GLOBAL RESEARCH ARCHIVE
Neocloud Expansion (Good Read), Token Cost Reductions, Hynix Capacity Build, & More—Headline Review (Semis & IT Hardware)
Research evidence excerpt
Neocloud Expansion (Good Read), Token Cost Reductions, Hynix Capacity Build, & More—Headline Review (Semis & IT Hardware)
Neocloud Expansion (Good Read), Token Cost Reductions, Hynix Capacity Build, & More—Headline Review (Semis & IT Equity ResearchHardware)
• A few recent examples of rising token costs include Uber burning through its entire 2026 token
budget in April, Salesforce estimated to pay Anthropic ~$300M/yr., Walmart limiting employee
token usage limits, Microsoft reportedly recently canceled most internal Claude Code licenses, and
an unnamed company reportedly going through $500M in Anthropic Claude AI usage in a single
month.
3. KKR Launches Helix Digital Infrastructure With >$10B Committed for AI Data Center
Deployment
• Helix is being launched with investment from KKR, Kuwait Investment Authority, NVIDIA,
and Vistra (>$10B long-duration capital commitments to date). Helix will deliver integrated
infrastructure at the speed and scale hyperscalers need and aims to serve as a single coordination
point for hyperscalers' data centers, power, connectivity, and related needs.
• NVIDIA will also serve as a strategic partner to support deployment of NVIDIA DSX AI factory-
aligned infrastructure, noted to de maximize tokens per watt and achieve lowest TCO and
accelerating time to first token.
• Former CEO of Amazon Web Services, Adam Selipsky, will lead the company.
4. Arm vs. x86 Competitive Dynamics Poised to Heating Up?
• Yesterday AMD outlined benchmark details for its now launched next-generation Zen6-based
Venice EPYC server CPUs - outlining (emphasizing) its competitive positioning versus NVIDIA's 88-
core Arm-based Vera CPUs (see: AMD fires back at Nvidia, claiming 256-core Zen 6 'Venice' CPU
beats Vera by 3.3x in rack-level performance).
The English excerpt is extracted automatically from the cited source page and may contain layout or recognition errors. It is never batch translated.
Open report viewer