OpenAI targets $600B compute spend by 2030

OpenAI is reportedly telling investors it’s targeting roughly $600 billion in total compute spend through 2030. That’s a massive signal for the AI “picks and shovels” trade: chips, networking, servers, and data-centre power and cooling.

What happened

OpenAI is aiming for around $600B in compute spend by 2030 as it lays groundwork for a potential IPO, while also projecting more than $280B in revenue by 2030. The headline takeaway: demand for AI compute is still accelerating, but the market is increasingly focused on the economics of inference, margins, and who controls supply.

Why markets care

1. Compute is the bottleneck: GPUs, memory, and high-speed networking remain tight, and a spend target this large reinforces multi-year capex cycles.

2. Data centres are the new “factories”: power delivery, cooling, and grid equipment become critical constraints.

3. Margin pressure shifts winners: if inference costs rise, it can push more optimisation and custom infrastructure buildouts, which changes who captures the profit pool.

Winners

AI chips and memory

A $600B compute budget implies persistent demand for accelerators and the memory stack that feeds them (HBM/DRAM), plus continued tight supply dynamics.

Names: $NVDA (NVIDIA), $AMD (Advanced Micro Devices), $MU (Micron Technology)

High-speed networking and interconnect

Training and inference at scale require ultra-fast networking (switches, optical, silicon) to keep expensive GPUs utilised. As clusters grow, networking becomes a bigger slice of total AI system cost.

Names: $ANET (Arista Networks), $AVGO (Broadcom), $MRVL (Marvell Technology)

Data-centre power, cooling, and electrical equipment

Why this group wins: Bigger AI clusters mean higher power density per rack and tougher cooling requirements. Suppliers tied to UPS systems, thermal management, switchgear, and electrical buildouts can see rising order books.

Names: $VRT (Vertiv), $ETN (Eaton), $PWR (Quanta Services)

Losers

Legacy IT services and “hours-based” consulting

Why this group loses: As AI budgets shift toward compute and automation, some traditional consulting and IT services can face pricing pressure, slower growth, or project deferrals as clients try to “do more with less labour.”

Names: $ACN (Accenture), $IBM (IBM)

Smaller or highly levered cloud/hosting players

Why this group loses: When hyperscale AI spend ramps, it can widen the moat. Smaller providers may struggle to match capex, source advanced hardware, or compete on AI-ready infrastructure pricing.

Names: $RXT (Rackspace Technology), $AKAM (Akamai Technologies)

Software names with heavy AI inference costs

Why this group loses: If inference costs remain high and usage scales faster than pricing power, some app-layer companies can see gross margin pressure until they optimise models, renegotiate compute, or raise prices.

Names: $DUOL (Duolingo), $SHOP (Shopify)

Quick trading lens

* If this theme stays hot, watch for follow-through in “infrastructure beneficiaries” (chips, networking, power/cooling) on strong volume days.

* If the market pivots to “AI profitability,” be selective: the best setups often appear in companies selling essential infrastructure with clearer near-term revenue capture.

#StockMarket #Trading #Investing #DayTrading #SwingTrading #AI #ArtificialIntelligence #OpenAI #DataCenters #Semiconductors #GPUs #Networking #CloudComputing #PowerInfrastructure #Earnings #MarketSentiment

The $600 Billion Compute Blueprint: Winners of the AI Supercycle

Powered by Wave AI