Why AI's best investment returns may already be behind us

What happened

In a May 2026 Bloomberg interview, Pictet Asset Management's Chief Strategist Luca Paolini delivered a verdict that should land hard for anyone still holding AI-themed positions bought at peak euphoria: the easy money era is over. Speaking with Merryn Somerset Webb, Paolini acknowledged that markets remain surprisingly resilient given geopolitical noise, persistent inflation fears, and mounting fiscal risk in the US. But he was clear that this resilience does not mean the next decade will look like the last one.

His core argument: the AI investment boom, strong corporate earnings, and US economic dominance have collectively supported elevated valuations. All three of those props are showing cracks. US exceptionalism is fading. Diversification is no longer optional advice. And investors who anchor long-term return expectations to the 2010–2024 playbook are making a structural mistake.

Why the market reacted this way

Markets are resilient right now partly because they have been forced to price in a new normal rather than a crisis. The AI spending cycle from Microsoft, Alphabet, Amazon, and Meta continues at a scale that has no historical parallel. Microsoft alone committed to $80 billion in AI infrastructure spending for fiscal year 2025, a figure that filters through earnings beats across the semiconductor and cloud supply chain.

But Paolini's warning points to something underneath the earnings headlines. The gap between what AI is spending and what AI is generating in revenue-per-dollar-deployed is still wide. NVIDIA's data center revenue has been extraordinary. The returns on that deployed hardware, measured in enterprise productivity gains, have been slower to materialize. Markets have been willing to accept this gap on faith. That patience has a shelf life.

The deeper structural shift is about US fiscal risk. The US debt load and deficit trajectory have historically been dismissed as abstract concerns. The Congressional Budget Office projects the federal deficit will reach $1.9 trillion in 2025, roughly 6.5% of GDP. That's not an abstract number anymore. It creates a real ceiling on how long US equity valuations can sustain a premium over global peers simply by virtue of being American.

What it means for brand visibility

Here is where the investment thesis intersects directly with GEO strategy, and why this Bloomberg conversation matters beyond portfolio allocation.

If capital becomes more selective and return expectations compress, the AI platforms themselves face a tighter monetization imperative. ChatGPT, Perplexity, Gemini, and Grok have all been operating in a phase of aggressive user acquisition with relatively loose commercial pressure. That changes when interest rates stay elevated and investor patience for unprofitable growth shortens.

When AI platforms shift toward monetization, brand visibility inside those engines stops being a feature and starts being a product. Sponsored citations, ranked answers, and premium placement in AI-generated responses become real revenue lines. Brands that have invested early in source authority and GEO fundamentals will be positioned to negotiate from a place of existing credibility rather than starting from scratch inside a pay-to-play system.

The analogy to early Google is imperfect but instructive. Brands that built organic authority before AdWords dominated the conversation had structural advantages that paid-only competitors could not easily replicate. The same dynamic is forming now inside AI engines.

Tools like winek.ai are measuring exactly this: which brands hold citation authority across ChatGPT, Perplexity, Gemini, and Claude before the monetization wave arrives. The brands with high GEO scores today are building the equivalent of domain authority in 2004.

Winners and losers

If Paolini is right about lower returns and fading US exceptionalism, the winners are not the companies with the biggest AI marketing budgets. They are the companies with the clearest, most authoritative information presence across AI engines.

Winners:

• Enterprise software brands that have invested in structured, citable content. When AI engines answer procurement questions, they pull from dense, specific sources. Vague brand messaging loses.
• Global diversified brands. If US exceptionalism fades and capital flows toward European, Asian, and emerging market equities, brands with strong non-English AI visibility gain disproportionate reach. The brands that have solved multilingual GEO early own this arbitrage.
• Niche specialists. AI engines consistently cite category-specific authorities over generalists. As competition inside AI answers intensifies, depth beats breadth.

Losers:

• AI infrastructure plays at stretched valuations. NVIDIA, ASML, and the broader semiconductor complex have priced in years of sustained capex growth. Any slowdown in hyperscaler spending hits multiples hard.
• Brands relying on paid search moats. If AI search continues eroding traditional click-through rates, brands that never built organic AI presence face a double compression: lower traffic and higher cost to replace it. Zero-click search dynamics are already hitting specific industries harder than others.
• Late-stage AI model companies without clear monetization. The funding environment Paolini describes is not friendly to Series C and D rounds for AI companies still in the "build trust, monetize later" phase.

