① Hyperscalers are in a capacity arms race: $600B committed to infrastructure, Duke Energy adding 2.7GW (7.6GW total data center power agreements), and Spain approving a €1.226B datacenter project through 2030—power is the binding constraint. ② The chip bottleneck persists despite aggressive supply expansion: Nvidia's B200 systems are sold out, Micron's HBM is fully booked, and AMD's data center revenue surged 107% YoY to 58% of total; Samsung is advancing HBM4E and HBM5 roadmaps to chase the deficit. ③ Vertical integration is reshaping the industry—Anthropic entered AI chip design, SpaceX committed exclusively to Nvidia while deploying $18.4B quarterly capex for AI cloud, and Amazon's $50B OpenAI investment consolidates capital around compute infrastructure.
Hyperscaler capex $600B · Amazon-OpenAI $50B · AMD data center +107% YoY · SpaceX capex $18.4B · Anthropic compute $10B
Hyperscalers committed nearly $600 billion to AI infrastructure amid surging demand for compute capacity.
The collective capex surge across Microsoft, Meta, Amazon, and Google demonstrates AI infrastructure buildout is now a dominant driver of enterprise capex, reshaping power, datacenter, and chip supply chains globally.
The collective capex surge across Microsoft, Meta, Amazon, and Google demonstrates AI infrastructure buildout is now a dominant driver of enterprise capex, reshaping power, datacenter, and chip supply chains globally.
Duke Energy secured 7.6 GW of electric service agreements with data center firms, adding 2.7 GW since Q4 2025.
The rapid pipeline expansion by a major US utility signals severe power constraints on datacenter siting; AI infrastructure now defines regional power grid planning.
Amazon completed a $50 billion investment in OpenAI, accelerating joint AI infrastructure and cloud deployment.
The massive capital commitment between a hyperscaler and leading frontier AI lab signals structural demand for GPU compute infrastructure and integrated AI-cloud services.
SpaceX posted $7.8B Q1 revenue but revealed $18.4B quarterly capex surge for AI datacenter expansion.
The $18.4B capex spike—among the highest quarterly spend by any company—confirms SpaceX is executing a hyperscaler-scale AI compute build while maintaining aerospace revenue, fundamentally reshaping capital intensity benchmarks.
UAE sovereign wealth fund considering $6.3 billion investment in AI data center infrastructure in Japan, per Bloomberg reporting.
Major sovereign capital commitment to non-US AI infrastructure, signaling geopolitical diversification of gigawatt-scale deployment away from China/US duopoly.
Envision commissions Galaxy Campus in Ulanqab, China, operationalizing a new model for gigawatt-scale AI infrastructure as a single integrated campus facility.
Adds major gigawatt-scale compute capacity in China; validates single-campus campus model for hyperscale AI data center deployment.
NVIDIA is reducing next-generation GPU specifications in response to persistent high-bandwidth memory (HBM) shortage, forcing design compromises in AI accelerators.
Supply-chain constraint in HBM directly threatens GPU performance headroom and may delay capacity deployment across the AI infrastructure buildout.
Business Insider obtained internal Amazon documents revealing major structural overhaul of one of the world's largest AI data centers.
Signals potential AI infrastructure strategy shift at key hyperscaler; may indicate new facility design, workload consolidation, or efficiency improvements.
Governor orders audit and pauses all new datacenter grid connections, citing infrastructure strain concerns.
Hard regulatory constraint on AI datacenter expansion in key US regions; grid infrastructure becoming capacity binding constraint before power generation.
Core Scientific, a major compute infrastructure provider, is pivoting $14 billion away from bitcoin mining to build an AI data center hub, marking a complete business model shift.
Major peer (Core Scientific) reallocating entire capital to AI infrastructure; signals market consolidation and capacity buildout acceleration among tier-1 operators.
OpenAI accounts for over 50% of Microsoft's AI revenue, disclosed in rare corporate transparency; raises questions about Microsoft's strategic dependence on single vendor.
Exposes critical revenue concentration in OpenAI; highlights Microsoft's vulnerability to OpenAI partnership disruption and potential antitrust/policy risk.
SK Hynix and SanDisk publish HBF (High Bandwidth Flash) specification: 512 GiB flash cache co-packaged with GPUs to address AI inference memory wall; endorsed by Google and Tenstorrent.
Establishes new memory tier between HBM and SSD, enabling cost-effective model serving at scale; unblocks inference scaling bottleneck.
Bitdeer secures 4.7 billion USD lease for a major AI data center facility in Norway, establishing long-term AI infrastructure capacity in Europe.
Significant AI capacity expansion in Europe via a multi-billion dollar infrastructure commitment; demonstrates sustained private investment in European AI compute availability.
Envision's Wulansuhai Xinghe AI data center (claimed as world's largest single compute facility by computing density) begins production in China.
Major Chinese AI infrastructure milestone; significant GPU/compute capacity now operational; indicates accelerating Chinese AI infrastructure deployment.
NVIDIA-backed startup (founded 7 months ago) secures $68 billion compute contract from Anthropic
Massive multi-year capacity commitment validates GPU cloud provider scaling model
Alphabet announced $205 billion capital expenditure program, predominantly for AI infrastructure expansion and compute capacity buildout.
Largest single capex commitment from any hyperscaler for AI infrastructure; likely to strain global GPU supply chains and power grids further.
Data Centers
Volta secured a reported $10B AI compute infrastructure deal with Anthropic for a Norwegian datacenter.
The Norway build signals Anthropic's intent to operate independent inference infrastructure and escape Nvidia/hyperscaler dependency; sovereign AI compute now extends to frontier model labs.
CleanSpark signed a $6.6 billion AI data center lease, its first HPC infrastructure deal.
Asturias, Spain approved a €1.226 billion datacenter project with 3,000 construction jobs, operational by 2030.
Chips & Hardware
AMD reported $11.5B Q2 revenue with data center revenue surging 107% year-over-year, now 58% of total revenue.
AMD's accelerating data center growth signals sustained demand for CPU-based accelerators and GPU-adjacent products; the company is capturing hyperscaler AI workloads previously assumed Nvidia-only.
Anthropic announced ambitions to design and manufacture AI accelerators in-house, joining Nvidia as a chip developer.
Nvidia's B200 GPU systems are sold out across forecasted shipments as demand crushes available supply.
SpaceX posted $7.8B Q1 revenue but revealed $18.4B quarterly capex surge for AI datacenter expansion.
The $18.4B capex spike—among the highest quarterly spend by any company—confirms SpaceX is executing a hyperscaler-scale AI compute build while maintaining aerospace revenue, fundamentally reshaping capital intensity benchmarks.
Subscribe to The Daily Brief
Delivered every morning · Curated across 5 global regions · Real-time breaking alerts