Hyperscaler bundling: Google Gemma/Arm/MSFT Maia/Oracle/AWS/SK sovereign + DigitalOcean + cloud AI shifts
Key Questions
What is Oracle's backlog and AI capacity plans?
Oracle has a $553 billion backlog, projecting $90 billion FY27 revenue with 10GW AI capacity. This reflects hyperscaler bundling trends.
How is Uber leveraging AWS for AI?
Uber is using AWS Trainium and Graviton chips to boost AI training and computing, hedging hyperscaler dependencies.
What differentiates AI cloud offerings like those from Snowflake and Azure?
AI cloud shifts include differentiators like Snowflake-OpenAI integrations and Perplexity-Azure bundles, amid superapp consolidation.
What issues are reported with Microsoft's Copilot?
Windows 11's Copilot app is RAM-intensive as a web app via Edge, contributing to bloat concerns in hyperscaler products.
How do Anthropic's deals hedge hyperscaler lock-ins?
Anthropic's Google/Broadcom TPU deals add multi-GW capacity, providing alternatives to single hyperscaler reliance like AWS or Azure.
What AI agent expansions is DigitalOcean pursuing?
DigitalOcean acquired Katanemo Labs and its Plano open-source project to deepen AI agent focus beyond infrastructure.
What are Google Gemma 4's capabilities?
Gemma 4, a 31B model under Apache 2.0, competes with giants 20x its size in benchmarks, supporting offline and edge AI.
How is AI redefining hyperscaler competition?
AI drives cloud competition through custom chips like MSFT Maia, Arm integrations, and sovereign clouds from AWS/Oracle/SK, with bundling accelerating adoption.
Oracle $553B backlog $90B FY27 10GW; Uber AWS Trainium/Graviton; AI cloud diff like Snowflake-OpenAI Perplexity-Azure; superapps/consolidation; MS Copilot bloat; Anthropic Google/Broadcom TPU deals hedge hyperscaler lock-ins.