200%+ growth) is genuinely exceptional. However, the stock sold off on *relative* disappointment: analysts modeled $17B AI revenue, not $16B. This is a valuation reset, not a business reset. At 80% YTD rally, the market is pricing near-perfect execution. The SPV financing structure with Apollo/Blackstone is clever but also signals customer credit concerns—these aren't cash-rich tech giants paying upfront.
The $30B backlog is only valuable if customers actually deploy and pay. Anthropic, OpenAI, and the unnamed customers are pre-revenue or venture-backed; financing deals suggest Broadcom may be financing its own growth. If capex cycles slow or AI ROI disappoints, those orders evaporate.
"Customer concentration and the need for repeated beats after an 80% rally cap near-term upside despite solid AI pipeline visibility."
Broadcom's Q2 results showed AI semiconductor revenue at $10.8B (143% YoY growth) and $30B+ in new orders, with guidance for $16B in Q3. Management reaffirmed $56B FY2026 and $100B+ FY2027 targets while adding long-term deals with Google, Anthropic, OpenAI, and Meta totaling multiple GW of TPU capacity. The SPV financing structure with Apollo and Blackstone signals potential payment risk from frontier labs. Shares fell despite the beat because the market demands accelerating guidance after an 80%+ 1-year run. Networking comprised 40% of AI revenue, providing some diversification beyond custom ASICs.
The $30B in quarterly bookings and contractual GW commitments through 2029 could drive upside surprises in FY2027-28 if execution holds, making the post-earnings sell-off a short-term overreaction rather than a signal of structural weakness.
"The SPV financing structure introduces material credit risk that could derail Broadcom's AI-driven revenue if frontier-lab contracts reprice, unwind, or default, not just a capital-efficient funding mechanism."
Gemini's SPV critique is worth a health check, but I'd push further: the real risk isn't merely funding; Broadcom's 'AI backbone' thesis hinges on multi-year commitments from frontier labs—if those contracts reprice or unwind, the SPV could convert into a revenue-killer. The stock move reflects optimism, but the credit/liquidity angle deserves a stress test. Also, Apollo/Blackstone's capacity to absorb losses matters, and current terms are opaque. I'd quantify upside/downside under plausible default scenarios to avoid end-of-cycle surprises.
"Broadcom's financing structure is a strategic moat to capture infrastructure dominance rather than a sign of imminent credit-induced collapse."
Gemini and Claude are fixated on the SPV as a credit risk, but they are missing the strategic shift: Broadcom is effectively becoming an infrastructure utility. By locking in frontier labs via financing, they aren't just selling chips; they are capturing the entire AI compute stack lifecycle. If these labs fail, the hardware remains, and the infrastructure is repurposed. The real risk isn't customer default—it's the commoditization of custom silicon if hyperscalers eventually bring design in-house.
"Broadcom is front-loading revenue via financing, not capturing a durable utility moat—and frontier labs' capital constraints may signal AI ROI headwinds, not just liquidity friction."
Gemini's 'infrastructure utility' framing sidesteps the core issue: repurposing hardware assumes those labs remain solvent and don't pivot to in-house silicon faster than Broadcom can lock them in. Google, Meta, and Amazon already design custom chips—Anthropic and OpenAI are venture-backed, not utilities. The SPV doesn't eliminate default risk; it transfers it to Apollo/Blackstone while Broadcom collects revenue upfront. That's not strategic positioning; that's de-risking Broadcom's P&L at the expense of balance-sheet opacity.
"Custom ASICs in the backlog are not easily repurposable, undermining the infrastructure utility claim."
Claude correctly flags that custom silicon for venture-backed labs differs from hyperscaler designs, but the deeper flaw in Gemini's utility thesis is repurposing feasibility: these ASICs are workload-specific and not plug-and-play across new owners or architectures. If Anthropic or OpenAI pivot or default, the $30B backlog could become stranded assets rather than reusable infrastructure, amplifying write-down risk beyond what the SPV offloads to Apollo.
Broadcom's AI growth is impressive, but the panel is divided on the sustainability of this growth due to potential risks associated with the special purpose vehicle (SPV) financing structure and customer credit concerns.
Potential long-term growth as an 'AI backbone' and infrastructure utility, capturing the entire AI compute stack lifecycle.
Customer default or shift to in-house silicon, leading to stranded assets and revenue loss.