ARPU
arpunow.bsky.social
ARPU
@arpunow.bsky.social
Thinking out loud about semiconductor, AI, and cloud.

Newsletter: arpu.hedder.com
But rivals like OpenAI & Google can't risk their development pipeline being visible to a competitor, so they are forced to walk away.

In the AI arms race, your supply chain is now a target.
June 18, 2025 at 3:09 PM
With Oracle’s cloud backlog surging and reports of them taking any capacity they can get, this program locks in a new, high-value customer segment that requires immense security and compliance—Oracle's sweet spot. (3/3)
June 18, 2025 at 2:59 PM
Its new 'Defense Ecosystem' program onboards small AI vendors, guides them through complex procurement, and—most importantly—gets them to build their solutions on Oracle Cloud Infrastructure (OCI). (2/3)
June 18, 2025 at 2:59 PM
Why this matters: Even the most advanced "AI factories" have a dirty secret—they're built on a mountain of these "boring" chips.

As NVIDIA's own CEO Jensen Huang said in the release, they're partnering with TI for their AI infrastructure.
June 18, 2025 at 2:54 PM
Texas Instruments is making a different bet: dominating the U.S. supply chain for foundational analog & embedded chips—the workhorses inside every car, phone, and server.
June 18, 2025 at 2:54 PM
The semiconductor market is now operating at two speeds. All eyes are on the race for 2nm AI chips from TSMC. But that's a brutal, capital-intensive fight with huge onshoring challenges (see TSMC in Arizona).
June 18, 2025 at 2:54 PM
The AI chip market is projected to be a $500B+ opportunity by the late 2020s. Intel has to be a player to survive and thrive.
June 18, 2025 at 2:46 PM
This is about more than just hiring smart people. Intel is acquiring specific DNA.

▶️ The new head of AI Chip Architecture comes from Google, a leader in custom silicon (TPUs).
▶️ The new head of AI SoC Engineering comes from Rain AI, a specialized startup.
June 18, 2025 at 2:46 PM
Let's be clear: Intel is in a tough spot. They're losing PC/server market share to AMD & Arm, their foundry is years behind TSMC, and their AI accelerators like Gaudi haven't meaningfully challenged Nvidia's dominance in the data center.
June 18, 2025 at 2:46 PM
AI's reliability is still in question. High failure rates for AI projects and instances like Klarna having to re-hire humans after AI-driven cuts highlight ongoing challenges.
June 18, 2025 at 3:44 AM