.png){kind=logo}
In Friday dealing, Broadcom’s drive to scale custom AI silicon commands attention, with Burghley Capital highlighting management’s ambition to push annual revenue above $100 billion by the end of 2027; the shares trade about 4% higher over the session.
The immediate catalyst is a confirmed five-year development and supply agreement that keeps Broadcom at the centre of Google’s Tensor Processing Unit programme, pairing bespoke processors with the networking components that link AI clusters. The arrangement runs through 2031, signalling that cloud operators increasingly prefer long-horizon procurement for the kit that trains and serves demanding models, even as policymakers tighten their grip on what can be shipped where.
Anthropic indicates plans to secure about 3.5 gigawatts of TPU-based capacity from 2027 onwards, and one Wall Street estimate links the partnership to roughly $23.4 billion of AI revenue over the fiscal year under way, with potential towards about $46.9 billion in the following year.
Broadcom’s latest reported quarter puts numbers behind that narrative. AI-related revenue runs at about $9.4 billion in the latest quarter, up 106% versus the same quarter a year earlier, while the AI order backlog reaches roughly $81.5 billion at quarter-end.
The backlog is where today’s investment story meets tomorrow’s delivery schedule, and James Barker, Director of Private Equity at Burghley Capital Pte. Ltd., describes the figure as “a contract for confidence in a sector that often sells promise, because it forces attention onto what ships over the next several quarters”.
Valuation still matters, because fast-growing narratives can get ahead of themselves. Broadcom trades at about 27 times forward earnings based on analysts’ forecasts for the coming four quarters. In the most recently completed fiscal year, revenue expands 44% to about $57.6 billion, supported by AI semiconductor sales of roughly $13.6 billion over that year, and external projections point to AI revenue around $45.1 billion for the fiscal year under way, about 92% higher than the prior year. Management keeps the bar high with a target for AI chip revenue above $111.6 billion by the end of 2027, and Barker frames the market’s calculus as “a question of whether the pipeline converts on time, because the multiple only holds if the delivery curve stays steep”.
The argument for resilience rests on how far Broadcom can smooth the semiconductor cycle. The $81.5 billion backlog recorded at the latest quarter-end provides visibility across multiple years, and Broadcom generates about $8.9 billion of free cash flow in the latest quarter, equal to 41% of revenue in that period. Leading AI platforms outline capital expenditure plans of roughly $457.7 billion over the next full-year spending cycle, about 80% above the previous cycle, a scale that keeps the supply chain under pressure. Barker argues that “when hyperscalers write bigger cheques for compute, suppliers that can commit capacity into the late 2020s gain leverage, even as export controls and security reviews complicate who gets what”.
Custom silicon is where Broadcom tries to differentiate itself from Nvidia’s general-purpose GPU franchise. Broadcom’s model centres on application-specific integrated circuits that are tuned to particular workloads, a route that tends to lower total cost of ownership once volumes justify the upfront design work, and industry researchers forecast Broadcom holds around 60% of AI server compute ASIC design partnerships through 2027. Research and development spending runs at about $3.3 billion each year. A quarterly dividend of about $0.9 per share rises 11% in the most recent fiscal year, extending a run of annual increases that spans 14 consecutive years. Over the past year, the shares rise 67.2%, and analyst targets for the next year sit around $510.2 a share from a latest close near $354.5, implying upside above 35% over that horizon. Barker describes that gap as “the market’s way of pricing optionality, but it still demands proof that customer concentration does not turn into margin compression”.
What investors watch next is straightforward: whether hyperscaler contracts translate into volume shipments on the timetable implied by the backlog, and whether capacity commitments protect profitability rather than dilute it. For Burghley Capital, the practical test is whether contracted demand and operational discipline keep pace with the politics of chips and the economics of power, and Barker concludes that “this is no longer a purely technological race, because the winners are those who can execute through regulation, supply bottlenecks and the funding cycle”.
Established in 2017, Burghley Capital Pte. Ltd. (UEN: 201731389D) is a Singapore-headquartered investment management firm focused on long-only strategies. It provides market intelligence, tailored investment analysis and financial advisory support for institutional and private clients worldwide, with an emphasis on disciplined risk management and long-term resilience. Additional insights are available at https://burghleycapital.com/resources. Media enquiries may be directed to Martin Wei at m.wei@burghleycapital.com or via https://burghleycapital.com.
