Morgan Stanley Says $3 Trillion in AI Infrastructure Investment Is Coming — Here Is Where That Money Is Actually Going

It is easy to think of the artificial intelligence boom as a story about software, chatbots, and clever new apps. Morgan Stanley's research division wants investors to see something different: a colossal, physical, industrial buildout that is reshaping the global economy in ways more comparable to the railroad era or the original electrification of cities than to any previous software trend. According to Morgan Stanley Research, nearly $3 trillion in AI-related infrastructure investment will flow through the global economy by 2028, and more than 80 percent of that spending has not even happened yet. That is not a forecast about some distant, speculative future. It is a description of capital that is, right now, being committed to chips, buildings, power plants and cooling systems on a scale that is already large enough to move national economic growth figures.

Understanding where that money is actually going, and why, matters for anyone trying to make sense of the AI investment landscape rather than simply reacting to headlines about chatbots and stock price swings. This article breaks down Morgan Stanley's projections, the specific categories absorbing the bulk of this spending, the geopolitical forces accelerating it, and what the bank itself is telling investors to do about it.

A Number Too Big to Treat as a Tech Story

Morgan Stanley's research team has been explicit that artificial intelligence has stopped being simply a technology theme and become what the bank describes as a macro variable, something now significant enough to influence GDP, corporate earnings, credit markets and geopolitics at an industrial scale. The headline figure behind that claim is striking on its own. Morgan Stanley Research estimates roughly $2.9 trillion in global data center construction costs alone through 2028, a figure driven by demand for computing power that continues to vastly exceed the available supply. That spending is not happening in a vacuum disconnected from the rest of the economy. Morgan Stanley estimates this buildout is contributing approximately 25 percent of total US GDP growth this year alone, a remarkable share for a single category of corporate investment to claim within a $29 trillion national economy.

The scale of physical expansion required to support this spending is just as dramatic as the dollar figures themselves. Morgan Stanley's research projects that global data center capacity will need to increase by a factor of roughly six times over the next several years to keep pace with the demands of AI and cloud computing. That is the kind of growth rate more commonly associated with an entirely new industry being built from scratch, not an expansion of existing infrastructure.

Where the Money Is Actually Going

Breaking down Morgan Stanley's nearly $3 trillion figure reveals a buildout with several distinct, equally critical components, and understanding each piece helps explain why this spending wave touches so many different sectors of the economy simultaneously. The most visible category is the data centers themselves, the physical buildings, racks, networking equipment and specialized hardware needed to house and run AI workloads at scale. This category alone, according to Morgan Stanley's estimates, accounts for roughly $2.9 trillion of construction cost through 2028, encompassing everything from the largest hyperscale campuses operated by companies like Microsoft, Amazon and Google to smaller, specialized facilities built by infrastructure developers serving those same hyperscalers under long-term lease agreements.

Sitting underneath that data center spending is an even more fundamental constraint: energy. AI-driven data centers are expected to account for nearly 20 percent of global power demand growth, with annual consumption from this single category approaching the size of Canada's entire national power demand. Morgan Stanley's research separately forecasts that US data center demand could reach 74 gigawatts by 2028, against a projected shortfall of roughly 49 gigawatts in available power access, a gap that is forcing massive new investment in everything from natural gas plants and nuclear power to battery storage and entirely off-grid power solutions built specifically to serve individual data center campuses. This energy bottleneck is not a side issue. Morgan Stanley describes it as simultaneously a constraint and a catalyst, one capable of triggering what the bank calls a trillion-dollar energy buildout in its own right, layered on top of the data center spending itself.

Cooling systems represent a related but distinct category of investment, since the dense, high-powered chips driving modern AI workloads generate far more heat than traditional computing equipment, requiring increasingly sophisticated liquid cooling and thermal management systems rather than the simple air-conditioning approaches that sufficed for earlier generations of data centers. Finally, semiconductors, the chips that actually perform AI computations, sit at the foundation of the entire buildout, with global demand for advanced AI chips continuing to outstrip manufacturing capacity even as companies like Nvidia, AMD and a growing list of custom chip designers race to expand production.

How Geopolitics Is Pouring Fuel on the Fire

One of the more distinctive elements of Morgan Stanley's 2026 analysis is how directly it ties this infrastructure buildout to geopolitical competition, particularly between the United States and China. The bank's research describes competition across chips, compute, energy and data as elevating what it calls the strategic premium on secure domestic infrastructure, meaning governments and corporations alike are increasingly willing to pay more, and move faster, to build AI capacity within their own borders or among trusted allies rather than relying on supply chains that cross geopolitical fault lines.

This is not simply a business decision happening in isolation from politics. Morgan Stanley explicitly frames the AI buildout as a national security imperative as much as a commercial one, one that is already shaping trade flows, investment incentives and the allocation of trillions of dollars in capital over the coming decade. Recent geopolitical flashpoints, including tensions in the Middle East, have only reinforced this dynamic, with Morgan Stanley noting that national security, energy, supply chains and technology are becoming increasingly interrelated considerations for investors rather than separate categories to be analyzed independently. In practice, this means countries are racing to secure their own semiconductor manufacturing capacity, their own energy infrastructure to power AI systems, and their own access to critical materials, with the United States and China each pursuing parallel but distinct versions of self-sufficiency across these same categories.

