open_in_newOriginal article
Google could partner with MediaTek for next gen AI chips: Report

Google could partner with MediaTek for next gen AI chips: Report

  • 18.03.2025 13:39
  • androidpolice.com
  • Keywords: AI, Google, MediaTek, Broadcom, NVIDIA, TSMC, OpenAI, Taiwan

Google is considering switching TPU partners from Broadcom to MediaTek for cost optimization in next-gen AI chips. MediaTek would handle I/O modules while Google focuses on in-house design, reducing dependency on Broadcom and enhancing competitiveness in the AI market.

Alphabet ReportsAVGOsentiment_dissatisfiedNVDAsentiment_neutral

Estimated market influence

Google

Negativesentiment_dissatisfied
Analyst rating: N/A

Google is considering switching TPU partners from Broadcom to MediaTek for cost optimization and reducing dependency on Broadcom.

MediaTek

Positivesentiment_satisfied
Analyst rating: N/A

MediaTek could gain a partnership with Google, which may lead to increased revenue and market presence in AI chips.

Broadcom

Broadcom

Negativesentiment_dissatisfied
Analyst rating: Strong buy

Broadcom's dependency on Google as its biggest revenue source might decrease if the switch happens, potentially affecting its financials.

NVIDIA

NVIDIA

Neutralsentiment_neutral
Analyst rating: Strong buy

Google is still a customer of NVIDIA but may reduce reliance over time if the partnership with MediaTek succeeds.

TSMC

Neutralsentiment_neutral
Analyst rating: N/A

TSMC continues to handle production, maintaining its role despite changes in design partners.

OpenAI

Neutralsentiment_neutral
Analyst rating: N/A

Mentioned as a rival using NVIDIA chips but not directly impacted by Google's partnership change.

Context

Analysis of Google's Potential Partnership with MediaTek for Next-Gen AI Chips

Key Facts and Data Points

  • Potential Switch in TPU Partners: Google may transition from Broadcom to MediaTek for producing Tensor Processing Units (TPUs) to optimize costs.
  • Division of Responsibilities: MediaTek will handle I/O module design, while Google focuses on in-house TPU design.
  • Cost Advantage: MediaTek charges less per chip compared to Broadcom, potentially reducing Google's expenses.
  • Production Collaboration: TSMC will continue production, with MediaTek liaising and handling quality control.

Financial Implications

  • TPU Spending: Google spent $6–9 billion on TPUs in the last year alone.
  • NVIDIA Investment: Google placed an order for over $10 billion worth of NVIDIA's Blackwell chips.
  • Cost Optimization: Reducing dependency on Broadcom could help Google maintain competitive pricing in cloud AI services.

Competitive Landscape

  • Market Rivals: Competitors like OpenAI rely on NVIDIA GPUs (e.g., for ChatGPT).
  • Broadcom's Position: Broadcom remains a major player with clients including Meta, Arm Holdings, Apple, and OpenAI.

Strategic Considerations

  • In-House Design Focus: Google is recruiting chip design engineers in Taiwan, signaling a shift toward more internal TPU development.
  • Dual TPU Strategy: Google produces two types of TPUs—one for model training and another for services like Search, YouTube, and Gemini.

Market Trends and Long-Term Effects

  • AI Compute Demand: Growing demand for AI compute modules positions Google to benefit from reduced dependency on Broadcom.
  • Production Shifts: The partnership with MediaTek could streamline production processes with TSMC, enhancing efficiency.

Regulatory and Industry Impact

  • Reduced Dependency Risks: Diversifying chip suppliers minimizes risks associated with sole reliance on Broadcom.
  • Potential for Cost Leadership: Lower costs could enable Google to offer more competitive cloud AI services.

This potential partnership underscores Google's strategic focus on cost optimization, in-house innovation, and long-term competitiveness in the AI hardware market.