"Alphabet could spend as much as $185 billion in capital expenditures in 2026, nearly double last year's level. That money is funding: AI data centers, Custom Tensor Processing Units (TPUs), Cloud infrastructure, Gemini model training, AI-powered search products. Here's what the hyperscaler spending picture now looks like: That's a staggering amount of capital being deployed into a business model that is still searching for durable monetization outside cloud subscriptions and advertising."
"That raises an uncomfortable question for investors: How long can hyperscalers keep pouring hundreds of billions into AI infrastructure before shareholders demand better returns? Alphabet's AI spending spree has become one of the biggest stories in tech this year. Alphabet may be trying to offload part of that burden onto users themselves through edge computing. That's a fancy way of saying AI tasks run locally on your device instead of entirely in Google's cloud infrastructure."
"Reports surfaced this week that Google Chrome has been quietly downloading a roughly 4GB file called "weights.bin" onto some users' computers. The file powers Gemini Nano, Google's local AI model used for features like writing assistance, summarization, scam detection, and autocomplete. Chrome matters here because it dominates global browser usage. Windows Central cited StatCounter data showing Chrome controls nea"
Hyperscalers including Alphabet, Microsoft, Amazon, and Meta are projected to spend about $710 billion to $725 billion this year on data centers, networking systems, and AI chips. The scale of spending raises concerns about how long investors will tolerate returns that are not yet clearly durable beyond cloud subscriptions and advertising. Alphabet’s capital expenditures could reach roughly $185 billion in 2026, funding AI data centers, custom TPUs, cloud infrastructure, Gemini model training, and AI-powered search products. To reduce cloud burden, Google is moving some computation to edge devices. Chrome is reported to download a 4GB weights.bin file that enables Gemini Nano, a local AI model for writing assistance, summarization, scam detection, and autocomplete.
Read at 24/7 Wall St.
Unable to calculate read time
Collection
[
|
...
]