By the numbers

• $80 billion is what Microsoft committed to AI infrastructure spending in fiscal 2025 alone (Microsoft Blog, January 2025). This is the single largest annual capex commitment by any technology company in history, and it sets the floor for what "serious AI investment" means at scale.

• $1.9 trillion is the projected US federal deficit for 2025, approximately 6.5% of GDP (Congressional Budget Office, 2025). This is the fiscal risk that Paolini flags as a structural ceiling on US equity premium valuations.

• $200 billion+ in combined AI capex is expected from the four major US hyperscalers (Microsoft, Alphabet, Amazon, Meta) in 2025, according to analyst estimates compiled by BrightEdge Research. The question is not whether AI investment is real. The question is what the return timeline looks like.

• 6.3x is the estimated valuation premium of US equities over global developed market peers as of early 2026, measured by cyclically adjusted price-to-earnings ratios (Research Affiliates, 2025). Paolini's thesis hinges on this premium compressing toward historical norms.

• 57% of enterprise AI budgets in 2025 are allocated to infrastructure and model licensing, with only 18% going to content, training data, and knowledge management (Gartner, 2025). This imbalance is exactly why brand visibility inside AI engines remains underfunded relative to its strategic importance.

What to watch next

Four signals worth tracking closely over the next six to twelve months:

1. Hyperscaler capex guidance in Q3 2025 earnings calls. If Microsoft, Alphabet, or Amazon begins trimming AI infrastructure spending projections, it signals that the ROI timeline is stretching. That triggers a repricing across the entire AI investment chain.

2. Perplexity and ChatGPT monetization moves. Any announcement of sponsored citations, brand placement products, or answer ranking APIs inside major AI engines is a direct signal that the GEO window for organic authority building is narrowing.

3. Non-US AI platform emergence. Paolini's diversification thesis maps to AI search: if capital flows toward European and Asian markets, regional AI engines (and brand visibility within them) become material. DeepSeek's trajectory in Asian enterprise markets is a leading indicator here.

4. Spread between AI spend and enterprise productivity data. McKinsey, BCG, and Gartner all publish quarterly surveys on enterprise AI adoption and measured productivity gains. When those numbers start matching the infrastructure investment narrative, the valuation debate shifts. Until they do, Paolini's caution is the more defensible position.

The investment thesis for AI was always going to evolve from "bet on the infrastructure" to "bet on who captures the value." That transition is happening now. For brands, capturing value in AI means being cited, trusted, and visible inside the engines where decisions get made. The window to build that organically is open. It will not stay open indefinitely.

Frequently asked questions

Q: Why does Luca Paolini say the easy money era in AI investing is over?

A: Paolini argues that the initial phase of AI investment, driven by infrastructure spending, US economic dominance, and earnings surprises, has already been priced into markets. Going forward, investors need actual productivity returns from AI deployment to justify current valuations. That ROI is materializing more slowly than the capital flowing in, which compresses future return expectations.

Q: What does US exceptionalism fading mean for tech stocks?

A: US tech stocks have traded at a significant valuation premium over global peers for over a decade. If fiscal deficits, geopolitical risk, and slowing AI ROI compress that premium, the repricing is not limited to a few names. It affects the entire index-heavy, US-tech-concentrated portfolio that most retail and institutional investors hold.

Q: How does a tighter investment environment affect AI search platforms?

A: When capital becomes more selective, AI platforms face stronger pressure to monetize their user bases faster. This means the current model of organic, unpaid brand citations inside AI-generated answers is likely to evolve toward a hybrid model with paid placement options, similar to how Google evolved from purely organic results to a blend of paid and organic.

Q: Which brands are best positioned if AI platform monetization accelerates?

A: Brands that have already built strong citation authority across multiple AI engines, through authoritative content, structured data, and consistent GEO investment, will negotiate from a position of existing credibility. They can participate in paid models without being entirely dependent on them, which lowers their cost of visibility.

Q: Is AI infrastructure investment still a good bet despite Paolini's warning?

A: The physical infrastructure buildout (data centers, power, semiconductors) is likely to continue regardless of near-term ROI debates. The risk is in the multiples, not the business. Companies like NVIDIA are genuinely profitable. The question is whether their current valuations already reflect a decade of growth, leaving limited upside from here.

Q: How should brands use the current window before AI search monetizes?

A: Treat it like the pre-AdWords era of Google. Build organic authority now through consistent, citable, structured content across the topics your brand needs to own. Measure visibility across ChatGPT, Perplexity, Gemini, Claude, and Grok using tools like winek.ai. The brands with the highest GEO scores when monetization arrives will have the most negotiating leverage.