The AI industry is currently fueled by a massive, unsustainable bubble built on speculative data center projects and unproven productivity claims.

Key points

• Data center mirage: While the industry touts hundreds of gigawatts of "planned" capacity, only a small fraction (roughly 5GW globally) is actually under construction. Most projects are speculative land deals or permit applications that lack the power infrastructure to ever go online.
• The GPU bottleneck: NVIDIA is selling hundreds of billions of dollars in chips, but these GPUs are being purchased years in advance for data centers that are nowhere near completion. This creates a massive supply glut and raises questions about why companies are buying hardware that will be obsolete by the time it is installed.
• Power and construction delays: Building data centers is a slow, complex process that requires years of permitting and power grid upgrades. The industry is currently struggling to meet the extreme power demands of AI, with many projects stalled indefinitely.
• "Tokenmaxxing" and software rot: Major tech companies are forcing employees to use AI coding tools to inflate "productivity" metrics. This has led to a culture of "vibe coding," where workers generate massive amounts of low-quality, unmaintainable code, resulting in security breaches and service outages.
• Lack of profitability: Despite the hype, there is little evidence that AI compute is generating significant revenue. Most AI startups and hyperscalers are losing billions, and the actual demand for AI services appears to be weak.

The AI industry is currently burning through vast amounts of capital on infrastructure that may never be finished and software practices that degrade product quality. This suggests a massive, impending financial correction as the reality of slow construction timelines and lack of genuine AI utility finally catches up to the hype.