Broadcom Q3 FY2026 Earnings: AI Revenue More Than Triples as Long-Term Outlook Expands

Broadcom reported record fiscal third-quarter revenue, operating income and free cash flow, led by a 221% increase in AI semiconductor revenue to $16.7B. The company’s near-term outlook also accelerated, with fourth-quarter AI revenue projected at $21.7B and management raising its fiscal 2026 AI revenue outlook to $58B.

The more consequential development was Broadcom’s long-range forecast: management expects AI semiconductor revenue to double to about $115B in fiscal 2027 and double again to $230B in fiscal 2028, subject to supply and data-center deployment constraints. That outlook substantially increases the importance of execution across custom accelerator programs, networking, component supply and customer financing.

As Reported on the Call

The Bigger Story

Broadcom’s quarter reinforces that it is becoming increasingly dependent on custom AI accelerators and the networking infrastructure surrounding them. AI semiconductor revenue represented 56% of total company revenue in the third quarter, up from 49% in the prior quarter, while AI-related revenue accounted for roughly four-fifths of Semiconductor Solutions revenue.

Management’s fiscal 2027 and fiscal 2028 AI forecasts are the key change in the earnings story. Broadcom expects approximately $115B of AI semiconductor revenue in fiscal 2027 and $230B in fiscal 2028, each roughly double the preceding year. Management said it has secured supply to support those targets, though demand exceeds its current supply outlook.

These are management projections rather than contracted revenue disclosures, and their realization depends on more than Broadcom’s ability to ship chips. Management repeatedly cited data-center land, power and building availability, alongside leading-edge wafers, substrates, high-bandwidth memory and system memory, as constraints that can affect the timing of deployments.

Revenue and Operating Performance

Semiconductor Solutions revenue reached a record $20.8B, rising 127% from a year earlier. The principal driver was AI semiconductor revenue of $16.7B, which more than tripled year over year and increased 54% sequentially.

Custom accelerators, or XPUs, accounted for 73% of AI semiconductor revenue during the quarter. Broadcom said XPU shipments increased more than 3.5 times from a year earlier, while AI networking revenue grew more than 2.5 times.

The company highlighted several customer programs that are now moving through production ramps:

• ·Broadcom delivered Google’s Ironwood TPU version 7 in high volume to Google and Anthropic.
• ·It began production shipments of Google’s next-generation TPU version 8i.
• ·It shipped Jalapeno, OpenAI’s first custom accelerator.
• ·It expects production shipments of Meta’s MTIA accelerator in the fourth quarter.

Management made several performance comparisons between its customer-specific accelerators and competing GPU platforms. Those comparisons reflect management’s view and were not independently quantified on the call. The broader economic point is clearer: Broadcom argues that customer-specific silicon can improve performance and cost efficiency for targeted AI workloads, supporting continued demand for custom accelerators.

AI networking is also becoming a more important part of the revenue opportunity. Broadcom said Tomahawk 6 Ethernet switches are deployed across nearly all AI hyperscalers building XPUs with the company and are also used by some customers that do not use Broadcom-designed accelerators. The company also said it taped out Tomahawk 7, a 200-terabit-per-second Ethernet switch, and expects Tomahawk Ultra adoption in scale-up applications to begin this quarter and continue into fiscal 2027.

Outside AI, semiconductor revenue was more stable than transformative. Non-AI semiconductor revenue was $4.2B, up 5% year over year and flat sequentially. Broadband and server-storage demand increased, partially offset by lower wireless revenue. Management expects non-AI semiconductor revenue to rise modestly to approximately $4.3B in the fourth quarter.

Infrastructure software revenue increased 29% year over year to $8.8B, with annual recurring revenue growing 15%. The segment remains a large and high-margin contributor, but management expects revenue to stabilize at about $8.7B in the fourth quarter, implying that near-term company growth will remain overwhelmingly driven by AI semiconductors.

Profitability and Margins

Broadcom generated $20.1B of operating income, up 92% from a year earlier, and reported non-GAAP diluted EPS of $3.32, up 96%. Consolidated operating margin expanded 240 basis points year over year to 67.9%, demonstrating substantial operating leverage as revenue grew faster than operating expenses.

