Broadcom Q2 2026 Earnings: $100 Billion AI Target Already Looks Conservative as Bookings Surge to $30 Billion in a Single Quarter
Q2 fiscal 2026 earnings call, June 3, 2026 — Broadcom blows past estimates and raises the stakes for fiscal 2027 and beyond
Broadcom reported fiscal second-quarter revenue of $22.2 billion, up 48% year-on-year and above guidance, but the headline number almost undersells what happened. AI semiconductor revenue hit $10.8 billion in the quarter, up 143% year-on-year, and the company is now guiding to $56 billion in AI chip revenue for the full fiscal year 2026. More striking still, Broadcom reaffirmed and effectively reinforced its guidance for AI semiconductor revenue "in excess of $100 billion" in fiscal 2027 — and CEO Hock Tan made clear that number is a floor, not a ceiling. "It will exceed very easily $100 billion in 2027," he said, adding that fiscal 2028 is expected to be "a substantial growth" from 2027.
$30 Billion in AI Bookings in a Single Quarter
The single most important data point from the call was largely buried: in Q2 alone, Broadcom received over $30 billion in AI semiconductor bookings against $10.8 billion shipped. That is not a typo. Backlog is now so deep that management's visibility has extended from 2027 three months ago to 2028 today. Hock Tan was unambiguous about what is driving this: "Customers have realized that to get compute, you need lead time. They are placing orders early, and they are placing orders in fairly huge demand." The $30 billion single-quarter booking figure implies that the widely-cited $73 billion 18-month backlog disclosed in December 2025 has been functionally superseded. Analysts who modeled that number linearly over six quarters were always going to be wrong; the reality is running well ahead of even the most aggressive interpretations.
Six Customers, Gigawatts of Commitments, and a New Financing Platform
Broadcom now has six named AI customers, and the commitments disclosed on this call represent a qualitative step-change in the company's strategic positioning. With Google, a long-term agreement announced in April covers multiple generations of TPUs and AI networking, with Tan describing the dollar commitment as "very, very substantial." He acknowledged that Google will maintain some supplier diversity but said Broadcom's contractual commitment from Google is firm and large. With Anthropic, Broadcom is providing over one gigawatt of TPU-based compute in 2026 and has signed an agreement for another five gigawatts of next-generation compute beginning in 2027. With OpenAI, silicon has been delivered and production is on track for late 2026, with a contractual commitment to deploy 1.3 gigawatts in 2027 as part of a larger 10-gigawatt-by-2029 framework. With Meta, Broadcom announced in April a partnership for multiple generations of MTIA XPUs, with three gigawatts to be deployed through end-2028 and an initial purchase order for one gigawatt already received.
Two additional unnamed customers are expected to begin shipments in late 2026, with purchase orders totaling $6 billion already received. Tan confirmed that the aggregate 2027 plan remains approximately 10 gigawatts of shipments, back-half weighted, which sets up what he described as a particularly strong trajectory into 2028.
To help frontier model customers fund and access the sheer volume of compute they need, Broadcom is launching what it calls the AI XPV platform, a vehicle co-created with Apollo and Blackstone and other institutional investors designed to deploy more than 20 gigawatts of compute capacity through 2028. The first tranche, valued at $35 billion, is being launched by Apollo. Tan was precise about what this is and is not: "We are a chip business. What we are doing is creating a vehicle to have these chips funded for LLM players who might otherwise have difficulty getting access to our technology, which provides them with the lowest power and the lowest cost."
Networking at 40% of AI Revenue — But That May Be the Peak Mix
Networking represented approximately 40% of Q2 AI revenue, a figure that Goldman Sachs analyst Jim Schneider probed on the call. Tan's answer was notably candid: "I see that as probably as high as that percentage of total AI revenue would go. The more expected percentage as a share of total AI revenue for networking would be closer to around 30%." He explained that the current elevated mix reflects a favorable period where Broadcom is selling networking into non-XPU footprints simultaneously with XPU-driven networking growth — a confluence of factors unlikely to persist at the same intensity. The company has been shipping the industry's only 100-terabit Ethernet switch, Tomahawk 6, for over a year, and is taping out a 200-terabit next-generation switch this quarter. Broadcom describes itself as having "at least one generation of technology and product leadership" in AI networking, and the co-packaged optics ecosystem — 1.6-terabit DSPs, CW and EML lasers — is described as the "de facto standard in the industry."
