Hyperscaler capex/infra: $1T+ Big Tech CapEx/backlogs, Broadcom-Google-Anthropic chips, Nutanix hybrid, NVIDIA GTC/Firmus/Intel/Uber/AI factories
Key Questions
What is the projected Big Tech CapEx for 2026?
Big Tech hyperscalers are projected to spend over $1T on CapEx in 2026, with Amazon at $200B, Google at $180B, Microsoft at $145B+, Meta at $72-135B, and Alibaba at $53B. Backlogs exceed $240B. This reflects massive investments in AI infrastructure.
What are the details of Broadcom's chip deals with Google and Anthropic?
Broadcom has expanded custom chip deals with Google through 2031 and with Anthropic for 3.5GW TPU capacity, targeting $50-100B in revenue. These deals position Broadcom in the $60B hyperscaler AI infrastructure market. The partnerships highlight the new battlefield in AI infrastructure.
What new capabilities did Nutanix announce?
Nutanix expanded its Agentic AI Platform with hybrid multicloud infrastructure, supporting Kubernetes, RDMA, and S3. This enables agentic workloads across hybrid environments. It addresses the need for flexible AI deployment.
What were key announcements at NVIDIA GTC 2026?
NVIDIA GTC featured product launches for AI infrastructure, including advancements in AI factories requiring 10x more energy. Everest Group reviewed these announcements. They underscore NVIDIA's role in scaling AI compute.
What is the status of Firmus, the AI datacenter builder?
Firmus, backed by NVIDIA and dubbed 'Southgate', reached a $5.5B valuation. It focuses on building AI datacenters in Asia. This reflects surging demand for AI infrastructure.
How is Uber utilizing Amazon's AI chips?
Uber adopted Amazon's AI chips for its infrastructure, optimizing delivery and rideshare operations. This follows other hyperscalers' custom chip strategies. It demonstrates growing reliance on specialized AI hardware.
What is the ROI for AI infrastructure projects according to Gartner?
Gartner reports that only 28% of AI infrastructure projects fully pay off. This highlights challenges in achieving expected returns despite heavy investments. Factors include TCO, power, and geopolitics.
What role does Microsoft play in Canadian AI infrastructure?
Microsoft is building AI infrastructure in Canada with a community-first approach. This involves collaborations to address local needs. It ties into broader hyperscaler CapEx plans.
$1T+ 2026 CapEx (Amazon $200B/Google $180B/MS $145B+Canada/Meta $72-135B/Alibaba $53B); backlogs $240B+; Broadcom-Google custom chips thru 2031/Anthropic 3.5GW TPU/Broadcom $50-100B; Nutanix Agentic hybrid multicloud (K8s/RDMA S3); NVIDIA GTC/AI factories 10x energy/Gartner 28% ROI. Ongoing chips/geopol/power TCO/NVIDIA/Nutanix/P2P.