The margin picture is becoming more nuanced as AI chips occupy a larger share of revenue. Consolidated gross margin was 75%, down 210 basis points sequentially, primarily because XPUs contain more memory and carry lower gross margins than parts of Broadcom’s historical mix, particularly infrastructure software.

Management expects this trend to continue in the fourth quarter. It guided to a consolidated gross margin of approximately 73%, down from 78% a year earlier, while expecting operating margin of approximately 66%, flat year over year. The company’s position is that revenue growth and disciplined operating spending can preserve operating margins even if gross margins decline as AI hardware becomes a larger share of sales.

There was one notable disclosure inconsistency on the call. The CFO initially stated that Semiconductor Solutions gross margin was approximately 76% in the third quarter, then later corrected that figure to approximately 67%. Broadcom’s reported consolidated gross margin of 75% was unchanged, but investors should rely on the company’s formal financial statements for the final segment-level figure.

Demand, Customers and Pricing

The demand backdrop remains unusually concentrated. Broadcom identified six XPU customers, with Google, Anthropic, OpenAI and Meta representing the most prominently discussed programs. Management expects Anthropic to become its largest XPU customer in 2027, followed by OpenAI in 2028 based on its current deployment outlook.

Google remains a foundational customer. Broadcom said it recently entered a long-term agreement to develop and supply future TPU generations and associated AI networking, with plans to deliver “multi-tens of billions of dollars” of TPUs annually over the next several years.

Management outlined significant prospective deployment plans for its AI lab customers:

• ·Anthropic is expected to deploy 1 GW of Ironwood in 2026, another 5 GW of TPU version 8i in 2027, and an incremental 10 GW in 2028.
• ·OpenAI is expected to deploy 1.3 GW of Jalapeno in 2027 and more than 5 GW of Jalapeno and successor products in 2028.
• ·Meta is expected to receive three generations of MTIA accelerators through the end of 2027, with management seeing a path to 3 GW of deployments through 2028.

These figures represent management’s stated line of sight to capacity deployments, not revenue recognized in the current quarter. Broadcom also cautioned that not all identified gigawatts will necessarily be installed and producing within the fiscal periods used in its revenue outlook.

Management said Broadcom’s AI semiconductor content is generally expected to remain in the range of $20B to $30B per gigawatt of deployed compute. Faster chips can command higher selling prices, but they also consume more power per unit, reducing the number of accelerators needed for a given gigawatt. As a result, management expects revenue per gigawatt to remain relatively stable while the total number of gigawatts rises.

What is Broadcom’s guidance and outlook?

For fiscal Q4, Broadcom guided for consolidated revenue of $34.8B, up 93% year over year. Semiconductor revenue is expected to reach approximately $26.1B, including $21.7B in AI semiconductor revenue. Infrastructure software revenue is expected to be about $8.7B.

The Q4 AI forecast implies another sharp sequential increase from $16.7B in Q3. Management expects both XPU and AI networking revenue to triple year over year in the quarter, driven by higher shipments of Ironwood to Anthropic, a high-volume ramp of Google’s TPU version 8i, continued OpenAI shipments and the start of production shipments to Meta.

For fiscal 2026, management raised its AI semiconductor revenue outlook to $58B from its previous $56B target. It then provided an unusually long forecast, targeting approximately $115B in fiscal 2027 AI revenue and $230B in fiscal 2028. Management said it does not intend to update those longer-term targets each quarter.

The main limitation is supply and deployment readiness. Broadcom said its outlook incorporates its assessment of customer data-center readiness, including land, power and building availability, as well as the supply chain for leading-edge silicon, substrates and HBM memory. The company will begin using its Singapore facility for substrate production in fiscal 2027 and is expanding optical-component capacity, including EML, CW laser and VCSEL production.

Balance Sheet, Cash Flow and Capital

Free cash flow reached a record $13.7B, equal to 46% of revenue. Broadcom ended the quarter with $24B of cash, up from $19.6B in the prior quarter, and spent $532M on capital expenditures.