Gross Margin Compression Is Real but the Operating Leverage Story Holds
Gross margin came in at 77.1% in Q2, down 230 basis points year-on-year, and will compress further to approximately 74% in Q3 as AI semiconductor revenue accelerates to $16 billion, up over 200% year-on-year. CFO Kirsten Spears, who announced her retirement effective June 12, was direct about the mechanics: custom XPUs and TPUs carry lower gross margins than networking, and as semiconductors grow to represent a larger share of the revenue mix relative to infrastructure software — which runs at 93% gross margin — consolidated gross margins will face structural pressure. Spears explicitly recommended that investors model the two segments separately. The saving grace is that operating margins remain firm: Q2 operating margin was a record 67.3%, up 200 basis points year-on-year despite the gross margin headwind, and Q3 operating margin guidance is flat at 67%, demonstrating what management rightly calls strong operating leverage. Adjusted EBITDA in Q2 was $15.2 billion, or 69% of revenue.
Software Acceleration Is an Underappreciated Catalyst
Infrastructure software, the VMware-anchored segment, generated $7.2 billion in Q2, up 9% year-on-year, with ARR growth sustaining at 17% year-on-year. But the Q3 guide of approximately $8.9 billion, up 31% year-on-year, signals a meaningful step-up in the growth rate. Tan attributed this to the strong wave of server deployments globally: "With strong server demand globally, the deployment of VCF 9.1 for on-prem cloud computing is extremely strong, driving robust revenue growth." The just-released VMware Cloud Foundation 9.1 adds heterogeneous compute support across AMD, Intel, and NVIDIA platforms, enabling enterprise customers to run AI, Kubernetes, and traditional virtualized workloads on a common private cloud. Tan also pushed back on concerns about agentic AI disrupting the software business: given that VMware products sit at the hypervisor layer — essentially at the hardware abstraction level — he sees limited structural risk from changes in the application or AI software layer above it.
Content Per Gigawatt Is Set to Rise Significantly
TD Cowen analyst Joshua Buchalter flagged an apparent tension: if Broadcom is shipping roughly 10 gigawatts in fiscal 2027 at the previously cited $15 billion to $20 billion per gigawatt of compute content, the implied revenue would substantially exceed the $100 billion guidance. Tan's response was instructive. Revenue per gigawatt will increase over time as XPU architectures evolve — incorporating embedded CPUs, more SRAM, and multi-die configurations with HBM — but these content increases track generational transitions rather than quarterly increments. The $100 billion figure for fiscal 2027 therefore already embeds some conservatism relative to the raw gigawatt math, and the content trajectory into 2028 suggests further upside as next-generation architectures mature.
Non-AI Semiconductors Quietly Recovering
Non-AI semiconductor revenue of $4.2 billion in Q2 grew 6% year-on-year, with bookings of over $6 billion during the quarter providing what management called "a clear indication we are on the path towards a full cyclical recovery." Broadband, server storage, and enterprise networking were all up, partially offset by a seasonal wireless decline. Q3 guidance of approximately $4.5 billion, up 12% year-on-year, suggests the recovery is progressing steadily if not dramatically.
Supply, Balance Sheet, and Capital Returns
Broadcom ended Q2 with $19.6 billion in cash, up from $14.2 billion in Q1, and generated a record $10.3 billion in free cash flow, representing 46% of revenue. Inventory rose to $4.3 billion, with days on hand increasing from 68 to 86 days as the company deliberately builds buffer ahead of the AI acceleration in the second half. On supply, Tan was measured but confident: "We are very comfortable that we have been able to secure supply for our needs in '26 and '27, and we are working on '28 and '29 right now." The company paid $3.1 billion in dividends in Q2 at $0.65 per share. Incoming CFO Amie Thuener was introduced on the call but did not speak.
Q3 consolidated revenue guidance is $29.4 billion, up 84% year-on-year. With AI semiconductor revenue of $16 billion, non-AI semiconductor revenue of approximately $4.5 billion, and software revenue of $8.9 billion, Broadcom is entering the back half of fiscal 2026 with a revenue run rate that was almost unimaginable twelve months ago — and management's own language suggests the trajectory into 2027 and 2028 has, if anything, steepened since last quarter.