TITLE: Broadcom’s $100B Moment Is 11 Weeks Away
SUBTITLE: The custom silicon king reports September 2. Two engines are running. One number decides everything.
BODY (HTML):
The GPU war gets all the headlines. The custom silicon war is where Broadcom (NASDAQ: AVGO) is quietly making its most consequential moves.
While investors spend August debating general-purpose GPU allocation and inference token economics, Broadcom is running a fundamentally different business model. It designs purpose-built accelerators, called XPUs, for the world’s largest AI spenders and collects on both the compute layer and the network fabric that connects it. September 2 is when the market finds out whether that model is delivering on its most ambitious promise yet.
Market Temperature
The macro backdrop heading into September is complicated but not hostile. The Federal Reserve held rates at 3.50% to 3.75% in July, and its Jackson Hole conference next week is drawing close attention from bond traders anxious about the path of the September 15-16 meeting. Inflation has cooled from May’s 4.2% spike, but the Fed’s 2% target remains elusive.
For hardware-intensive AI companies, this creates a specific tension. Capital expenditure from hyperscalers is accelerating. Alphabet raised its 2026 capex guide to $195 to $205 billion from $180 to $190 billion. That spending must flow somewhere. A meaningful share flows to Broadcom.
Healthcare and financial stocks have led the August rotation, giving some air cover to AI infrastructure names that sold off through July. Broadcom fell roughly 12.6% in a single session after its June 3 Q2 report, even after beating revenue and EPS estimates. The market wanted a higher $100 billion target. It did not get one. The stock has since recovered and sat near $392 heading into this week.
Company Introduction
Broadcom is two businesses operating under one ticker. The first is a custom semiconductor company that designs XPUs, Ethernet switches, and co-packaged optics for hyperscale AI clusters. The second is an enterprise software company anchored by VMware, which runs the private cloud infrastructure of most of the Fortune 500.
CEO Hock Tan calls it a two-engine model. What makes it strategically unusual is that both engines reinforce each other. VMware Cloud Foundation generates the recurring software cash flow that funds the XPU roadmap without straining the balance sheet. The XPU wins deepen relationships with the same hyperscalers buying VMware subscriptions. The attachment rate compounds in both directions.
This is not a story about riding one wave. It is a story about building infrastructure that hyperscalers cannot easily walk away from.
Data-Driven Deep Dive
The numbers do not require editorial embellishment. In Q2 fiscal 2026, Broadcom reported revenue of $22.2 billion, up 48% year-over-year, with GAAP net income rising 88% to $9.3 billion. Adjusted EBITDA reached $15.2 billion, representing a 69% margin, above guidance and up 52% from a year earlier. Cash from operations was $10.5 billion for the quarter alone.
The AI semiconductor line is the one analysts are watching most closely. In Q1 fiscal 2026, AI revenue hit $8.4 billion, up 106% year-over-year. In Q2, it reached $10.8 billion, a 143% increase. For Q3, management guided AI semiconductor revenue to $16 billion, which would represent over 200% growth from the year-ago period. That is the number September 2 must deliver.
- Q3 FY2026 revenue guidance: approximately $29.4 billion, up roughly 84% year-over-year
- Q3 AI chip revenue target: $16 billion
- Fiscal 2027 AI revenue target: in excess of $100 billion
- Q2 AI bookings: over $30 billion, extending revenue visibility into 2028
- VMware Infrastructure Software: management guided Q3 infrastructure software revenue to approximately $8.9 billion
- Free cash flow, Q2: approximately $10.3 billion
The customer roster is deep and expanding. Broadcom has said it is co-designing custom AI accelerators and AI networking with its hyperscale partners, but it does not publicly provide a definitive customer list in earnings releases. Public disclosures and reporting have linked Broadcom’s custom silicon work to major AI and cloud buyers including Google and Meta, and Broadcom has also discussed work tied to leading LLM developers, including Anthropic.
The networking layer deserves equal attention. Broadcom’s Ethernet switching business, anchored by the Tomahawk 6 switch, which Broadcom has said is shipping in production volume at 102.4 terabits per second, and its Jericho fabric products, accounted for about 40% of Q2 AI semiconductor revenue. This means Broadcom captures revenue from both GPU and XPU clusters. It is paid regardless of which accelerator wins the compute war.
The VMware side of the business is further along than most investors acknowledge. In Broadcom’s fiscal Q4 2025 commentary, management cited an infrastructure software operating margin of 78%. Since the acquisition, VMware by Broadcom has moved to subscription licensing and ended the sale of perpetual licenses. Analysts continue to debate how quickly subscription conversion and customer churn settle into a steady state.
