Broadcom 2027–2028 Gigawatt Roadmap Shows $11.5B per GW and Steady Deployed Content

Broadcom’s 2027–2028 Gigawatt Projections and Implied Content

1. Gigawatt projections

The approximately 10 GW for fiscal 2027 and 20 GW for fiscal 2028 figures are directionally correct, but they represent the customers’ expected deployment roadmaps and demand visibility—not necessarily gigawatts that will be fully installed and in production during those fiscal years. Management explicitly said the customer roadmaps total roughly 30 GW across 2027–2028, while cautioning that some portion may not be deployed quickly enough for Broadcom to ship against it during that period. ¹

The figures are also consistent with the more detailed customer disclosures: Anthropic represents 5 GW in 2027 and 10 GW in 2028, while OpenAI represents 1.3 GW in 2027 and more than 5 GW in 2028. ³ Meta contributes a further 3 GW through 2028, although the filing does not provide the complete Google, Meta, or other-customer breakdown needed to independently reconstruct the full approximately 10 GW and 20 GW totals. ⁴

2. Revenue implied by the roadmap

Broadcom is guiding to approximately $115 billion of AI semiconductor revenue in fiscal 2027 and $230 billion in fiscal 2028, or approximately $350 billion cumulatively across the two years. ⁵⁴

Using the customer-roadmap denominators of 10 GW and 20 GW produces the following arithmetic:

Thus, the question’s estimate of roughly $11 billion–$12 billion per gigawatt is mathematically reasonable if revenue is divided by the full roadmap gigawatts. ²⁵

However, this should not be interpreted as Broadcom’s expected steady-state dollar content per gigawatt. Management specifically cautioned that the approximately 30 GW is not equivalent to 30 GW fully deployed into production during 2027–2028, because data-center shells and other infrastructure must be ready before chips and racks can be shipped. ¹ Consequently, the $11.5 billion figure is best understood as AI semiconductor revenue divided by the broader customer roadmap, not necessarily the revenue attached to each gigawatt actually deployed in that year.

3. Management’s stated content level

Management separately characterized Broadcom’s content at approximately $20 billion–$30 billion per deployed gigawatt, and said it expected that range to remain relatively sustainable. ⁶

This creates an important distinction:

• ~$11.5 billion per roadmap GW is the arithmetic result of dividing annual guided revenue by the full customer roadmap. ²⁵
• ~$20 billion–$30 billion per deployed GW is management’s stated estimate of Broadcom’s dollar content associated with a gigawatt of actual deployment. ⁶

The difference is economically meaningful. It suggests that the 10 GW and 20 GW figures should not be treated as fully revenue-bearing shipments in the same fiscal period. Some roadmap capacity may be delayed, staged, or deployed after the relevant fiscal year, even though customer demand is already visible. ¹

4. How content should trend as XPUs evolve

Management’s framework implies a relatively stable dollar content per gigawatt, rather than a steadily rising figure:

1. XPU performance and silicon cost increase. Each successive XPU or GPU becomes more powerful, uses more leading-edge silicon, and is therefore more expensive to manufacture; management said this increases ASPs per XPU. ⁶

2. Power consumption also increases. Higher-performance XPUs consume more power per chip, meaning fewer advanced XPUs fit within one gigawatt of power capacity. ⁶

3. The two effects broadly offset at the gigawatt level. Although the price per XPU rises, the number of XPUs per gigawatt declines, leaving Broadcom’s dollar content per gigawatt “relatively stable” in the stated $20 billion–$30 billion range. ⁶

4. Total opportunity grows through more gigawatts. Management expects the number of gigawatts deployed to accelerate because customers require exponentially more compute for frontier AI workloads, even if the dollars attached to each gigawatt remain broadly stable. ⁶

In simplified form:

Higher price per XPU × fewer XPUs per GW ≈ broadly stable dollars per GW ⁶

This is a more nuanced outlook than assuming that every new XPU generation automatically produces proportionately higher revenue per gigawatt. The expected growth driver is primarily the expansion in total gigawatts, while per-gigawatt content remains broadly within the $20 billion–$30 billion range. ⁶

5. Overall interpretation

The projections can therefore be reconciled as follows:

• The customer roadmap is approximately 10 GW in 2027 and 20 GW in 2028, or approximately 30 GW combined. ²¹
• Broadcom expects approximately $115 billion of AI semiconductor revenue in 2027 and $230 billion in 2028. ⁵⁴
• Dividing those revenue figures by the roadmap gigawatts gives approximately $11.5 billion per roadmap GW. ²⁵
• Management’s more relevant steady-state content estimate is $20 billion–$30 billion per gigawatt actually deployed, because not all roadmap capacity is expected to be operational and revenue-bearing within the two fiscal years. ¹⁶
• As XPUs advance, ASP and power per chip rise together, causing dollar content per gigawatt to remain broadly stable even as the number of required gigawatts expands rapidly. ⁶

The central analytical takeaway is that the $11 billion–$12 billion figure is a valid roadmap-based revenue calculation, but it likely understates Broadcom’s content on an actual deployed-gigawatt basis. The company’s stated long-term framework is closer to $20 billion–$30 billion per deployed gigawatt, with growth driven by a rapidly increasing gigawatt base rather than substantial inflation in content per gigawatt. ¹⁶