$AVGO EXECUTIVE CALL SUMMARY: Broadcom Inc (03/04/26)
Broadcom reported fiscal Q1 2026 results above internal expectations and guided to a sharp acceleration in fiscal Q2 2026, with AI semiconductors and AI networking as the dominant incremental growth vectors. Consolidated revenue reached $19.3 billion (+29% y/y) and adjusted EBITDA was $13.1 billion (68% margin), with management explicitly attributing the revenue beat to “better-than-expected growth in AI semiconductors.” Based on disclosed growth rates, the implied year-ago revenue baseline was approximately $15.0 billion, indicating approximately $4.3 billion of incremental revenue in Q1; the disclosed AI semiconductor growth implies approximately $4.3 billion of incremental AI semiconductor revenue, suggesting that approximately 100% of consolidated y/y growth was AI-driven, with non-AI semiconductors and Infrastructure Software collectively near-flat y/y. Q2 guidance implies a further step-up to approximately $22.0 billion of revenue (+47% y/y; +14.0% q/q) with adjusted EBITDA maintained at approximately 68%. Semiconductor Solutions is expected to drive essentially all incremental growth, guided to $14.8 billion (+76% y/y; +18.4% q/q) including AI semiconductor revenue of $10.7 billion (+140% y/y; +27.4% q/q). Management escalated the medium-term ambition by stating: “Today, in fact, we have line of sight to achieve AI revenue for – from chips, just chips, in excess of $100 billion in 2027,” supported by claims of multi-year customer roadmaps and supply-chain capacity secured for calendar ’26 through ’28. Infrastructure Software remained a high-margin ballast at $6.8 billion (+1% y/y) with Q2 guided to $7.2 billion (+9% y/y), and VMware was highlighted via 13% y/y revenue growth, >$9.2 billion of Q1 total contract value bookings, and 19% y/y ARR growth.
FINANCIAL PERFORMANCE AND QUALITY OF EXECUTION
The quarter displayed a combination of demand-driven upside in AI semiconductors and continued operating discipline. Non-GAAP gross margin was 77% of revenue and consolidated operating expenses were $2.0 billion (including $1.5 billion of R&D). Non-GAAP operating income reached $12.8 billion (+31% y/y), driving operating margin expansion of 50 bps y/y to 66.4%. The implied year-ago operating income was approximately $9.8 billion, indicating approximately $3.0 billion of incremental operating income on approximately $4.3 billion of incremental revenue; this implies an incremental operating margin of approximately 70% y/y, highlighting strong operating leverage. Adjusted EBITDA of $13.1 billion equated to 68% margin and exceeded the prior margin guide of 67%, indicating that incremental AI revenue was accretive to the operating model despite increased leading-edge technology content and supply-chain complexity.
Cash conversion remained a central element of the equity story. Free cash flow was $8.0 billion (41% of revenue) against $250 million of capex, consistent with a structurally asset-light model. Capital returns totaled $10.9 billion in the quarter ($3.1 billion of dividends and $7.8 billion of repurchases), exceeding free cash flow and implying incremental balance-sheet usage. The repurchase pace implied approximately 23 million shares retired in the quarter, or roughly 0.5% of the guided Q2 non-GAAP diluted share count of 4.94 billion. Cash ended at $14.2 billion, and an incremental $10.0 billion repurchase authorization through the end of calendar 2026 signaled continued emphasis on shareholder distributions alongside AI capacity investments.
Working-capital positioning shifted more defensively toward supply assurance. Inventory increased to $3.0 billion and days of inventory rose to 68 days from 58 days in fiscal Q4, explicitly framed as pre-positioning “in anticipation of accelerating AI semiconductor growth.” This inventory build supports near-term revenue conversion in a constrained supply environment but increases exposure to demand timing risk if customer deployment schedules slip or if component procurement commitments prove misaligned with actual pull.
SEMICONDUCTOR SOLUTIONS: AI ACCELERATION AND MIX SHIFT
Semiconductor Solutions revenue was $12.5 billion (+52% y/y) and represented 65% of consolidated revenue. Segment gross margin was approximately 68% (+30 bps y/y), operating expenses were $1.1 billion (8% of revenue), and operating margin expanded to 60% (+260 bps y/y), reflecting meaningful operating leverage on accelerating AI volume. The disclosed growth implies Semiconductor Solutions revenue of approximately $8.2 billion in fiscal Q1 2025, indicating approximately $4.3 billion of incremental semiconductor revenue y/y; this nearly matches consolidated revenue growth, consistent with Infrastructure Software remaining near-flat y/y.
