2026 Thematic mid-year update

AI continues to be the key theme of 2026, reshaping industries and influencing nearly every corner of the global economy. Our 2026 Thematic Mid-Year Update features key charts exploring how the rapid advancement and adoption of AI is transforming sectors, creating new opportunities and risks, and redefining where and how we invest.

• AI’s value chain continues to expand as adoption deepens, but winners are beginning to separate amongst supply constraints.
• Rising demand for AI compute is intensifying the need for critical materials and power, exposing bottlenecks along the way but also potentially opening new frontiers like space.
• New capabilities emerging in robotics and healthcare may impact our daily lives, while AI & digital assets may reinforce each other through daily use cases.

The AI investment opportunity is shifting from a concentrated technology buildout toward a broader, increasingly physical economy, with implications and applications for all business sectors and consumers. The next phase of thematic investing may depend on identifying both the beneficiaries of AI adoption and the constraints that determine how quickly the AI economy can scale.

For much of the AI buildout, the investment story has centered on the companies building the infrastructure. But the next phase may become about what that infrastructure enables. We are beginning to see signs that years of AI capex are translating into revenue.¹

Cloud revenue growth for the largest providers has been accelerating as AI infrastructure comes online, showing evidence that demand has begun to be monetized.² In just one year, the combined cloud backlog, which represents signed customer commitments, of Amazon, Alphabet, Microsoft, and Oracle nearly tripled, rising from $800 billion to $2.3 trillion.³

But spending on AI and creating value from AI are not necessarily the same thing. As AI adoption matures, opportunities may increasingly favor companies that can successfully deploy AI across their businesses, not simply those investing in AI technology.

Kellanova offers a real-life example. The company behind the popular Pringles snack, invested $4–5 million in AI to make every chip more consistent (potato chips that is, not to be confused with AI chips). The result? A 10% increase in quality, 13% reduction in waste, and 40% return on investment.⁴

We believe the next set of AI beneficiaries may include not only the companies building the technology, but also those with the strategy, talent and operating models to put it to work. As a result, AI’s economic value may be measured not by how much technology companies buy, but by what they are able to do differently because of it.

That shift may already be spreading through the broader enterprise economy faster than in previous technology cycles. In the first three years after the launch of ChatGPT in late 2022, 29% of Fortune 500 companies and 19% of Global 2000 companies adopted AI.⁵ What may be even more telling is where that adoption is beginning to take hold. Coding may be the obvious starting point, but faster-growing enterprise use cases are also emerging in legal and medical administration, industries that by some measures have been historically slower to adopt new technologies, as evidenced in the chart below:

AI bottlenecks aren’t just limited to semiconductors, demand is extending well beyond chips and data centers to the raw materials that underpin them. Copper, lithium, nickel, and more all sit within broader supply chains supporting electrification, batteries, power infrastructure and advanced technologies. Yet expanding supply is not simply a question of finding more resources.

New mines can face long development timelines, permitting hurdles, declining ore grades and significant capital requirements, creating the potential for supply to lag demand. One example? Copper, where current project pipelines could leave a quarter of 2035 demand unmet.⁹

Did you know?

121 GW: projected U.S. data center IT power demand by 2030.


McKinsey, July 31, 2026.

And copper isn’t the only pain point. The funnel for materials is narrowing at refining; currently the leading refining country accounts for an average of 72% of global refined supply of critical raw materials, including copper, lithium, nickel, cobalt, graphite, and manganese.¹⁰ In our view, not only is more supply needed, but refining must diversify.

Beyond critical materials, the physical demands of AI extend directly to the power system. As data centers scale, electricity demand is beginning to accelerate after decades of relatively muted growth, creating what may be the largest expected step-up in U.S. electricity demand growth in a century.

As the physical constraints around AI become more visible, space is emerging as a potential frontier for solving some of those challenges. Falling launch costs and improving space infrastructure are lowering the barriers to experimentation, enabling companies to test ideas that would have seemed far less practical even a decade ago.

Did you know?

95%: less expensive to launch a rocket to space now versus 65 years ago.


“How Much Does It Cost to Launch a Rocket?,” Orbital Radar, visited August 2, 2026. Figures based on company list prices and data from NASA, FAA, Bryce Tech and CSIS Aerospace Security.

Orbital computing is one example. While the technology remains highly experimental and significant engineering challenges persist, placing computing infrastructure in space could potentially offer access to more abundant and consistent solar energy, while reducing dependence on terrestrial land, transmission and grid connections. Solar panels in space can generate roughly four to ten times more power¹¹ than comparable systems on Earth, and suitable orbits can provide near-continuous sunlight rather than outputs that vary with weather and seasons.

The broader implication extends beyond data centers. As access to orbit becomes cheaper, and more active satellites are launched, entirely new space-enabled markets may become more economically viable.

The technology may still be early, but the combination of falling costs and growing activity is creating a flywheel; cheaper access enables more experimentation, which can support new infrastructure, applications and ultimately new investable opportunities.

Robotics may be one of the clearest examples of AI’s move beyond the digital world and into the physical one. Robotics bring superintelligence into the real world by allowing machines not just to process information, but to perceive, move and act on it. What was once concentrated in controlled industrial settings is increasingly expanding into transportation, manufacturing and, over time, everyday consumer use cases.

Did you know?

30X: the cost to build a humanoid robot has fallen more than 30X in a decade.


Barclays, “Impact Series 14: AI Gets Physical,” January 2026.

The market remains in its early stages, but the breadth of potential use cases points to a vast long-term addressable opportunity.

AI has entered its next chapter, no longer beheld to just a single sector, but instead reshaping industries and influencing nearly every corner of the global economy. As the AI  value chain continues to evolve, the opportunity is broadening beyond the companies building AI to those best positioned to put it to work.

The next phase of AI is also becoming increasingly physical. Rising compute demand is intensifying the need for critical materials and power, exposing bottlenecks along the way. At the same time, falling costs and advancing technology are opening new frontiers across space and defense, expanding the industries and companies participating in the AI buildout.

As AI moves into everyday life, we are seeing new use cases for AI especially where growing demand meets limited resources; from robotics and autonomous systems to disrupting the healthcare system as we know it. Financial markets are also evolving, and AI & digital assets may be able to reinforce each other via daily use cases.

For investors, capturing these themes may increasingly require looking across sectors, understanding where constraints and adoption are emerging, and being selective about where value is ultimately created to fully capitalize upon the AI economy.