Morgan Stanley's Four-Part Strategy for 2026

Rather than simply describing the scale of this buildout, Morgan Stanley's research has laid out a specific, four-part recommended strategy for investors navigating 2026, attributed to Stephen Byrd, the bank's Head of Global Thematic Research. The first element of that strategy is to focus on companies positioned to benefit as nations pursue self-sufficiency across energy, critical materials, manufacturing capacity and AI capabilities, reflecting the geopolitical dynamics described above. The second is to invest directly in AI infrastructure itself, the data centers, power generation and supporting physical assets that form the backbone of the entire buildout, rather than focusing exclusively on the software and applications layer sitting on top of that infrastructure.

The remaining elements of Morgan Stanley's guidance extend into less obvious corners of the market. The bank's fixed-income strategists have specifically pointed investors toward opportunities in structured credit and asset-backed financing tied to contracted AI infrastructure, alongside diversification benefits available through highly rated, cash-generative hyperscaler bonds in investment-grade credit markets. This reflects a broader theme running through Morgan Stanley's analysis: as AI capital expenditure continues rising, the bank expects debt financing to play an increasingly important role, particularly for infrastructure-heavy projects that smaller or earlier-stage companies cannot fund entirely through their own cash flow. Morgan Stanley has highlighted its own role advising on landmark transactions of this kind, including a $27 billion structured joint venture with Meta for a US AI data center campus, as evidence of how seriously credit markets are now engaging with AI infrastructure financing.

Underlying all four elements of this strategy is a consistent message from Morgan Stanley's Chief Investment Officer for Wealth Management, Lisa Shalett, who has cautioned investors against simply chasing broad technology exposure and urged them instead to differentiate the genuine winners from companies merely benefiting from association with the AI theme. That distinction matters enormously given how correlated many sector themes have become to the scale and scope of the data center buildout, making genuine portfolio diversification more difficult to achieve even as it becomes more necessary.

Where Ordinary Investors Can Realistically Participate

For everyday investors without access to the private credit markets or structured joint ventures that dominate headlines about the largest AI infrastructure deals, the realistic avenues for participation are narrower but still genuinely meaningful. Publicly traded companies across the categories described above, established semiconductor manufacturers, data center real estate investment trusts, utility companies expanding power generation capacity, and infrastructure developers building data center campuses on long-term lease contracts, all offer direct, accessible exposure to this buildout through ordinary brokerage accounts.

Diversified technology and infrastructure-focused exchange-traded funds offer a lower-effort way to gain exposure across many of these categories simultaneously, without requiring investors to pick individual winners among chip manufacturers, data center builders or utility companies. This approach aligns reasonably well with Morgan Stanley's own caution that picking individual winners in a sector this complex and fast-moving is genuinely difficult, even for professional analysts with access to far more granular data than most individual investors will ever see.

It is worth being equally honest about the risks layered into this opportunity. Morgan Stanley's research itself acknowledges that the AI trend is large enough to trigger valuation resets and sector rotation, and that markets are actively weighing the benefits of this buildout against its potential to disrupt existing workers and industries. The bank's own analysis of 3,600 stocks found that while 21 percent of S&P 500 companies now cite some AI-related benefit, up sharply from just 10 percent in 2024, the market is increasingly distinguishing between companies merely mentioning AI and those actually monetizing it through measurable margin expansion and revenue growth. Investors chasing this theme without that distinction in mind risk paying premium prices for AI exposure that never translates into the kind of cash flow gains Morgan Stanley's research associates with genuine adopters and infrastructure beneficiaries.

The Bottom Line

Morgan Stanley's nearly $3 trillion projection is less a single prediction than a map of an entire industrial transformation already underway, spanning chips, buildings, power plants, cooling systems and the credit markets financing all of it simultaneously. More than 80 percent of that spending still lies ahead, which means the categories described in this article, data centers, energy infrastructure, advanced cooling and semiconductor manufacturing, will likely remain central to both economic growth figures and investment portfolios for years to come, not just through the remainder of 2026. The geopolitical competition accelerating this buildout shows no sign of cooling, and Morgan Stanley's own recommended strategy reflects a belief that the winners from this cycle will be determined less by who simply mentions artificial intelligence and more by who actually builds, finances and operates the physical infrastructure this technology depends on. For investors trying to position themselves within that transformation, understanding exactly where the money is flowing, rather than simply that an enormous amount of money is involved, is the more useful starting point.

*This article is for informational purposes only and does not constitute financial or investment advice. Always conduct your own research or consult a licensed financial advisor before making investment decisions. Data and quotes are sourced from Morgan Stanley Research and Morgan Stanley Institute publications as of June 2026.*