AI semiconductor revenue was $8.4 billion (+106% y/y) in Q1 and guided to $10.7 billion (+140% y/y) in Q2. AI has become the dominant driver of the segment: AI semiconductors represented approximately 67.2% of Semiconductor Solutions revenue in Q1 (versus an implied ~49.6% in the year-ago quarter), rising to approximately 72.3% in the Q2 guide. At the consolidated level, AI semiconductor revenue represented approximately 43.5% of Q1 revenue, rising to approximately 48.6% in the Q2 guide. The non-AI semiconductor business was $4.1 billion in Q1 (flat y/y) and guided to approximately $4.1 billion in Q2 (+4% y/y). In Q1, enterprise networking, broadband, service, and storage were described as up y/y, offset by a seasonal decline in wireless, implying that non-AI stability was driven by broad-based demand rather than a single end-market rebound.
A central narrative in the prepared remarks was the breadth and scale of the custom accelerator franchise. The custom accelerator business was stated to have grown 140% y/y in Q1, and management described the “ramp of custom AI accelerators across all our 5 customers” as “progressing very well,” later updated to 6 customers with the addition of OpenAI. Specific customer disclosures were unusually direct for Broadcom’s historical style and appeared designed to reinforce both demand visibility and durability:
•For Google, demand for the “7th-generation Ironwood TPU” was described as strong in 2026 with “even stronger demand” expected from subsequent generations.
•For Anthropic, management stated: “we are off to a very good start in 2026 for 1-gigawatt of TPU compute” and “for ’27, this demand is expected to search in excess of 3-gigawatts of compute.”
•For Meta, management stated: “Contrary to recent analyst reports, Meta’s custom accelerator, MDIA roadmap, is alive and well. We’re shipping now,” and the next-generation XPUs were framed as scaling to “multiple gigawatts in ’27 and beyond.”
•For an additional 2 undisclosed customers, management stated that shipments in 2026 are expected to be strong and “more than double in 2027.”
•For OpenAI, management stated: “We expect OpenAI deploying in volume their first-generation XPU in 2027 at over 1-gigawatt of compute capacity.”
These disclosures imply that the 2027 demand set is increasingly anchored in compute-capacity planning (gigawatts) rather than an abstract dollar TAM, which improves the auditability of the narrative but also increases execution and expectation risk. In Q&A, management acknowledged the relevance of this framing: “you can look at it at gigawatts, which is the right way to look at it instead of dollars,” and added that “if you look at it by gigawatt in ’27, we are seeing it getting close to 10-gigawatts.” If realized, 10 gigawatts of deployments across 6 customers supports the >$100 billion chips-only ambition; however, the range of “dollars per gigawatt” was explicitly described as variable by customer, limiting direct extrapolation and leaving meaningful model sensitivity around mix, system content, and pricing.
Management emphasized durable competitive barriers in custom accelerators, particularly against internal customer tooling (COT) initiatives. The Q&A response articulated a multi-factor moat: leading-edge silicon design capability, high-speed SerDes, advanced packaging, networking architecture, and volume manufacturing at yield. A notable manufacturing execution quote was: “Anybody can design a chip in a lab that works well. Can you produce 100,000 of those chips quickly at yields that you can afford?” This positioning suggests that Broadcom views its competitive edge as less about access to the customer and more about repeated, high-yield, high-volume delivery across generations, consistent with historical execution but sensitive to ecosystem constraints in foundry capacity, advanced packaging, and high-bandwidth memory.
AI NETWORKING: SCALE-OUT/ SCALE-UP POSITIONING AND MIX IMPLICATIONS
Networking was positioned as a major incremental driver and as an area of share gains. Q1 AI networking revenue grew 60% y/y and represented approximately 33% of total AI revenue, implying approximately $2.8 billion of Q1 AI networking revenue within the $8.4 billion AI semiconductor total. For Q2, management projected that AI networking would “grow to 40% of total AI revenue” and added: “We are clearly gaining share in networking.” On the Q2 guide of $10.7 billion AI semiconductor revenue, a 40% mix implies approximately $4.3 billion of AI networking revenue, up approximately 53% q/q. Under that mix shift, the implied sequential growth in AI compute silicon is approximately 15% q/q, indicating that networking attach and content are rising faster than compute in the near term.
