Broadcom's $230bn 2028 Guidance Is a Capacity Claim

What Broadcom reported for the quarter ended August 2, 2026

Broadcom posted consolidated revenue of $29.6 billion for the fiscal quarter ended August 2, 2026, an increase of 86 percent from the same quarter a year earlier. The company disclosed the figures in an exhibit to its Form 8-K filed with the US Securities and Exchange Commission, and CEO Hock Tan walked through the same numbers on the earnings call that followed.

AI semiconductor revenue was the reason the quarter looked nothing like a typical chip cycle. It reached $16.7 billion, up 221 percent year over year and 54 percent from the prior quarter, and for the first time it made up more than half of Broadcom's entire business at 56 percent of total revenue.

Custom silicon, not merchant chips, is doing the growing

Broadcom's AI semiconductor business is built almost entirely on custom accelerators it co-designs with a small number of hyperscale cloud customers, not on general-purpose chips it sells off a shelf. These are the application-specific processors, often called XPUs, that Google, Meta and other large cloud operators use alongside merchant GPUs to run their own AI workloads at a lower cost per chip.

That distinction matters for how the guidance should be read. A merchant-chip forecast reflects an estimate of open-market demand; a custom-silicon guide reflects work Broadcom's engineering teams are already committed to delivering for named customers on multi-year design and manufacturing cycles. When Hock Tan raises the number, he is describing capacity his largest customers have already reserved, not guessing at a market that might show up.

The number that outranks the beat: $115bn, then $230bn

Broadcom's guidance for the fiscal quarter now underway calls for $21.7 billion in AI semiconductor revenue, up 236 percent year over year, and in an ordinary quarter that alone would have been the headline.

Tan went further, telling investors the company now has "line of sight" to $115 billion in AI semiconductor revenue for fiscal 2027 and $230 billion for fiscal 2028, about EUR 212 billion at current exchange rates, guidance that reaches 18 to 24 months past the quarter Broadcom had just closed.

Why guidance this far out functions as a capacity claim

A revenue guide issued 18 to 24 months in advance is not really a forecast of demand; it is a statement about how much foundry and advanced-packaging output Broadcom's hyperscaler customers have already reserved on Broadcom's behalf.

Custom silicon at this scale requires committed wafer and packaging capacity booked years ahead, so a $115bn and $230bn figure is Broadcom telling the market its customers have effectively locked that much of a shared, constrained pipeline. Every other buyer drawing on the same foundry and advanced-packaging pool, including EU cloud resellers and enterprises assembling their own inference clusters, is now competing for what is left over after those reservations clear.

What this means for procurement timelines and pricing leverage in 2027 and 2028

A mid-size EU enterprise or cloud reseller trying to secure AI compute capacity for 2027 or 2028 is now negotiating against demand Broadcom has effectively pre-sold, which changes both the timeline and the leverage on any deal signed today.

The practical shift is that a capacity commitment now needs to be treated like a multi-year reservation, not a spot purchase: lock in allocation and pricing as early as possible, because the pool available a year from now will be smaller and pricier than the pool available today. Expect quoted lead times to lengthen and expect vendors to start pricing scarcity into contracts rather than just hardware cost.

For a buyer without Broadcom-scale leverage, the sensible response is to diversify supply relationships now, move capacity budgeting and sign-off cycles forward by two to three quarters, and treat any 2027-2028 delivery slot offered today as worth securing even before the full internal business case is finished, because the alternative is bidding for what is left once the hyperscalers named in this guidance have already been served.

Servola Journal

We do this for everyone trying to keep up with what technology is doing to our lives. The people who build it, and the people it happens to. The Servola Journal exists so that what we learn belongs to all of them.

Nobody pays us for this. No ads, no paywall, free to everyone. We just believe that understanding what's happening to all of us shouldn't depend on who can afford to pay for it.

If it gave you something today, tell us to keep going. Follow us, leave a like, or write a positive comment. We read every one, and they are what keeps us going.

Read next: Nvidia Turns Hugging Face Into a Chokepoint  |  Broadcom's AI Debt Bet Triples to $100 Billion