The Easy Money In AI Investing Is Over

The first phase of AI investing was easy, relatively speaking. Buy chip stocks. Buy cloud stocks. Watch them go up. That phase isn't gone entirely, but multiple major investment houses are now saying the next stage of returns is going to require actually picking winners rather than just riding the obvious names.

Here's what fund managers at Eastspring Investments, HSBC Private Bank, and OCBC are all telling their clients right now about where the real opportunity sits from here.

Why AI is genuinely different from past tech cycles

Eastspring Investments makes an interesting distinction worth sitting with: AI isn't just another wave of technology like the internet or mobile were. Those earlier waves changed how people accessed, communicated, and consumed information. AI is different because it changes how people think, decide, and create a much deeper kind of disruption.

It's also uniquely broad in its applicability. Because AI builds on existing digital infrastructure rather than requiring entirely new hardware ecosystems, it's cheaper and faster to deploy across almost any industry, which is part of why adoption has moved faster than previous technology cycles.

But that speed cuts both ways. Eastspring warns that because AI adoption is both faster and broader than past cycles, it compresses the normal market cycle timeline, leading to sharper, quicker swings in expectations and valuations. That makes picking actual winners from laggards more important than it's been in previous tech booms, not less.

The historical warning nobody wants to hear

Eastspring's pulling directly from tech history here, and the lesson is blunt: every major technology cycle has gone through periods where optimism ran ahead of reality. Leadership among companies shifts over time, value becomes concentrated among fewer winners, and periods of excessive optimism create real concentration and valuation risk. That's exactly why disciplined stock selection matters so much right now.

They frame AI not as some standalone investment theme, but as the latest chapter building on everything that came before it, the internet-enabled connectivity and platforms, mobile extended that through devices and app ecosystems, and cloud computing now serves as the actual backbone AI runs on.

Why comparisons to the dot-com bubble don't quite hold up

The dot-com parallel is inevitable at this point, but Eastspring pushes back on it directly. Their read: AI disruption is real, and monetization is increasingly visible, it's just still uneven, concentrated among a relatively narrow group of leading companies rather than spread broadly across the market.

The bigger structural difference, in their view, is around supply and demand. Demand for AI computing continues to outpace available capacity, and hyperscale tech companies are increasingly competing to actually secure production capability rather than competing on price. As Eastspring puts it: "With production capacity now the constraint, it is no longer about price, but about who can deliver quickly and at scale." That's a fundamentally different dynamic than the dot-com era, where companies were often burning cash chasing customers rather than fighting over genuinely scarce physical infrastructure.

Where the actual value is expected to accumulate

Eastspring expects the AI cycle to unfold across multiple phases over many years, rather than as one short-lived investment boom, which they say creates a genuinely broad set of opportunities rather than a narrow window that's already closing.

That said, they're clear about the main risk right now: overpaying at the wrong point in the cycle. Their advice is to focus specifically on companies that are already monetizing AI, or that have clear pricing power and the ability to scale, execute, and compound returns as the broader cycle matures, rather than betting on companies riding pure hype without demonstrated earnings.

They're also pushing investors to think beyond the obvious US-listed tech giants. While the US clearly led AI's creation, Asian supply chain manufacturers have formed the actual backbone enabling AI's progress over the past four years. Manufacturers across Taiwan, South Korea, China, and Japan are expected to stay integral as AI moves beyond generative applications into what's being called agentic and physical AI. Understanding where value actually gets created across that supply chain, and how Asia specifically captures it, is core to their thesis.

HSBC's take: look for earnings, not just infrastructure spending

HSBC Private Bank and Premier Wealth's chief investment officer, Jonathan Sparks, shares a broadly similar view that the AI trade is broadening beyond just chips and cloud infrastructure toward companies that can actually show AI-enabled earnings growth. This earnings season specifically, he says investors will be looking for real proof that AI is cutting costs, streamlining workflows, and improving customer service, not just promises about future potential.