Management attributed the networking momentum to technology leadership in switching silicon, SerDes, and optical DSPs. The scale-out thesis highlighted “Tomahawk 6 switch at 100-terabit per second” and “200G SerDes” as capturing hyperscaler demand across both XPU- and GPU-based clusters. A 2027 extension was flagged via Tomahawk 7 with “double the performance.” In scale-up, Broadcom emphasized an architectural advantage in extending direct attached copper through higher SerDes speeds, arguing that remaining on copper reduces cost and power versus optical: “As we next step up to 400G SerDes in 2028, our XPU customers will likely continue to stay on direct attached copper, and this is a huge advantage as the alternative of going to optical is more expensive and requires significantly more power.” In Q&A, this positioning was sharpened as a pragmatic delay of CPO adoption: “You don’t need to go run into some bright shiny objects called CPO, even as we are the lead in CPOs. CPOs will come in its time. Not this year, maybe not next year, but in its time.”
The networking narrative has 3 investment implications. First, Broadcom’s AI exposure is not limited to custom compute accelerators and can benefit even when customers deploy GPU architectures, potentially reducing reliance on a single compute paradigm. Second, the shift toward Ethernet as the standard protocol in both scale-out and scale-up clusters expands the addressable footprint for Broadcom’s merchant silicon relative to proprietary interconnect alternatives. Charlie Kawwas reinforced this direction: “Ethernet is the scale-out of choice” and the “right answer is Ethernet” for scale-up based on current customer asks, though timing and breadth of adoption remain uncertain and could be influenced by competing ecosystems. Third, a rising networking mix could create model uncertainty around gross margin if networking margins differ from compute; management attempted to neutralize this risk by rejecting the premise that system-level shipments or AI mix will dilute margins. The clearest on-call margin defense was: “Our gross margin is solidly at the number Kirsten reports. We will not be affected… by more and more AI products going out.” Kirsten Spears added that the mix impact “is actually not going to be substantial at all.”
SUPPLY CHAIN AND VISIBILITY: KEY ENABLER AND EMBEDDED DOWNSIDE LEVER
The call placed unusual emphasis on supply assurance as a strategic differentiator. Management stated: “We have fully secured capacity of these components for ’26 through ’28,” explicitly referencing leading-edge wafers, high-bandwidth memory, and substrates. In Q&A, supply security was attributed to early action and strategic partnerships, including early lockups of “T-Glass” and substrates. Charlie Kawwas added that deep, multi-year engagements with 6 customers provide demand visibility “over the next 2 to 3 years, sometimes 4 years,” enabling multi-year capacity procurement and technology co-development with suppliers.
This supply posture lowers the probability that near-term AI growth is constrained by ecosystem shortages, a common limiting factor across the AI hardware value chain. The same posture increases exposure to contractual commitments, prepayments, or take-or-pay structures that can become adverse in a demand-down scenario. The inventory build and higher days-on-hand are consistent with an intentional strategy to prioritize revenue capture over working-capital minimization. Monitoring future changes in inventory levels, prepayment balances, and customer lead-time behavior will be important for assessing whether supply security is being achieved efficiently or at the cost of latent downside leverage.
INFRASTRUCTURE SOFTWARE: VMWARE EXECUTION AND AI POSITIONING
Infrastructure Software revenue was $6.8 billion in Q1 (+1% y/y) and guided to $7.2 billion in Q2 (+9% y/y), representing 35% of consolidated revenue in Q1. Segment gross margin was 93% and operating margin was 78% (+190 bps y/y), sustaining the consolidated margin profile and providing cash flow stability that can fund semiconductor growth and capital returns. Within the segment, VMware revenue grew 13% y/y and bookings were described as strong, with total contract value booked in Q1 exceeding $9.2 billion and ARR growth sustained at 19% y/y.
Management’s strategic claim was that Infrastructure Software is structurally complementary to AI rather than threatened by it. VMware Cloud Foundation was described as “the essential software layer in data centers” and management asserted: “VCF cannot be disintermediated or replaced” and “the growth in generative and Agentic AI will create the need for more VMware, not less.” The implied strategy is to position VCF as an abstraction and orchestration layer that integrates heterogeneous compute (CPUs, GPUs, XPUs), storage, and networking into a private-cloud operating environment, which could benefit from AI-driven data-center complexity and enterprise preference for control, compliance, and predictable cost.
