Morgan Stanley Investment Management has drawn a new map of the AI trade for investor readers, and the center of gravity has moved. The firm recently laid out 10 key truths about AI, saying the story is no longer just about chips but about the layers of supply, infrastructure and usage that make the technology work.
The timing matters because the money is already flowing faster. MSIM says roughly $2.3 trillion has been committed to AI capex since 2017, and token consumption grew by more than 10 times in 2025 alone. That is why the keyword investor is showing up so often now: the debate is no longer whether AI is big, but where the next indispensable piece will be found.
MSIM argues that the old semiconductor playbook is not enough. The bottlenecks started with chips, then shifted to power, memory, networking and cooling, which means the pressure points have spread deeper into the stack. In that sense, the firm is telling investors to think less about a single winner and more about a sequence of layers that become critical as demand moves through them. The same logic appears in its view of data centers, which it calls modern-day factories with tokens as the output.
The firm is also pushing the time frame far beyond the current cycle. At the present rate, it says AI systems will be 250 times more powerful by 2028, a pace that sounds almost mechanical until MSIM adds the warning that history offers no reliable template for compounding at this rate. That is the friction inside the bullish case: the numbers point to an extraordinary buildout, but the map for valuing it is still being written.
For investors, that leaves a tougher question than simply owning AI exposure. MSIM says the next wave of value will come from AI usage, inference, orchestration, applications and workflows that can produce more durable recurring revenue, while the winners in vertical SaaS will be the ones that control data, domain and distribution. It also says AI is shifting from being reactive to becoming autonomous, so the investment case is moving from software that answers prompts to systems that act, including robotics, autonomous vehicles, drones and industrial automation. That is why the firm describes AI as a cross-sector theme spanning infrastructure, models, applications, robotics and power.
The clearest takeaway is that the next AI winner may not be the one most people are watching now. MSIM’s message is to identify the indispensable layer before consensus does, but it does not name that layer. That is the gap investor readers have to live with: the opportunity is broad, the bottlenecks are changing, and the real edge may belong to whoever sees the next constraint before the market prices it in.
