Morgan Stanley Warns AI Compute Demand Is Three Times Outpacing Supply

AI Demand Explosion Outstrips Supply

Morgan Stanley has released a sweeping research report warning that the artificial intelligence revolution is accelerating far beyond what financial markets have anticipated. The investment bank finds that demand for computing power is growing approximately three times faster than NVIDIA's forecasted supply growth rate, signalling a persistent and intensifying chip shortage across the AI industry.

The report highlights what Morgan Stanley calls a non-linear leap in AI capabilities as a defining theme of twenty twenty-six. Data from the OpenRouter platform shows global weekly token usage surged from six point four trillion in early January to twenty-two point seven trillion by March, an increase of roughly two hundred and fifty percent in just three months. Several major AI providers have been forced to impose usage caps to manage the explosive demand.

Infrastructure Under Strain

The supply crunch extends well beyond semiconductors. Morgan Stanley projects that United States data centres will face a fifty-five gigawatt electricity shortfall between twenty twenty-five and twenty twenty-eight, up from earlier estimates of thirty-six gigawatts last October and forty-seven gigawatts in February. The bank estimates nearly two point nine trillion dollars in global data centre construction will be needed through twenty twenty-eight to keep pace.

GPU rental prices are already climbing sharply. The Neo Cloud B200 index rose twenty-two percent over the past three months, whilst the H100 index climbed twenty percent, reflecting intense competition for scarce compute resources.

Early Job Market Disruption

The report also documents early but measurable workforce disruption. A survey of roughly one thousand executives across five countries found that the most AI-exposed industries experienced an eleven percent cut in jobs due to AI over the past twelve months, with another twelve percent of positions left unfilled. After accounting for eighteen percent new hiring, the net job reduction stood at approximately four percent. Morgan Stanley estimates that ninety percent of occupations will eventually be affected by AI automation to some degree.

A separate tracker found that AI currently adds at most zero point one percentage points to aggregate unemployment, though the impact is concentrated among younger workers in high-exposure roles.