Buzzberg Cup Live

An Interview with CoreWeave Executives: AI Demand Seems to 'Intensify' Every Day

Tae Kim · Key Context by Tae Kim · June 17, 2026 at 12:35 · ⏱ 7 min read  | Read on Substack ↗
Summary
CoreWeave executives describe AI demand intensifying daily, driven by agentic AI and inference workloads, with powered shell availability as the primary bottleneck. The neocloud's customer-led model and strong execution differentiate it from hyperscalers, while cost pass-through for HBM protects margins. For markets, this reinforces sustained GPU demand and validates Nvidia's roadmap, with CoreWeave itself positioned as a key beneficiary.
  • •CoreWeave is the only independent AI cloud serving Anthropic, OpenAI, Meta, Google, Microsoft, and Nvidia as customers.
  • •Agentic AI demand began in Q4 2025 with a massive inflection in inference and AI consumption in Q1 2026.
  • •98%+ of CoreWeave's revenue is contractually driven, with customers defining infrastructure builds.
  • •Powered shells (data center buildings) are the biggest bottleneck, not GPUs or HBM; CoreWeave has 49 sites online and running.
  • •HBM cost increases are passed through to customers by locking in GPU pricing simultaneously with purchase orders.
  • •CoreWeave expects Vera Rubin servers to begin appearing in late 2026 with heavy ramp throughout 2027, mirroring the GB200/GB300 pattern.
Read time 7 min
Length 7,918 chars
Category finance
Ideas
Tae Kim Senior writer, Barron's; author of The Nvidia Way
The interview positions CoreWeave as the leading neocloud with unmatched customer base (9/10 top AI labs), platinum SemiAnalysis rating, and strong execution (49 sites). AI demand 'intensifies every d
The interview positions CoreWeave as the leading neocloud with unmatched customer base (9/10 top AI labs), platinum SemiAnalysis rating, and strong execution (49 sites). AI demand 'intensifies every day' and the company's contract-driven model provides visibility and pricing power. Public listing assumed by 2026. Risk: Competition from hyperscalers (Azure, AWS, GCP) and other neoclouds; powered shell bottleneck could slow growth if supply chain doesn't improve.
CRWV
Tae Kim Senior writer, Barron's; author of The Nvidia Way
CoreWeave expects to be an early adopter of Nvidia Vera CPUs and Vera Rubin servers, and notes demand for Nvidia GPUs remains dominant. The interview validates Nvidia's roadmap continuity and strong c
CoreWeave expects to be an early adopter of Nvidia Vera CPUs and Vera Rubin servers, and notes demand for Nvidia GPUs remains dominant. The interview validates Nvidia's roadmap continuity and strong customer pull for next-gen hardware. Risk: ASIC competition (Amazon Trainium, Google TPU) could erode Nvidia's share in inference over time.
NVDA
Tae Kim Senior writer, Barron's; author of The Nvidia Way
HBM memory costs and shortages are discussed, with CoreWeave confirming they pass through price hikes to customers. This suggests sustained pricing power for HBM suppliers like Micron as AI GPU demand
HBM memory costs and shortages are discussed, with CoreWeave confirming they pass through price hikes to customers. This suggests sustained pricing power for HBM suppliers like Micron as AI GPU demand grows. Risk: HBM supply may loosen if powered shell constraints cap GPU deployments, reducing near-term demand.
MU
Tae Kim Senior writer, Barron's; author of The Nvidia Way
CoreWeave's fleet is currently 'mostly AMD' for CPUs, but they expect 'ton of interest for Vera CPU' and that mix will shift over time based on customer demand. This implies AMD's share in CoreWeave's
CoreWeave's fleet is currently 'mostly AMD' for CPUs, but they expect 'ton of interest for Vera CPU' and that mix will shift over time based on customer demand. This implies AMD's share in CoreWeave's CPU deployments could decline as Nvidia Vera gains traction. Risk: AMD could maintain share if its own CPU roadmap (e.g., Turin) competes effectively or if Nvidia Vera faces delays.
AMD
More from Key Context by Tae Kim
Why the RTX Spark Shows NVIDIA's Genius
Jun 15, 20:42
Brief Thoughts on the SpaceX IPO
Jun 12, 18:45
Intel AI Executive on Why the Chipmaker's CPUs Are Well Positioned for Agentic AI
Jun 11, 01:00
Nvidia Executive Is Bullish on CPO Rollout, Seemingly Refuting SemiAnalysis Narrative
Jun 09, 17:49
7 Basic Portfolio Management Tips
Jun 08, 23:21

This newsletter, published June 17, 2026, features Tae Kim discussing CRWV, NVDA, MU, AMD. 4 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Tae Kim  · Tickers: CRWV, NVDA, MU, AMD