Broadcom is converting AI demand into record revenue, expanding margins, and massive free cash flow. The stock remains attractive on growth, but valuation is already demanding.
Broadcom(AVGO) is one of the clearest examples of a company that has turned AI demand into real revenue, real margins, and real cash flow rather than just a good story. In fiscal Q2 2026, revenue reached a record $22.187B, up 48% YoY, while non-GAAP diluted EPS was $2.44 and free cash flow was $10.262B. The engine is obvious: AI semiconductor revenue hit $10.8B in the quarter, up 143% YoY, and management guided that figure to $16.0B in fiscal Q3, up over 200% YoY.
The investment case rests on a simple but powerful mix. First, Broadcom has a deep position in custom AI accelerators and AI networking, where management said bookings for AI semiconductors were over $30B against $10.8B shipped in Q2. Second, the company pairs that growth with a software business that produced $7.2B of Q2 revenue, 93% gross margin, and roughly 79% operating margin. Third, the financial model remains unusually efficient. Broadcom generated $26.914B of free cash flow in FY2025 and $10.262B in Q2 alone, even while carrying a large debt load from acquisitions.
The main risk is that the stock already reflects a lot of success. AVGO trades at 59.9x trailing earnings, 19.8x forward earnings, and 23.9x EV/revenue. Those are rich numbers for any large-cap semiconductor name. Still, the forward P/E, PEG ratio of 0.42, and analyst target of $523.73 indicate that earnings growth is catching up to the valuation faster than the headline trailing multiple implies. For a balanced, moderate-risk investor with a medium-term horizon, Broadcom still looks attractive, but not at any price. The stock earns a Buy, with fair value anchored at $510.
Broadcom(AVGO) is a Palo Alto-based technology company with two operating pillars: Semiconductor Solutions and Infrastructure Software. The company employs 33,000 people and sells into enterprise and data center networking, AI infrastructure, wireless connectivity, broadband access, storage, industrial markets, cybersecurity, mainframe software, and private cloud. That breadth matters because it gives Broadcom multiple ways to monetize the same infrastructure cycle.
The business has become much larger over the last five years. Revenue rose from $27.45B in FY2021 to $63.89B in FY2025. Over the same period, net income increased from $6.74B to $23.13B, while free cash flow climbed from $13.32B to $26.91B. This is not a company chasing growth at the expense of profitability. It is scaling while keeping margins at levels most hardware companies can only admire from a distance.
Broadcom’s current identity is shaped by both product leadership and acquisition discipline. The semiconductor side remains the growth spear, especially in AI accelerators, Ethernet switching, optical interconnect, SerDes, and custom silicon. The software side, strengthened by VMware, gives the company recurring enterprise exposure and very high margins. In FY2025, Semiconductor Solutions generated $36.858B, or 57.7% of revenue, while Infrastructure Software contributed $27.029B, or 42.3%.
Semiconductor Solutions is Broadcom’s larger and faster-moving segment. In fiscal Q2 2026, segment revenue reached a record $15.0B, up 79% YoY, and represented 68% of total revenue. Segment gross margin was approximately 70%, and operating margin was 62%, up 460 basis points YoY. The growth driver was AI, not a broad cyclical rebound alone.
Within semiconductors, AI revenue was the centerpiece. Broadcom reported AI semiconductor revenue of $10.8B in Q2, up 143% YoY and equal to 49% of total company revenue. Non-AI semiconductor revenue was $4.2B, up 6% YoY, with management saying broadband, server storage, and enterprise networking were up, partly offset by seasonal wireless weakness. That split matters because it shows Broadcom is not just riding one wave. AI is the rocket booster, but the legacy franchise is recovering too.
Infrastructure Software is the stabilizer and margin machine. In Q2, software revenue was $7.2B, up 9% YoY and 32% of total revenue. Gross margin was 93%, operating expenses were $1.0B, and operating margin rose 310 basis points YoY to about 79%. ARR grew 17% YoY, and management guided software revenue to $8.9B in Q3, up 31% YoY. That is a serious acceleration for a business many investors still treat like a mature annex.
The segment mix has changed meaningfully over time. In FY2023, Infrastructure Software was 21.3% of revenue. By FY2024 it was 41.6%, and by FY2025 it was 42.3%. That shift gives Broadcom a more balanced model than a pure-play chip company. It also helps explain why Broadcom can absorb some mix pressure in gross margin while still preserving elite operating margins.
Get AI research on any stock
Instant reports, daily intelligence, and an AI analyst in your pocket.
Broadcom’s flagship growth products sit inside AI infrastructure. The most important are custom AI accelerators, AI networking silicon, and the surrounding connectivity stack. Management said networking represented almost 40% of Q2 AI revenue, though Hock Tan added that a more normal share would be closer to 30% over time. Either way, Broadcom is not just selling the engine. It is also selling the roads, bridges, and traffic system.
On the networking side, Broadcom highlighted Tomahawk 6 as the industry’s only 100-terabit Ethernet switch that has been shipping for over a year, and said its next-generation 200-terabit switch would tape out in the quarter. The company also described itself as the de facto standard in co-packaged optics with 1.6-terabit DSPs, CW, and EML lasers, and said Jericho3 and Jericho4 fabric solutions are enabling some of the world’s largest hyperscaler deployments.