Strategic Insight
The framing that matters is not GPU versus XPU. It is generalist versus bespoke. General-purpose GPUs, which one prominent competitor dominates, are optimized for flexibility. Custom XPUs are optimized for a specific workload at a specific power budget for a specific customer. As AI training and inference workloads mature and become more predictable, the economic case for bespoke silicon strengthens.
Broadcom has reported that AI revenue in all of fiscal 2024 was $12.2 billion. It is now on track to surpass that figure in a single quarter.
The $100 billion fiscal 2027 target is not a marketing number. Management has said it has line of sight to achieve AI chip revenue in excess of $100 billion in 2027, supported by multi-year customer demand visibility extending into 2028. The combination of compute silicon and networking infrastructure creates what may be the most defensible attachment rate in enterprise technology: every new XPU cluster a customer deploys also requires networking to connect it.
The June selloff was a sentiment event, not a fundamental one. Tan held the $100 billion target rather than raising it, and the market punished the stock despite a Q2 beat. That disconnect is worth examining. Broadcom has described an AI order book of roughly $73 billion to be delivered over the next 18 months. That level of near-term visibility is rare in the semiconductor industry.
Risks
Customer concentration is the sharpest risk on the list. A significant portion of Broadcom’s AI semiconductor revenue flows from a small number of hyperscalers. If Google or Meta reduces AI capex, or if one of Broadcom’s other large customers shifts strategy, the revenue impact would be immediate and visible. The June selloff, which was triggered by investor anxiety about a single guidance line, shows how sensitive the stock can be to even the perception of a slowdown.
Supply chain constraints represent a second layer of risk. Broadcom has flagged foundry capacity at advanced nodes as a potential bottleneck for custom accelerator production. Advanced packaging and co-packaged optics are similarly constrained across the industry. Revenue visibility is high, but delivery execution is not immune to manufacturing delays.
The VMware transition carries its own friction. Broadcom has concentrated its focus on the top 2,000 global accounts, and aggressive subscription pricing has prompted some customers to evaluate alternatives. Churn in the enterprise base would weaken the recurring revenue profile that funds the AI hardware roadmap.
Geopolitical risk is persistent. Trade investigations and shifting export controls could complicate future deliveries and demand patterns, particularly for legacy semiconductor lines and for customers with international deployment footprints.
Big Picture
The AI infrastructure buildout is entering a more selective phase. Hyperscalers are no longer buying compute indiscriminately. They are optimizing for power efficiency, total cost of ownership, and workload specificity. That is precisely the environment where custom silicon wins.
Big Tech is expected to spend heavily on AI infrastructure in 2026, and the routing of that spend is becoming increasingly deliberate. Broadcom is embedded in commitments from several of the largest spenders. The remaining question is not whether demand is real. It is whether Broadcom can manufacture and deliver at the pace its own backlog demands.
The VMware software engine adds a dimension that purely hardware-focused AI plays cannot match. As the subscription model matures, the software segment’s recurring cash flow could accelerate, giving management additional capital to deploy into the next generation of XPU co-design agreements.
Broadcom currently holds a Strong Buy consensus from Wall Street analysts, but the exact count of ratings and the average 12-month price target varies by data provider and day. Valuation multiples also move quickly with price and forward estimates.
Final Thought
Broadcom has earned its reputation as the quieter beneficiary of the AI capital expenditure wave. It does not manufacture the most talked-about chip. It designs the chip your cloud provider cannot build without it. Then it sells that provider the software to run everything else.
September 2 is the moment of reckoning. The $16 billion Q3 AI revenue target, an 84% year-over-year surge in total revenue, and the $100 billion fiscal 2027 ambition will either be validated or complicated in a single earnings session. The June selloff showed that the market expects perfection at this valuation. That is a risk worth sizing carefully.
But investors who have been waiting for a cleaner entry since June’s post-earnings flush may be watching the window close. The backlog is growing. The customer list is expanding. The VMware flywheel is accelerating. And the September 2 report could be the data point that finally closes the gap between what Broadcom’s fundamentals say and what its stock price currently reflects.
AVGO is worth a close look, not as a momentum trade, but as a structurally differentiated AI infrastructure franchise reporting its most consequential quarter in company history.
Subject Line: AVGO’s $16B Quarter Is 11 Days Away
Preheader: Custom silicon, VMware’s flywheel, and the $100B target that either gets confirmed or complicated on September 2.
Meta Description: Broadcom reports Q3 fiscal 2026 earnings on September 2. With a $16 billion AI revenue target, a $100 billion fiscal 2027 commitment, and a two-engine model that pairs custom XPUs with VMware software, here is what investors need to know before the number drops.
Disclaimer: This editorial is for informational purposes only and does not constitute investment advice. The analysis presented reflects publicly available information and does not represent a buy or sell recommendation. Investing in equities involves risk, including the possible loss of principal. Past performance is not indicative of future results. Always consult a qualified financial professional before making investment decisions.