The key diligence question is whether VMware’s ARR and bookings momentum reflects sustainable product value and platform entrenchment, or whether it is partially driven by near-term contracting dynamics, pricing actions, and license transitions that may normalize over time. The call did not provide detailed churn or renewal metrics, nor did it disclose the contribution of other Infrastructure Software assets, limiting visibility into whether VMware’s growth is offsetting declines elsewhere or driving the entire segment’s forward growth.
GUIDANCE, MODELING VARIABLES, AND CHANGES VS PRIOR POSITIONING
Q2 guidance implies a material acceleration in consolidated growth with stable profitability. Key guideposts included: consolidated revenue of approximately $22.0 billion (+47% y/y), Semiconductor Solutions revenue of approximately $14.8 billion (+76% y/y) including AI semiconductor revenue of $10.7 billion (+140% y/y), Infrastructure Software revenue of approximately $7.2 billion (+9% y/y), consolidated gross margin “flat sequentially at 77%,” adjusted EBITDA of approximately 68%, and a non-GAAP tax rate of approximately 16.5%. The guided share count of approximately 4.94 billion, excluding potential repurchases, implies that incremental buybacks would be additive to EPS.
The call provided limited numeric comparatives to prior-period guidance beyond the explicit statement that Q1 exceeded guidance and the disclosed EBITDA margin beat of 1 point (68% actual versus 67% guided). A notable qualitative change in positioning appeared in the gross margin discussion around potential rack shipments. Kirsten Spears referenced a change in assessment “relative to even comments that I did make last quarter,” concluding that the impact “is actually not going to be substantial at all.” This supports near-term margin confidence but increases the importance of verifying whether rack/system revenue is immaterial, structurally higher margin than assumed, or being priced to preserve blended margin.
The largest forward-looking variable introduced on the call was the 2027 AI revenue ambition. The statement “line of sight… in excess of $100 billion in 2027” represents a significant departure from typical 1-year guidance horizons and functions as an attempt to anchor investor expectations on a multi-year AI buildout. The implied scaling requirement is substantial: achieving >$100 billion of annual AI chip revenue implies an average quarterly AI chip run-rate >$25 billion in 2027, versus the Q2 2026 guided run-rate of $10.7 billion. The gap is large but potentially achievable if gigawatt deployments scale as indicated and if Broadcom’s silicon content per gigawatt remains robust; however, the acknowledged variability in dollars per gigawatt, limited customer-by-customer dollar disclosure, and the potential for competitive dynamics to influence share leave the >$100 billion figure as a high-impact, high-uncertainty anchor.
TECHNICAL AND COMPETITIVE SIGNALS FROM Q&A
Several technical themes in Q&A have direct implications for long-term share and ecosystem positioning.
Inference-driven demand and monetization: Management highlighted increasing demand not only for training but also for inference, describing it as “very, very interesting and surprising” and noting that “inference is driving a substantial amount of compute capacity.” This message attempts to rebut the concern that AI CapEx could stall pending ROI proof, by suggesting inference workloads tied to productization and monetization are already material and growing.
Workload specialization and the XPU roadmap: In response to questions about evolving architectures, management argued that a general-purpose GPU approach is increasingly inefficient for emerging workloads and that XPUs can be optimized for specific model types and phases (training, inference, prefill, decode, reinforcement learning, test-time scaling). The framing was that customers will increasingly develop specialized training and inference chips in parallel to shorten time-to-monetization and avoid being leapfrogged, which supports sustained multi-year silicon demand but increases R&D intensity and program complexity.
COT competitive threat: The company dismissed near-term share loss to COT initiatives, emphasizing execution and time-to-market as core differentiators and stating that competition “will come eventually” but is “still a long way off.” This is directionally consistent with Broadcom’s historical position in custom ASICs, but the speed of innovation and the scale of capital allocated by hyperscalers could compress timelines relative to prior cycles.
Ethernet standardization: Charlie Kawwas reinforced Ethernet’s primacy and suggested that the ecosystem is converging on Ethernet not only for scale-out but also for scale-up. If Ethernet scale-up adoption accelerates, Broadcom’s networking silicon could capture a larger portion of the AI cluster value chain while reducing the moat of proprietary GPU interconnect ecosystems.