Capital spending is set to increase materially. Management expects $1.4B of Q4 CapEx as it invests in semiconductor capacity, including substrates and optical components. This is consistent with management’s view that supply availability remains a limiting factor in meeting AI demand.

Broadcom repaid $5.6B of long-term debt during the quarter and an additional $1.5B of senior notes after quarter-end. Gross principal fixed-rate debt stood at $59.6B, with a weighted-average coupon rate of 4% and weighted-average maturity of 7.4 years.

The company also discussed its AI XPV financing platform with Apollo and Blackstone, designed to support more than 20 GW of compute infrastructure for Anthropic and OpenAI through the end of 2028. The first $35B tranche, related to Anthropic’s 1 GW deployment, closed in June.

Broadcom said third-party financial partners will independently underwrite and capitalize these projects, rather than Broadcom directly financing them. However, management acknowledged that it may provide modest residual-value guarantees where necessary. It did not provide new details on maximum exposure or future guarantees, saying financing terms will be assessed on a deal-by-deal basis.

What Investors Should Watch

1. Q4 AI Revenue Conversion

The immediate test is whether AI semiconductor revenue reaches the guided $21.7B in Q4. Meeting that figure would validate the production ramps at Google, Anthropic, OpenAI and Meta and support the increased fiscal 2026 AI revenue outlook.

2. Data-Center Deployment Readiness

Broadcom’s long-range targets depend not just on chip demand but on customers securing land, power and completed data-center capacity. Any delays in these projects could defer shipments even if customer demand and Broadcom’s supply availability remain strong.

3. Gross-Margin Pressure Versus Operating Leverage

The company expects AI mix and higher memory content to reduce gross margins further in Q4. Investors should watch whether operating margin remains near management’s 66% target as lower-margin hardware becomes a larger portion of revenue.

4. Supply Expansion and Component Constraints

Substrate availability, HBM memory, leading-edge wafers and optical components remain key bottlenecks. Broadcom’s Singapore substrate production and expanded indium-phosphide optical capacity are intended to ease some constraints, but execution on those investments will matter for the fiscal 2027 and 2028 outlook.

5. Customer Financing and Residual-Value Exposure

The AI XPV platform could help strategic customers accelerate deployment, but it also adds a new financing-related consideration. Investors should monitor future disclosures on residual-value guarantees, the scale of Broadcom’s contingent liabilities and the degree to which Anthropic and OpenAI can independently fund infrastructure.

AlgovestIQ Takeaway

Broadcom’s third quarter was stronger across revenue, earnings and cash flow, but the central investment development was the scale of management’s AI outlook rather than the reported quarter alone. AI semiconductor revenue has moved from a major growth contributor to the company’s dominant business driver, and management now expects that business to grow from $58B in fiscal 2026 to approximately $230B by fiscal 2028.

The quarter also showed that Broadcom’s AI opportunity is broader than custom accelerators. Networking, optical interconnects, switching and substrates can all benefit from rising AI cluster scale. The company is investing in manufacturing capacity where management sees industry supply constraints, particularly substrates and optical components.

At the same time, the growth profile carries greater concentration and execution risk. A limited group of customers is expected to drive most of the AI expansion, while the timing of their deployments depends on construction, power availability, financing and component supply. Management’s multi-year forecasts appear to incorporate these considerations, but they remain forecasts tied to a complex deployment ecosystem.

Margins are another issue to monitor. Broadcom is showing strong operating leverage, but its gross margin is declining as memory-intensive XPUs become a larger share of sales. Sustained operating margins near current levels would support management’s argument that scale can offset mix pressure; a sharper decline would challenge that thesis.

The key question is no longer whether Broadcom has meaningful AI exposure. It is whether the company can convert its stated capacity roadmap into shipments while preserving its operating model and limiting balance-sheet exposure from customer infrastructure financing. Evidence of on-time customer deployments, supply expansion and stable operating margins would strengthen the case. Delays in data-center readiness, tighter supply bottlenecks or larger financing commitments would be the principal counterweights.