AMD

AMD’s second-quarter record is less interesting as a victory lap than as a change in the burden of proof. Data Center supplied 58% of company revenue, and its $6.7 billion was more than double the prior-year level. Management now guides to roughly $13 billion for the third quarter. Those numbers establish commercial acceleration; they do not yet reveal customer concentration, shipment mix or whether today’s demand can preserve margins as supply and competition move.

AMD will secure more than 500MW of United States data-centre capacity from Core Scientific for customer deployments beginning in 2027, with an ability to expand up to 2.5GW. The agreement gives AMD a physical route for systems built around its chips and software. It does not say that the maximum is committed, built, powered or occupied.

AMD and Anthropic have agreed on an accelerator programme of up to 2GW, beginning with as much as 1GW of MI450 capacity in the first half of 2027. AMD may also invest up to $5 billion as deployment milestones are met. The scale is large, but every headline number is conditional: capacity must be engineered, financed, sourced and switched on before either side receives the economics implied by the announcement.

AMD's Instinct accelerators and ROCm software no longer have to prove that an isolated benchmark can look impressive. The harder production test is whether an AI infrastructure team can take a training or inference workload, make it run on AMD hardware, reproduce the result under known versions, recover when the stack moves, and still save enough money after porting, validation, cloud capacity and fallback plans are counted.

OpenAI’s compute procurement affects hyperscale cloud demand, data centre buildout, GPU supply chains and power infrastructure planning.

Salute says AI data centre growth needs more engineers as liquid cooling and AI Factory deployments raise operational demands.

China plans a $295bn AI data centre buildout led by state telecom carriers and domestic suppliers.

Megaport secures four AI infrastructure deals and plans a $594m raise to build an enterprise inference cloud.