AI Compute

Sector thesis

AI Compute is the hardware and infrastructure that powers artificial intelligence—the chips, data centers, and networking equipment that train and run AI models. Right now, it's one of the most capital-intensive sectors in tech because every company building AI needs somewhere to run it, and the demand has outpaced supply. The megatrend is simple: AI is moving from research labs into real products. That requires enormous amounts of computing power. Data centers need specialized chips (mostly GPUs and custom processors), cooling systems, and power infrastructure. This isn't hype—it's a genuine structural shift in how computing resources are allocated. Within AI Compute, there are three main buckets. First: semiconductor makers who design and manufacture the chips themselves—this is the most capital-intensive and competitive part. Second: data center operators and cloud providers who buy chips and rent computing power to AI companies. Third: infrastructure suppliers—companies making cooling systems, power management, networking gear, and other supporting hardware. The biggest risks are real. Chip demand could slow if AI adoption stalls or if companies build their own custom chips instead of buying from established makers. Margins can compress as competition increases. There's also geopolitical risk—export restrictions on advanced chips create uncertainty. And frankly, some of the valuations in this space assume growth that may not materialize. For a retail portfolio, this sector works best as a thematic position rather than a single bet. You might own a diversified tech fund that includes compute exposure, or pick one or two established players with proven revenue. Watch for quarterly earnings reports showing whether AI compute demand is actually translating to sales, not just hype. Look for signs of oversupply in data centers or margin pressure—those are warning flags. This is a real trend, but it's also crowded, so patience and selectivity matter.