On the software side, VMware Cloud Foundation 9.1 is the flagship platform. Management said the release focuses on infrastructure efficiency, security, and support for enterprise AI inferencing workloads. It adds heterogeneous compute support across AMD, Intel, and NVIDIA platforms, allowing customers to run AI, Kubernetes, and traditional virtualized workloads on a common private cloud environment. That is corporate language for one very practical point: Broadcom wants VMware to remain the operating layer for enterprise infrastructure even as AI changes the workload mix.
Broadcom’s moat starts with design depth. Custom AI silicon is not an off-the-shelf business. Once a hyperscaler commits to a multi-generation roadmap, the switching costs become high because the chip, networking, software stack, and deployment plan are all tied together. Management underscored that with Google, saying Broadcom entered a long-term agreement in April to develop and supply multiple generations of TPUs and AI networking.
The second moat is networking leadership. Hock Tan said Broadcom has at least one generation of technology and product leadership in networking. That claim is supported by named products and deployment status: Tomahawk 6 in 100-terabit Ethernet, a 200-terabit tape-out underway, Jericho fabric solutions in hyperscaler deployments, and leadership in co-packaged optics and high-speed SerDes. In AI clusters, compute gets the headlines, but networking often decides whether the system scales cleanly or turns into an expensive traffic jam.
The third moat is execution speed. Broadcom and OpenAI unveiled Jalapeño, described as OpenAI’s first Intelligence Processor, and the collaboration moved from design to tape-out in nine months. That kind of cycle time matters because the AI infrastructure race is not patient. Customers are not buying abstract roadmaps. They are buying delivered silicon.
Finally, Broadcom’s software business adds a different kind of moat: installed-base lock-in. VMware, mainframe software, cybersecurity, and enterprise automation products sit close to mission-critical systems. In Q2, software gross margin was 93%. Businesses do not produce that kind of margin unless customers find replacement painful, risky, or both.
Broadcom runs a fabless model and depends on contract manufacturing and a limited supplier base, which creates exposure to supply constraints and geopolitics. That risk is real, but management sounded confident on supply. On the Q2 call, Hock Tan said Broadcom is comfortable it has secured the needed supply for 2026 and 2027 and is working on 2028 and 2029. He also said customers have been coming incrementally for more supply over the last few months and that, by and large, Broadcom can support that demand.
Inventory trends show the company is preparing for that ramp. Broadcom ended Q2 with inventory of $4.3B, and days of inventory on hand rose to 86 from 68 in Q1. Management said this was done in anticipation of accelerating AI semiconductor growth in the second half of the year. That is a deliberate build, not a sign of weak sell-through.
Capital intensity remains low relative to revenue scale. Q2 capex was $231M against $22.187B of revenue, and FY2025 capex was just $623M against $63.887B of revenue. That is one reason Broadcom converts so much revenue into free cash flow. The company is in the chip business, not the foundry business, and the distinction is worth billions.
Broadcom operates in a semiconductor market that several industry forecasts place between roughly $668B and $740B in 2025 to 2026, with long-term growth around mid-single digits. That broad market view matters less than Broadcom’s specific slice of the market, which is AI accelerators, AI networking, optical interconnect, and custom silicon. That slice is growing much faster than the industry average.
Broadcom’s own numbers show the difference. AI semiconductor revenue grew from $8.4B in Q1 FY2026 to $10.8B in Q2, and management guided $16.0B for Q3. For full-year FY2026, management reiterated AI semiconductor revenue of $56B, up about 180% from FY2025, and said FY2027 AI semiconductor revenue should exceed $100B. Those are not market-share dreams. They are explicit company targets tied to named customer programs.
The software market is also evolving in Broadcom’s favor. Enterprises are trying to run AI workloads inside private cloud environments for cost control, data governance, and security. VMware Cloud Foundation 9.1 is aimed directly at that need. Broadcom is effectively selling both the picks and the plumbing: chips for hyperscalers and software for enterprises that want AI without handing the whole stack to public cloud providers.
Like what you're reading?
Get full access to AI-powered research reports, market analysis, and portfolio tools.
Broadcom’s customer base is concentrated at the high end of infrastructure spending. On the semiconductor side, management identified six core AI customers and specifically discussed Google, Anthropic, OpenAI, and Meta. These are not casual buyers. They are among the largest and most technically demanding infrastructure customers in the world.
The customer relationships are increasingly multi-year and capacity-based. Broadcom said it has a long-term agreement with Google for multiple generations of TPUs and AI networking. For Anthropic, Broadcom said it is providing over 1 gigawatt of TPU-based compute in 2026 and signed an agreement in April for another 5 gigawatts of next-generation TPU-based compute beginning in 2027. For OpenAI, Broadcom said it has delivered silicon, is on track for production in late 2026, and has a contractual commitment to deploy 1.3 gigawatts in 2027 as part of a larger 10-gigawatt agreement by 2029. For Meta, Broadcom announced a partnership to deliver multiple generations of MTIA XPUs and expects to deploy 3 gigawatts through the end of 2028.
