Cloud Infrastructure & Data Processing

Computing Infrastructure Providers (NAICS 5182) covers the cloud-computing, data-processing, web-hosting, and colocation industries — the physical and virtual infrastructure that powers digital services. The hyperscaler segment is dominated by Amazon Web Services (~$110 billion annual revenue), Microsoft Azure (~$80 billion+ from cloud services), Google Cloud (~$45 billion), and Oracle Cloud (~$25 billion), with IBM, Alibaba Cloud, Tencent Cloud, and Huawei Cloud as significant secondary players. A new wave of AI-specialized clouds (CoreWeave, Lambda Labs, Crusoe Energy, Nebius/Nebius Group, Voltage Park) has emerged to serve AI training and inference workloads at competitive GPU rates. The colocation segment (Equinix, Digital Realty, CoreSite) provides the underlying real estate. Combined global cloud-infrastructure revenue exceeds $300 billion annually and is the fastest-growing segment of the broader IT industry, driven primarily by AI workload demand.

On May 4, 2026, both frontier US labs — Anthropic and OpenAI — simultaneously launched finance-industry-backed enterprise AI services joint ventures using a Palantir-style forward-deployed engineer model that embeds lab engineers directly inside customer organizations. Anthropic's joint venture is backed by Blackstone, Hellman & Friedman, and Goldman Sachs (plus Apollo, Sequoia, GIC, General Atlantic, and Leonard Green); OpenAI's parallel "Development Company" raised $4 billion at a $10 billion valuation from TPG, Brookfield, Advent, and Bain Capital, alongside 15 other investors. The dual launches mark a structural shift in how cloud-resident frontier models are distributed to mid-market customers: rather than selling consumption-based API access alone, the labs are now also selling integration services as a separate, capitalized business — a pattern that may compete with the systems-integrator and consultancy partners that hyperscalers have historically relied on for large-deal motion.

AI infrastructure is rebuilding the cloud industry in real time. Hyperscalers are spending unprecedented capital on GPU clusters — Microsoft and Meta announced over $80 billion each in 2026 capital expenditure, primarily for AI infrastructure. AWS Trainium and Inferentia, Azure ND-series GPU instances, Google Cloud TPU pods (now at the v6 generation), and Oracle Cloud's NVIDIA GB200 deployments anchor the major-cloud AI offerings. Specialized AI clouds (CoreWeave, Lambda Labs, Crusoe) have raised tens of billions in equity and debt to build NVIDIA-only GPU clusters tuned for foundation-model training. Inference-optimized clouds (Cerebras Inference, Groq Cloud, Together AI, Baseten) compete on tokens-per-second and cost-per-token rather than raw FLOPs. Beyond the GPU layer, every major cloud now offers managed AI platforms (AWS Bedrock, Azure AI Foundry, Google Vertex AI, Oracle OCI AI) that abstract foundation-model access for enterprise customers. Power, cooling, and water availability have become binding constraints on deployment.

Anthropic's compute supply diversification has produced the largest disclosed AI cloud contract to date: a three-year lease of 300 megawatts and over 220,000 NVIDIA GPUs at xAI's Colossus 1 data center in Memphis (now a SpaceX subsidiary after the February 2026 merger) at $1.25 billion per month through May 2029 — worth over $40 billion in total — surfaced via SpaceX's S-1 filing. The deal sits alongside Anthropic's existing AWS and Google Cloud relationships, and signals that frontier labs are diversifying away from single-cloud relationships and now buying compute from direct competitors' parents — a structural shift toward neocloud economics where labs without consumer-product traction monetize their training clusters by renting them to rival labs. Cerebras Inference, Groq Cloud, Lambda Labs, CoreWeave, and now xAI are competing along the same axis.

Cloudflare announced in May 2026 that it is cutting more than 1,100 jobs (roughly 20% of headcount) despite a record Q1 — CEO Matthew Prince explicitly framed it as a structural shift to "the agentic AI era" rather than a cost-cutting exercise, citing internal AI usage up over 600% in three months. As a profitable infrastructure company attributing the cut to internal AI productivity rather than revenue pressure, Cloudflare becomes the cleanest market signal yet that the cloud sector itself is restructuring around agentic-AI-driven workflows.

The AI build-out is the most capital-intensive infrastructure cycle the cloud industry has seen — comparable in scale to the original cloud build-out of the 2010s but compressed into 3-5 years. Site-selection, power-procurement, and grid-interconnect specialists have become bottleneck roles; many planned data centers face multi-year delays waiting for utility power. Hyperscaler operations roles are growing in tandem (AWS, Azure, Google Cloud each hired tens of thousands of new infrastructure engineers in 2024-2025). Specialized roles in AI-infrastructure software (CUDA programming, distributed-training engineering, GPU-cluster ops) command compensation premiums. Traditional data-center roles (hardware techs, network engineers) are stable but face increasing automation. Sustainability and water-usage concerns drive growth in nuclear-power and on-site-generation specialist roles. The competitive dynamic between hyperscalers and AI-specialized clouds remains in flux as model providers (OpenAI, Anthropic) build their own infrastructure relationships and own bespoke deployments.