Broadcom secures 60% AI chip market dominance as hyperscalers race to build custom silicon

Broadcom Dominates Custom AI Chips Market With 60% Share

Broadcom is projected to retain its leadership position as an artificial intelligence server compute ASIC design partner with 60% of market share through 2027, according to Counterpoint Research 4

4

. This commanding position comes as hyperscalers including Google, Amazon Web Services, Meta, and Microsoft accelerate their shift away from general-purpose GPUs like Nvidia's offerings toward Application-Specific Integrated Circuit solutions tailored for specific AI workloads 4

4

.

The momentum reflects a fundamental shift in AI infrastructure spending. While Nvidia's GPUs continue to dominate, large companies are increasingly turning to custom AI chips that offer stronger performance and greater energy efficiency for specific tasks 3

3

. Broadcom provides the building blocks and critical intellectual property that enable customers to transform their chip designs into physical products manufactured at scale 3

3

.

CEO Hock Tan Reports Insatiable Demand Across Product Lines

In a recent virtual meeting with TD Cowen analysts, Broadcom CEO Hock Tan and CFO Kristen Spears delivered reassuring guidance that demand is growing not only for the company's XPUs but also for its AI networking solutions 1

1

. Tan emphasized that Broadcom faces "seemingly insatiable demand for its products" and expressed clear confidence in meaningful upside to the company's recently disclosed backlog figure of $73 billion expected to ship over the next 18 months 1

1

.

TD Cowen analysts, led by Joshua Buchalter, reiterated a buy rating with a $450 price target, suggesting 35% potential upside 1

1

. The firm noted that bookings have grown meaningfully since December's disclosure, with the multi-year chip production deal with OpenAI likely not even included in that $73 billion backlog 1

1

. Tan appeared "fully confident" that the OpenAI deal would be deployed between 2027 and 2029 1

1

.

Broadcom Holds 18-Month Lead Over Customer-Owned Chip Programs

JPMorgan analyst Harlan Sur maintained an Overweight rating with a $475 price target, highlighting that Broadcom has an 18-month-plus head start over internal customer-owned tooling efforts where customers handle about 90% of the design work 2

2

. He pointed to Google's 3nm Zebrafish program as an example, noting it has faced repeated delays and may not produce first silicon until the end of this quarter or early next quarter 2

2

.

In contrast, Broadcom's next-generation 3nm Tensor Processing Units, code-named Sunfish, were evaluated and qualified by Google earlier, providing a meaningful timing advantage 2

2

. Sur said he would "buy aggressively" at current levels, noting that tightening lead times for wafers, CoWoS, HBM, and advanced substrates make early qualification critical 2

2

.

Google Plans Massive TPU Deployment Powered by Broadcom Silicon

Google is now targeting deployment of 6 million to 7 million TPU processors in calendar 2027, up from prior estimates of 5 million units 2

2

. JPMorgan expects most of those 2027 shipments to support external customers through Google Cloud, with more than 95% of Google's TPU units in 2027 powered by Broadcom silicon 2

2

. Sur added that Broadcom already has multi-billions of dollars in purchase orders tied to the Sunfish ramp starting in the second half of this year 2

2

.

Broadcom's success with Alphabet's TPUs has led other hyperscalers to adopt its ASIC services. The company has indicated that its three earliest customers—believed to be Alphabet, Meta Platforms, and ByteDance—represent a $60 billion to $90 billion market opportunity in fiscal 2027 3

3

. Broadcom has also added OpenAI as a customer, signing a deal to supply custom AI accelerators capable of supporting data centers generating 10 gigawatts of computing power 3

3

.

AI Networking Sales Emerge as Underappreciated Growth Driver

Broadcom's AI networking sales are likely to become a larger part of total revenue than investors currently expect, according to TD Cowen 1

1

. The company holds leading positions in various components needed to build AI infrastructure, including Ethernet solutions, PCI Express switches used to connect hardware components, and optical interconnect solutions 1

1

. Tan suggested meaningful momentum across scale-up and scale-out in the company's Tomahawk franchise 1

1

.

Citigroup analysts estimate that Broadcom's AI revenue could surge from around $20 billion last fiscal year to over $50 billion this fiscal year, then potentially double again to $100 billion in fiscal 2027 3

3

. Given that Broadcom produced just $63.9 billion in total revenue in fiscal 2025, this represents enormous growth potential 3

3

. AI server compute ASIC shipments are expected to triple by 2027 as AI accelerator deployments accelerate across major cloud providers 4

4

.