Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
IndustryInsight · On RadarBloomberg Tech→ original

Broadcom targets $100 billion in AI chip sales by 2027

Broadcom CEO Hock Tan has said the company expects AI chip revenue to exceed $100 billion in 2027. It is an ambitious forecast and a clear push into territory long held by Nvidia, the entrenched leader in AI accelerators. Broadcom is betting on custom chips (ASICs) designed for specific customers such as Google and Meta*, offering an alternative to Nvidia's general-purpose GPUs. If the forecast comes true, the balance of power in AI infrastructure could change dramatically. *Meta has been designated an extremist organization and is banned in Russia.

AI-processed from Bloomberg Tech; edited by Hamidun News
Broadcom targets $100 billion in AI chip sales by 2027
Source: Bloomberg Tech. Collage: Hamidun News.
Listen to article

When the head of one of the world's largest semiconductor companies publicly names a figure of $100 billion, the market takes notice. That is exactly what happened after a statement from Broadcom CEO Hock Tan: the company expects its revenue from AI chip sales to pass that psychologically important mark in 2027. For an industry in which Nvidia has looked like an unshakeable monopolist, it sounds like a declaration of war.

To grasp the scale of Broadcom's ambitions, it helps to recall where the company came from. Historically, Broadcom was known mainly as a maker of network controllers, switches and infrastructure chips: a reliable but not especially visible player in the shadow of Intel and Qualcomm. That changed when Hock Tan, known for an aggressive acquisition strategy, turned it into a semiconductor giant with a market capitalization in the hundreds of billions of dollars. The 2023 purchase of VMware for $69 billion was merely the loudest of his deals. But Broadcom's real strategic pivot was toward custom AI chips, and that is the key to understanding today's statement.

Unlike Nvidia, which sells general-purpose GPU accelerators such as the H100 and B200 that suit a wide range of tasks, Broadcom bets on ASICs: application-specific integrated circuits designed for a particular customer and a particular workload. Its clients include top-tier tech giants: Google with its TPU (Tensor Processing Unit), Meta*, and, according to analysts, a number of other large cloud providers. The logic is simple: when you train GPT- or Gemini-scale models and spend billions of dollars a year doing it, even a ten percent gain in energy efficiency or performance per watt pays for itself many times over.

A custom chip tuned to your own architecture can deliver that gain and free you from reliance on a single supplier.

Fear of dependence on Nvidia has arguably been the main engine of Broadcom's growth in AI. For the past three years, the world's largest tech companies have found themselves in a position they strongly dislike: they queue for GPUs, accept Nvidia's terms and build their infrastructure around a single ecosystem, CUDA. For corporations used to controlling every layer of their technology stack, that is unacceptable. Broadcom offers a way out. It is not a replacement for Nvidia as such but an alternative path, in which the customer gets a chip designed for its needs, to its specifications and integrated into its infrastructure.

The $100 billion forecast is certainly ambitious and deserves a healthy dose of skepticism. By analysts' estimates, Broadcom's AI revenue in fiscal 2025 was around $12–15 billion. Growth to $100 billion in two years means roughly a sevenfold increase, a pace that requires not just organic growth but a fundamental expansion of the customer base and production volumes.

Still, there is reason to think Tan is not bluffing. Total investment in AI infrastructure is climbing fast: Microsoft, Google, Amazon and Meta* together plan to spend more than $300 billion on capital expenditure in 2026–2027, and a significant share of that money will go to chips. If Broadcom manages to capture even a third of that flow, the $100 billion figure stops looking fantastical.

For Nvidia, the statement is not a catastrophe but a serious signal. Jensen Huang's company still controls the overwhelming share of the AI accelerator market and has a formidable software ecosystem that cannot be reproduced in a year or two. But the very fact that a competitor publicly talks about hundred-billion-dollar prospects shows that the AI chip market is ceasing to be a monopoly. It is becoming a battlefield where Broadcom's custom designs, Google's and Amazon's in-house chips, and the efforts of AMD and new startups such as Cerebras and Groq are gradually eroding one player's dominance.

What does all this mean for the industry as a whole? First of all, healthy competition, which will inevitably push down the cost of AI compute. When customers have a real choice between general-purpose GPUs and custom ASICs, prices start to fall and innovation to rise. For companies building AI products, that means more affordable infrastructure. For end users, it means faster adoption of AI in everyday services. Hock Tan's statement is not just a corporate forecast. It is a marker of a new stage in the AI industry, in which the monopoly on compute gives way to a multipolar market. And that market is only beginning to take shape.

*Meta has been designated an extremist organization and is banned in Russia.

Story
ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Private AI Builders Community · Vibe Club

Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews

Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.

Join Vibe Club →Explore Hamidun Academy
What do you think?
Loading comments…