His framing is direct: companies delivering tangible productivity gains, margin expansion, and stronger free cash flow should see higher valuations and outperform going forward. That's a meaningfully different bar than "this company mentions AI in its earnings call."

Sparks also sees Asia as central to the AI hardware story specifically, spanning semiconductors, advanced packaging, memory, and the broader infrastructure powering AI development. Rising AI capital expenditure globally is boosting demand for Asian suppliers across chips, components, data centers, connectivity, and power networks creating what he describes as a broader, more lasting investment opportunity that also helps diversify portfolios beyond concentrated bets on US tech names.

The electricity angle most investors are missing

Here's a genuinely underappreciated part of the AI story that HSBC's flagging directly: AI and broader electrification are driving a multi-year surge in global electricity demand, fueled by power-hungry data centers, electric vehicles, industrial automation, and the wider energy transition happening alongside all of this.

This connects directly to something we've covered extensively the massive AI data center buildout that's been reshaping energy markets, from Google's huge solar and battery investments to xAI's unpermitted gas turbines. Sparks sees real opportunity specifically in power infrastructure: grid upgrades, transmission systems, substations, transformers, storage, and energy management sectors positioned to support resilient earnings well through the coming decade, essentially riding AI's energy demands without needing to bet on any single AI company directly.

OCBC's warning: software companies are facing real disruption risk

Not every corner of the AI trade is a clear buy, and OCBC Group's chief investment strategist Eli Lee is specifically flagging software as an area requiring real caution. While software earnings have stayed resilient in the near term, valuation multiples have already dropped significantly from their 2025 peak, driven by concerns about vulnerable long-term terminal values as AI increases competitive pressure and erodes the moats that used to protect established software companies.

Lee expects that competitive pressure will stay elevated, particularly as AI-native companies pursue public listings and accelerate rolling out large language model-based applications that directly compete with incumbent software providers. His guidance is to stay selective within software specifically, and to be prepared for a potentially extended period of weakness and volatility across vulnerable sub-segments.

There's one clear exception in his view: cybersecurity. Lee sees cybersecurity as one of the more resilient software segments precisely because AI increases both security risks and regulatory requirements simultaneously creating structural tailwinds rather than headwinds for the industry. He expects larger cybersecurity firms specifically to benefit from industry consolidation, while newer AI-native security companies are more likely to complement existing providers rather than outright replace them.

The shift toward AI inference as the next opportunity

Beyond software, OCBC sees the broader AI investment opportunity diversifying as demand shifts toward inference-related technologies the actual running of AI models in production, as opposed to training them. Lee views the broadening of the AI trade toward inference beneficiaries (application-specific integrated circuits, CPUs, memory, storage, networking) as a genuinely positive development, since it lets investors participate in AI's growth with real diversification across more sub-segments and companies, rather than concentrating everything in a handful of dominant chip names.

That said, Lee isn't giving inference-related stocks a blank check either. He cautions that investors need to factor valuations carefully into their risk-reward assessment here too, since recent strong momentum in this space could already suggest stretched positioning.

His more specific views: semiconductor foundries remain well-positioned as long as overall computing demand keeps growing, and major internet companies look attractive right now, partly because their share prices have underperformed relative to semiconductor stocks recently, while they retain genuine potential to benefit from AI applications both internally and in customer-facing products.

Across three different major investment houses, the message converges on a fairly consistent theme: the phase of AI investing where you could just buy the obvious chip and cloud names and profit is giving way to something that requires real discernment. Real earnings, real productivity gains, real pricing power, and genuine diversification across the broader supply chain spanning Asian manufacturers, power infrastructure, and inference-specific technology are what's expected to separate winners from laggards going forward.

The consistent warning running through all three views is the same one: valuations across parts of this trade have already run hot, and the risk of overpaying at the wrong point in the cycle is real. The opportunity hasn't disappeared. It's just gotten a lot less obvious about where exactly it's hiding.