Broadcom logo on smartphone screen digital background.

Broadcom's New Google Chips Could Be 40% Cheaper To Run Than Nvidia's, Analyst Says

Broadcom Inc (NASDAQ:AVGO) is gaining momentum in the artificial intelligence accelerator race thanks to Alphabet Inc (NASDAQ:GOOGL) (NASDAQ:GOOG) Google's push to scale Tensor Processing Unit-powered computing across both internal systems and future external customers.

• AVGO shares are under pressure. Track the latest developments here.

BofA Securities analyst Vivek Arya reiterated a Buy rating on Broadcom and increased the price target from $400 to $460.

Arya noted Google's growing reliance on its homegrown TPUs as a powerful tailwind for Broadcom, the key silicon design partner behind the chips.

Also Read: Three Big Catalysts Could Decide Broadcom’s Earnings-Day Pop, Analyst Says

After Google trained Gemini 3 entirely on TPUs and signaled plans to rent them out to external customers, the analyst analyzed how the latest TPUv7 stacks up against NVIDIA Corp's (NASDAQ:NVDA) new Blackwell Ultra graphics processing units and refreshed Broadcom's revenue model tied to TPU shipments and pricing.

He expects Broadcom's TPU business to accelerate meaningfully.

The current TPU average selling price of roughly $5,000 to $6,000 and an estimated two million units in calendar 2025 could grow to $12,000 to $15,000 and more than three million units in 2026.

If demand from both Google and new external customers like Anthropic and Meta Platforms Inc (NASDAQ:META) expands, shipments could stretch toward 3.6 million to 3.8 million units, as per Arya.

The analyst argued that TPUv7 can outperform Nvidia's GB300 on power efficiency for specific AI training tasks, such as FP8 large language models, producing about 5.4 TFLOPs per watt versus roughly 3.6 TFLOPs per watt for Blackwell Ultra.

That advantage can translate to up to 40% lower total cost of ownership for those optimized workloads.

Still, he noted performance varies heavily depending on workload type, model support and software optimization — areas where Nvidia's GPU ecosystem still leads.

What Does the Future Hold?

Looking ahead, Nvidia's next major architecture, Vera Rubin, could leapfrog TPU on memory technology and system cost when it arrives in late 2026.

Broadcom's TPU roadmap may not deliver a major semiconductor upgrade until TPUv8 in 2027, potentially narrowing competitive gains, Arya told.

The analyst also flagged risks.

If Google shifted from renting TPUs exclusively through Google Cloud to selling them directly, it could create fresh competition for Broadcom's other custom Application-Specific Integrated Circuit customers, including Meta, ByteDance and OpenAI.

Google could also expand its design partner roster by adding MediaTek for lower-tier TPU variants, reducing Broadcom's share.

Even with those uncertainties, he noted Broadcom remains well-positioned as AI compute spending scales.

Arya expected the company's custom ASIC and TPU products to account for up to 15% of the roughly $900 billion AI accelerator market by 2030, up from roughly 8% this year.

The analyst modelled more than $23 in Broadcom EPS power by 2030 and modestly adjusted gross margin expectations to reflect a higher mix of compute silicon.

Overall, he noted rising TPU leverage and broadening adoption as meaningful growth catalysts for Broadcom as AI infrastructure investment ramps through 2027 and beyond.

Arya projected fourth-quarter revenue of $17.445 billion (up from prior forecast of $17.405 billion) and adjusted EPS of $1.83 (down from prior guidance of $1.85).

AVGO Price Action: Broadcom stock was trading lower by 2.92% to $391.19 at publication on Monday.

Loading...
Loading...

Read Next:

Photo: Shutterstock

Market News and Data brought to you by Benzinga APIs

Comments
Loading...