A historic surge in artificial intelligence spending defined 2025, with capital pouring into chips, data centres, power systems and cloud infrastructure. As that investment wave matures, attention is shifting toward returns, balance sheets and where profits ultimately accrue.
BNN Bloomberg spoke with Nicholas Mersch, portfolio manager at Purpose Investments, about why 2026 could mark a turning point for AI investing, how rising debt is changing investor expectations and which parts of the ecosystem may benefit next.
Key Takeaways
- The AI buildout reached an inflection point in 2025, with capital spending approaching levels rarely seen in modern economic cycles.
- While last year’s rally lifted most participants, 2026 is expected to separate sustainable AI winners from weaker, overextended players.
- Growing use of debt financing is forcing investors to demand clearer paths to monetization and return on investment.
- Near-term value is likely to accrue to hardware bottlenecks, hyperscalers with scale and balance-sheet strength, and power-constrained infrastructure.
- AI adoption beyond technology — including healthcare, logistics and other industries — is emerging as a key driver of future productivity gains.
Read the full transcript below:
ROGER: As we approach the end of 2025, let’s get a recap of the year and a look ahead at what 2026 may bring and what our guest is watching for. We’re joined by Nicholas Mersch, portfolio manager at Purpose Investments. Nicholas, thanks very much for joining us.
NICHOLAS: Thanks so much for having me.
ROGER: All right, let’s look back first. It’s the way we normally do it. For you, what defined 2025? What were some of the key drivers?
NICHOLAS: I think flipping the page on the calendar year is always a great time to go back and ask what happened in 2025, what worked, what didn’t, and then look ahead. For me, the core theme of 2025 was really the inflection point in a multi-year capital spending cycle that’s helping drive what could be the next industrial revolution, centred around artificial intelligence.
If you think about the last decade, many mega-cap companies became very good at scaling asset-light business models, driven by high gross margins and low fixed costs in software. Now the game has changed. This is an arms race to build the best AI systems, and it’s being driven by physical capital expansion. That means the economy over the next 10 years is likely to look far more capital-intensive than it did over the past decade.
We saw the entire value chain start to take off, including semiconductors, cooling systems, power units and cloud providers. Aggregate buildout figures for 2025 are expected to approach roughly half a trillion dollars in capital spending. What was equally surprising was not just the magnitude of that spending, but how quickly the goalposts kept shifting quarter over quarter, with more and more capital required.
At the same time, we started to see very real concerns around debt financing and circularity. We saw Blue Owl, one of the private lenders, walk away from an Oracle financing, and we’ve seen credit spreads on Oracle widen sharply over the past few quarters. As a result, many higher-beta stocks in the space began selling off in November, and we’re seeing some of that again today.
So while 2025 was the hallmark year for AI investment, we’re now turning the page and asking for returns on investment to justify the capital spending. There are some cracks in the system, but we’re still in the early stages of building this out, and it remains an exciting setup heading into next year.
ROGER: Looking ahead now, let’s talk about capital spending. Is there still room for it, or do investors really want to start seeing returns?
NICHOLAS: I think there’s still some room to run on the capital spending side, but 2026 is shaping up to be a real “show-me” year that separates AI winners from losers. In 2025, a rising tide lifted all boats. In 2026, we’re likely to see that sorting process begin.
My focus remains on the enablers of AI — the companies providing the physical and digital layers that power these systems. That includes the picks-and-shovels players: chips, memory, networking and infrastructure. I still think there’s room to run in many of those names.
If you look at Nvidia, for example, its Blackwell chips remain sold out, and GE Vernova is booked out for years on natural gas turbines. These are critical bottlenecks in the ecosystem, and bottlenecks often come with pricing power. That said, they don’t last forever.
At some point in this technology cycle, the focus will shift from the enablers to the adopters — companies that use AI to increase efficiency by embedding it into workflows and expanding operating margins beyond just technology-focused firms. Industries such as pharmaceuticals, through drug discovery, and biotechnology, through genome sequencing, are examples.
I also like companies that sit at the intersection of enabling and adopting. Names like Microsoft, Google and Amazon are vertically integrated across the ecosystem, from silicon to applications. They capture value across the chain and are expected to remain free-cash-flow positive next year, even with heavy capital spending. They have the balance sheets to finance this buildout, whether it becomes a winner-take-most environment or turns into a defensive strategy where products are bundled aggressively.
The market has recognized Google’s performance this year, but Amazon and Microsoft have lagged the benchmark significantly. I think those names have potential to catch up.
ROGER: Does it feel like 2026 could be the year we see clear winners and losers?
NICHOLAS: Yes, absolutely. We’re already starting to see that play out. In the loser category, companies with too much leverage stand out. Oracle is a good example. It carries a significant amount of debt, and after its earnings report, the stock sold off sharply.
On top of that, Oracle disclosed roughly US$250 billion in lease commitments that begin before 2028, which don’t show up as traditional balance-sheet debt. When companies keep building capacity while taking on excessive leverage, it raises red flags across the ecosystem.
Oracle initially rallied on strong remaining performance obligations, or RPOs, which are essentially future bookings, reaching roughly half a trillion dollars. Much of that was tied to OpenAI. But investors are now questioning where the financing comes from, given OpenAI’s lack of a free-cash-flow-generating business, despite its ability to raise capital or potentially pursue an IPO.
What investors ultimately want are companies that can finance growth through free cash flow or cash on hand, rather than relying heavily on external funding.
ROGER: Stepping a bit outside of technology, are there other sectors that are early adopters of AI and starting to see benefits?
NICHOLAS: Yes. Supply chain and logistics companies are a good example. Many are integrating AI into inventory management and routing systems, driving major efficiency gains through predictive logistics.
Healthcare is another area, particularly in drug development, where AI is helping accelerate early-stage research and clinical processes. In biotechnology, genome sequencing is another promising application where AI is already being embedded into workflows.
What this ultimately does is increase operating leverage. We’ve even heard executives say the quiet part out loud, noting that headcount reductions are partly the result of employees becoming more productive. Over the past decade, revenue per employee across the S&P 500 was relatively flat. Since the launch of ChatGPT, revenue per employee has accelerated sharply.
That has implications for the labour market, which is already showing signs of strain in recent economic data. As we move through the second half of the year, I think investors need to focus on companies that are adopting AI most effectively and translating it into measurable productivity gains.
ROGER: We’ll have to leave it there. Nicholas, thanks very much for joining us.
NICHOLAS: Thanks so much for having me.
ROGER: Nicholas Mersch is a portfolio manager at Purpose Investments.
---
This BNN Bloomberg summary and transcript of the Dec. 17, 2025 interview with Nicholas Mersch are published with the assistance of AI. Original research, interview questions and added context was created by BNN Bloomberg journalists. An editor also reviewed this material before it was published to ensure its accuracy and adherence with BNN Bloomberg editorial policies and standards.
