AI Race: Power and Infrastructure Now More Critical Than Chips
Morgan Stanley warns that a massive power shortfall could hinder the AI boom, favoring tech giants over smaller players.
The global competition to dominate artificial intelligence (AI) is shifting. While the world has been obsessed with the "chip wars," a new report from Morgan Stanley suggests that the real bottleneck isn't the hardware itself, but the electricity and infrastructure needed to run it. The ability to secure massive amounts of power and build data centers is becoming the defining factor in who wins the AI race.
The Power Gap: A Looming Crisis
In an equity research report titled "What Does the Power Shortfall Mean for Semis?", Morgan Stanley warns that the rapid investment in AI is moving faster than the energy grid can handle. The report highlights a stark reality: companies are producing more semiconductors than there are functional places to plug them in.
The numbers are alarming. Morgan Stanley estimates that data-center developers in the US could face a 32-gigawatt (GW) power shortfall by 2028. This gap represents roughly 34 per cent of the total projected power requirements, even when accounting for new power-generation solutions. Looking further ahead, this cumulative shortfall could balloon to 68 GW by 2029.
Big Tech vs. The Rest: The 'Squeeze Out'
This energy crisis is creating a divide between the industry titans and smaller innovators. Access to electricity is increasingly becoming a privilege of the wealthy. Because power providers prefer "credit-worthy" players with predictable revenues, the largest technology companies are securing energy contracts more quickly.
The report predicts a massive concentration of power:
- 2026: Major tech firms and specialist cloud providers could control 94 per cent of global data-center capacity additions.
- 2027: This figure is expected to be around 64 per cent.
- 2028: The share is projected at 57 per cent.
Morgan Stanley warns that this trend could effectively "squeeze out" smaller developers who lack the financial muscle to compete for limited energy resources.
The Ripple Effect on Chip Makers
If data centers cannot be powered, they cannot install the chips they have ordered. This creates a dangerous feedback loop for the semiconductor industry. If deployment stalls, customers may begin to push back delivery dates or cancel orders entirely.
However, not all chip makers are equally at risk. Morgan Stanley believes Nvidia and Broadcom are likely to meet their 2027 forecasts due to their deep visibility into customer demand. The higher risks lie with suppliers of optical equipment, memory chips, and other supporting components, who may see a slowdown if infrastructure fails to keep pace.
Asia and India as Potential Havens
With the US grid struggling to keep up, the AI boom may migrate. The report suggests that technology companies might look toward Asian markets, including India, to expand their computing infrastructure.
Morgan Stanley's Asia data-center team estimates that approximately 14 GW of demand from US "hyperscalers" could shift to Asia by 2030. While this represents a significant opportunity for Asian infrastructure, the report clarifies that international expansion is a strategic move rather than a total solution to the power problems facing the US.
"Nvidia confirmed our thesis that there is not enough power to meet demand, and there are more semiconductors being built than there are places to put them."