Wedbush Says a Rumored Google-AMD Chip Deal Could Reshape AI - Here's Why

Semiconductor research firm SemiAnalysis told clients that "market chatter suggests" Alphabet Inc. (NASDAQ:GOOGL) is working with Advanced Micro Devices Inc. (NASDAQ:AMD) on a 10th-generation Tensor Processing Unit. Neither company has confirmed the project. Tom's Hardware, which reported the note, expects such a chip would be built for reinforcement learning and agentic AI workloads.

Wedbush analyst Matt Bryson called the potential collaboration significant, and a company quietly building its own chip leverage is among the signs of financial success that may not appear in any single earnings report.

The SemiAnalysis note and what it actually said

SemiAnalysis stated in a client note that "market chatter suggests" Google is working with AMD on a TPU project in the v10 generation, Tom's Hardware reported. Neither Google nor AMD has publicly commented on the report, and the scope of AMD's role remains unclear.

The reported collaboration would mark AMD's first major involvement in a custom AI accelerator project, according to SemiAnalysis, though AMD is already building a customized version of its MI450 accelerator for Meta. Google has developed eight generations of TPUs with Broadcom Inc. (NASDAQ:AVGO) as its primary silicon design partner and maintains a supply agreement with Broadcom that runs through at least 2031.

Why Wedbush's Matt Bryson called the collaboration significant

Bryson said a Google-AMD relationship would be significant and might indicate the growing relevance of skills in designing ASICs and reusable chip intellectual property, according to an August 18, 2026, Seeking Alpha report. A more prominent role on Google's TPU roadmap could give AMD another avenue to capitalize on hyperscaler AI spending, on top of its existing accelerator and processor business.

Bryson holds a 4.99-star analyst rating with a 77.5% success rate, StockAnalysis data shows. AMD shares traded near $484 on August 18, 2026, while Alphabet traded near $344 on the same date. Wedbush maintains an Outperform rating on AMD.

Reinforcement learning and agentic AI demand a different chip design

Standard AI model training relies heavily on GPU-style tensor compute, but reinforcement learning and agentic AI workloads require substantially more general-purpose CPU processing power. SemiAnalysis stated the design rationale for the 10th-gen TPU is directly tied to these emerging workload types.

Google has already increased CPU resources in its current TPU 8i systems by pairing one Google Axion CPU with every two TPUs, doubling the ratio used in seventh-generation hardware, Dataconomy noted. Tom's Hardware reported hearing that a 1:1 ratio of CPUs to accelerators is optimal in some cases.

AMD's MI300A and the hybrid blueprint this project appears to need

Analysts cited by Dataconomy said AMD's appeal lies in three capabilities that few other chipmakers combine in a single offering.

• x86 CPU intellectual property that integrates general-purpose processing directly alongside accelerators.
• Advanced packaging expertise for combining multiple chiplets in a single package.
• The MI300A chip, which already integrates x86 cores and AI accelerator chiplets together, providing a working prototype of the hybrid design this TPU reportedly needs.

AMD is the only vendor shipping a high-volume data center processor that puts x86 CPU cores and GPU chiplets on one package with a single unified memory pool. The 10th-gen TPU project reportedly needs exactly this kind of integration, which is why SemiAnalysis flagged AMD as the partner rather than a traditional GPU or ASIC designer.

How this could affect the Broadcom relationship that runs through 2031

Google's Broadcom partnership covers conventional TPU design and manufacturing through at least 2031, and analysts noted the AMD collaboration is unlikely to replace that arrangement. Broadcom handles silicon implementation, high-speed interconnects, and advanced packaging for Google's existing TPU roadmap.

The 10th-gen TPU project appears to target a different category of workload rather than a replacement for Broadcom's core role. Reinforcement learning chips with integrated CPU cores represent a parallel design track, not a substitution, and your assessment of Alphabet's chip strategy may benefit from viewing it through that lens.

What this signals about Alphabet's AI infrastructure margin over time

Alphabet has guided to $195–205 billion in 2026 capital expenditures and spent $80.6 billion in the first half, with the majority flowing toward AI infrastructure. Building custom silicon in-house reduces long-term dependence on third-party GPU suppliers and could widen the company's AI infrastructure margins as workloads shift toward agentic computing.

Google Cloud revenue reached $24.8 billion in the second quarter of 2026, growing 82% year over year, and its contract backlog stood at approximately $514 billion. Custom chip development at this scale is a structural advantage that compounds over multiple hardware generations, and the TPU 10 project with AMD may accelerate that trajectory.

Risks in relying on a report neither company has confirmed

SemiAnalysis attributed its information to "market chatter," and neither Google nor AMD has issued a public statement. Analyst calls based on unconfirmed reports carry inherent uncertainty, and chip development timelines often stretch years beyond initial reports. The 10th-generation TPU is likely a few years from production.

AMD also faces execution risk. Custom ASIC design is a different business from selling standard processors and GPUs, and scaling a new capability while maintaining its existing product roadmap adds operational complexity. You may want to weigh how much of your investment thesis depends on unconfirmed developments versus the fundamentals already on record.

Bottom line

Google's reported work with AMD on a 10th-generation TPU is unconfirmed but directionally significant, as Wedbush's Matt Bryson noted. The project targets reinforcement learning and agentic AI workloads that demand tighter CPU-accelerator integration than current chip designs provide, and AMD's MI300A already demonstrates the hybrid architecture this effort reportedly requires.

Analyst notes like this one often move stocks before the underlying products reach production, and having must-have investing apps configured to surface these reports in real time may give you a head start on evaluating their significance. The real test for Alphabet shareholders is not one chip collaboration but whether Google's long-term silicon strategy reduces its infrastructure costs enough to show up in margins across future earnings.

This article is for informational purposes only and should not be considered investment advice.

You May Also Like

Almost 80% of Americans Fear a Retirement Age Increase — Here’s the Real Reason Why

By FinanceBuzz Editors

10 Smart Money Moves to Retire

By FinanceBuzz Editors

The Easiest Way to Diversify Your Investment Portfolio With Fine Art

By FinanceBuzz Editors

7 Brilliant Ways to Build Wealth After 40

By FinanceBuzz Editors

5 Ways Seniors Are Throwing Away Money

By FinanceBuzz Editors

Best High-Yield Savings Accounts of September 2026

By FinanceBuzz Editors

Help Protect Your Savings From Stock Market Volatility With Gold

American Hartford Gold Benefits

• American Hartford Gold helps individuals protect their retirement by rolling over IRAs and 401(k)s into physical gold.
• Includes FREE IRA rollover and storage for up to 3 years.
• Get up to $20,000 in free silver on qualifying purchases.

Download your guide for free

---

Subscribe Today

Learn how to make an extra $200

Get vetted side hustles and proven ways to earn extra cash sent to your inbox.

or

Subscribe Now →

By submitting this form you agree to receive emails from FinanceBuzz and to the privacy policy and terms