Alphabet’s first-quarter results have revived the argument that hyperscalers with in-house silicon may be better placed for the AI buildout than the merchant suppliers serving them. But the “sell Broadcom, buy Google” case published by 24/7 Wall St. via AOL is more a valuation call than evidence that Broadcom is being pushed aside.
Alphabet reported $109.9 billion in first-quarter revenue, up 22% year over year, while Google Cloud revenue rose 63% to $20.0 billion. Per Alphabet’s SEC filing, Cloud backlog nearly doubled sequentially to more than $460 billion, and the segment produced $6.6 billion in operating income. Those numbers make Google’s vertical model—TPUs, cloud capacity, Gemini models and enterprise distribution—hard to ignore.
Google uses its Tensor Processing Units internally and offers them through Google Cloud, letting it optimize hardware, software and infrastructure as a single stack. For enterprise customers running AI workloads, that can mean access to an alternative to GPU-heavy deployments, particularly where Google’s models and managed cloud services are already part of the environment.
That does not mean every AI workload shifts to Google-designed hardware. Customers still buy a mix of Nvidia GPUs, custom accelerators, networking gear and cloud services, depending on model frameworks, availability, pricing and vendor commitments. The practical advantage for IT buyers is broader infrastructure choice, not a guaranteed end to merchant silicon demand.
Broadcom is not merely a generic chip vendor selling into a fading market. Its business includes custom silicon programs and AI networking, both of which are central to hyperscaler data-center expansion. A large cloud operator designing more of its own chips can still be a Broadcom customer for design support, networking components or other infrastructure.
The cited sale by Broadcom co-founder Henry Samueli also needs context. His June 24 stock transactions were executed under a Rule 10b5-1 trading plan adopted in December 2025, according to the Form 4 filing. Insider sales can matter, but a prearranged plan is not, by itself, evidence of a changing business outlook.
Investors can debate multiples, but administrators should expect continued pressure to evaluate cloud AI offerings, accelerator availability, network design and long-term vendor lock-in rather than assume a single chip supplier or hyperscaler will dominate the stack.
Alphabet reported $109.9 billion in first-quarter revenue, up 22% year over year, while Google Cloud revenue rose 63% to $20.0 billion. Per Alphabet’s SEC filing, Cloud backlog nearly doubled sequentially to more than $460 billion, and the segment produced $6.6 billion in operating income. Those numbers make Google’s vertical model—TPUs, cloud capacity, Gemini models and enterprise distribution—hard to ignore.
Google’s TPU advantage is real
Google uses its Tensor Processing Units internally and offers them through Google Cloud, letting it optimize hardware, software and infrastructure as a single stack. For enterprise customers running AI workloads, that can mean access to an alternative to GPU-heavy deployments, particularly where Google’s models and managed cloud services are already part of the environment.That does not mean every AI workload shifts to Google-designed hardware. Customers still buy a mix of Nvidia GPUs, custom accelerators, networking gear and cloud services, depending on model frameworks, availability, pricing and vendor commitments. The practical advantage for IT buyers is broader infrastructure choice, not a guaranteed end to merchant silicon demand.
Broadcom is still benefiting from custom AI chips
Broadcom’s own June earnings report points in the opposite direction from a simple disintermediation narrative. The company posted fiscal second-quarter revenue of $22.2 billion, up 48%, and said AI semiconductor revenue rose 143% to $10.8 billion. Broadcom projected AI semiconductor revenue of $16 billion for its fiscal third quarter, with demand coming from custom AI accelerators and networking.Broadcom is not merely a generic chip vendor selling into a fading market. Its business includes custom silicon programs and AI networking, both of which are central to hyperscaler data-center expansion. A large cloud operator designing more of its own chips can still be a Broadcom customer for design support, networking components or other infrastructure.
The cited sale by Broadcom co-founder Henry Samueli also needs context. His June 24 stock transactions were executed under a Rule 10b5-1 trading plan adopted in December 2025, according to the Form 4 filing. Insider sales can matter, but a prearranged plan is not, by itself, evidence of a changing business outlook.
What Windows and enterprise IT teams should watch
For Windows shops, the relevant development is the increasingly strategic nature of AI infrastructure procurement. Google’s acceleration in Cloud and TPU investment strengthens its position against Azure, AWS and private GPU deployments. Broadcom’s growth confirms that network fabrics and custom accelerators remain costly, capacity-constrained parts of that equation.Investors can debate multiples, but administrators should expect continued pressure to evaluate cloud AI offerings, accelerator availability, network design and long-term vendor lock-in rather than assume a single chip supplier or hyperscaler will dominate the stack.
References
- Primary source: aol.com
Published: 2026-07-13T18:49:16+00:00
Alphabet Vs. Broadcom: Rotate Out of AVGO and into GOOG as a Custom Chip Creator and Hyperscaler Partner - AOL
Broadcom just posted 143% AI revenue growth, yet its stock shed a quarter of its value in a single month while a founder quietly dumped over a million shares. The hyperscaler quietly disintermediating its own chip suppliers may explain why.www.aol.com