That concentration creates risk, but it also creates visibility. In Q2, bookings for AI semiconductors were over $30B, and Broadcom said it had received purchase orders totaling $6B from two additional customers whose shipments are expected to begin late 2026 and accelerate into 2027. In software, the customer profile is large enterprise, telco, and data center operators that rely on virtualization, private cloud, security, and mainframe tools. Those customers tend to be sticky, budgeted, and slow to rip out core infrastructure.
Broadcom competes across several markets rather than one. In AI networking and data-center interconnect, key rivals include Nvidia(NVDA), Marvell(MRVL), Cisco(CSCO), Intel(INTC), and smaller connectivity specialists. In custom silicon and ASICs, Nvidia, Intel, and Marvell are the most relevant competitors. In infrastructure software, the competition shifts toward Microsoft(MSFT), IBM(IBM)/Red Hat, Nutanix(NTNX), Oracle(ORCL), and public-cloud platforms.
Broadcom’s edge is focus. It is not trying to be everything in semiconductors. It targets niches where performance, integration, and switching costs matter more than commodity pricing. That is why management keeps emphasizing custom accelerators, Ethernet switching, optics, SerDes, and fabric solutions. These are not glamorous in the consumer sense, but they are mission-critical in the data center sense, which is where the money is.
The software side adds another layer of differentiation. VMware gives Broadcom a large installed base in virtualization and private cloud. That does not make the company immune to competition, but it does mean Broadcom starts from a position of embedded relevance. In infrastructure, being deeply embedded usually beats being loudly interesting.
Broadcom sits at the intersection of two major macro forces: the AI capex boom and the geopolitics of semiconductor supply chains. The first is a tailwind. Industry sources cited in the market context show AI, HBM, networking, and advanced packaging as the main growth engines in semiconductors. Broadcom’s Q2 results line up with that trend almost perfectly.
The second force is more complicated. Broadcom relies on contract manufacturing and a limited supplier base, which exposes it to export controls, trade restrictions, and regional disruptions. The company’s own risk framing highlights supply chain dependence, trade risk, and regulatory exposure. This is the cost of operating in the most strategic part of the global hardware stack.
There is also a cyclical macro risk. AI spending is strong now, but it is still capex. If hyperscalers slow deployment schedules or rework architectures, Broadcom’s AI growth could decelerate sharply. That risk is partly balanced by the company’s software cash flow and by non-AI semiconductor recovery, but it cannot be ignored. When a business is growing this fast, even a slowdown to merely very good can feel harsh in the stock.
Broadcom’s A- balance sheet reflects a large debt load, but $10.262B of Q2 free cash flow and $26.914B in FY2025 cash generation keep leverage manageable.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →Revenue hit a record $22.187B in fiscal Q2 2026, with AI semiconductor sales up 143% YoY to $10.8B and software margins still running near 93% gross.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →Management guided AI semiconductor revenue to $16.0B in fiscal Q3, implying more than 200% YoY growth, while software revenue is expected to reach $8.9B.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →AVGO trades at 59.9x trailing earnings and 19.8x forward earnings, but a PEG ratio of 0.42 suggests growth is starting to catch up to the multiple.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →The report’s fair value anchor is $510, with upside to $430 for a Buy call and $360 for a Strong Buy threshold.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →Broadcom(AVGO) is one of the rare large-cap technology companies where the bullish case does not need much decoration. Q2 FY2026 revenue was $22.187B. AI semiconductor revenue was $10.8B. Free cash flow was $10.262B. Q3 revenue guidance is $29.4B. FY2026 AI semiconductor revenue guidance is $56B, and FY2027 is above $100B. Those are hard numbers, not mood music.
The company’s strength comes from its structure. Semiconductor Solutions gives Broadcom direct exposure to the fastest part of infrastructure spending. Infrastructure Software gives it recurring revenue, high margins, and enterprise stickiness. The result is a business that can grow fast without looking financially reckless. That combination is rare, and the market knows it.
For medium-term investors, the key is price discipline. Broadcom deserves a premium, but premiums can still be overpaid. With a fair value estimate of $510, the stock remains attractive below that level and increasingly demanding far above it. In plain English, this is a high-quality AI compounder, not a blank check.
Get AI-powered research reports, daily market intelligence, and a personal analyst in your pocket.
Get Full Access →Not ready to subscribe? ·
Broadcom's latest Apple agreement gives investors something the AI bubble debate can't dismiss: multi-year, contract-backed demand from one of the world's biggest hardware buyers. That kind of revenue visibility matters even more when it sits on top of triple-digit AI growth and elite margins.
Broadcom Inc. (AVGO) rises sharply as investors react to a reported Apple chip supply agreement worth more than $30 billion through 2031. The move comes despite a weaker broader market, reinforcing Broadcom’s strength in custom silicon, AI infrastructure, and long-term revenue visibility.
Broadcom Inc. (AVGO) rises after Reuters reported an expanded Apple chip partnership through 2031, boosting revenue visibility and reinforcing its custom silicon and AI growth story. The move comes as Broadcom posts strong quarterly results and investors continue to reward long-duration demand.