INVESTMENT IMPLICATIONS
The call increased the explicitness of Broadcom’s AI growth thesis and attempted to compress the debate from “whether” to “how much” and “how durable.” The evidence presented supports a constructive near-term interpretation on revenue acceleration and margin durability, but it also increases exposure to expectation risk given the size and specificity of the 2027 ambition.
Key positive implications supported by disclosed data and management commentary:
•AI revenue scale and acceleration: The Q2 guide implies AI semiconductor revenue of $10.7 billion (+140% y/y), with AI approaching 50% of consolidated revenue. The implied sequential growth in AI of +27.4% is large for Broadcom’s scale and indicates near-term shipment visibility.
•Networking as a second growth engine: AI networking mix is expected to rise from approximately 33% of AI revenue in Q1 to 40% in Q2, implying networking revenue acceleration that may be partially decoupled from custom accelerator cycles and that can benefit both XPU and GPU deployments.
•Multi-year visibility narrative: Customer-specific disclosures, gigawatt framing, and the assertion of supply secured through ’28 collectively aim to de-risk the near-term growth path and suggest a higher confidence level than is typical for early-cycle hardware ramps.
•Margin resilience: Gross margin guidance held at 77% despite rapid AI mix shift, and EBITDA margin held at 68%. Management rejected concerns that racks or AI mix will dilute margins, implying either limited system revenue or pricing/cost structure that preserves profitability.
•Software cash flow and potential AI tailwind: VMware’s disclosed bookings and ARR growth provide recurring, high-margin cash flows that can stabilize consolidated results and fund AI investments, while management positions VCF as structurally required for AI-enabled private clouds.
Key negative implications and risk factors surfaced or implied by the call:
•Expectation risk around 2027: The >$100 billion AI chips ambition is a high-impact anchor that could increase stock sensitivity to any sign of customer deployment delays, competitive share shifts, or supply-chain hiccups. The implied requirement of >$25 billion quarterly AI chip revenue in 2027 versus $10.7 billion guided for Q2 2026 highlights the scale of the execution challenge.
•Customer concentration and program risk: The AI thesis is concentrated in 6 customers. Any single program delay, architecture change, or internalization effort can have outsized impact, and customer disclosure increases the market’s ability to track and front-run program progress.
•Limited transparency on racks/system content: Management declined to quantify chip versus rack content for large deployments and pushed back on margin concerns without providing granular breakdowns. This leaves uncertainty around revenue composition, gross margin by product, and potential working-capital implications of system-level offerings.
•Working-capital and supply commitments: Inventory days increased materially, consistent with supply assurance. If demand timing changes, inventory and contractual commitments can become a drag on cash flow and margin, particularly in a rapidly evolving AI hardware environment.
•Competitive dynamics in compute and networking: NVIDIA was framed as an aggressive and improving competitor. Merchant networking faces competition from other switch silicon providers and from vertically integrated GPU/network stacks. The pace of Ethernet scale-up adoption and the timing of optical transitions (including CPO) remain open variables.
•Software durability: VMware growth and bookings are strong, but sustainability of ARR growth depends on renewal behavior and customer satisfaction with licensing models; the call provided limited detail on churn, net retention, or customer sentiment.
Near-term indicators likely to drive the stock over the next 2 quarters:
•Q2 delivery against the $22.0 billion revenue and $10.7 billion AI semiconductor guide, with particular attention to whether networking reaches the 40% AI mix target.
•Commentary that validates or refines the 2027 >$100 billion AI chips line-of-sight, including updates on gigawatt deployment trajectories and any incremental customer disclosures.
•Evidence that gross margin remains stable at ~77% as AI volumes accelerate, including any increased disclosure on racks/system revenue and associated margin profile.
•Inventory trajectory and supply-chain commentary, including whether the company continues to build inventory or begins to normalize days-on-hand as capacity lockups convert into shipments.
•VMware bookings and ARR growth durability, including any signals on renewals, enterprise adoption of VCF, and the linkage between AI workloads and private-cloud demand.
•Hock E. Tan, President and Chief Executive Officer
•Kirsten Spears, Chief Financial Officer and Chief Accounting Officer
•Charlie Kawwas, President, Semiconductor Solutions Group
•Ram Velaga, President, Infrastructure Software Group
•Ji Yoo, Head of Investor Relations
