We use cookies that are necessary to make our site work as well as analytics cookie and third-party cookies to monitor our traffic and to personalise content and ads.
Please click “Cookies Settings” for details on how to withdraw your consent and how to block cookies. For more detailed information about the cookies we use and of who we work this see our cookies notice
Necessary cookies:
Necessary cookies help make a website usable by enabling basic functions like page navigation and access to secure areas of the website and cannot be switched off in our systems. You can set your browser to block or alert you about these cookies, but some parts of the site will then not work. The website cannot function properly without these cookies.
Optional cookies:
Statistic cookies help website owners to understand how visitors interact with websites by collecting and reporting information
Marketing cookies are used to track visitors across websites. The intention is to display ads that are relevant and engaging for the individual user and thereby more valuable for publishers and third party advertisers. We work with third parties and make use of third party cookies to make advertising messaging more relevant to you both on and off this website.
The AI cycle beyond the hype: earnings, bubble risks, and positioning
Artificial intelligence remains one of the market’s defining themes. Now, the conversation around the AI cycle is becoming more exacting, as an exuberant first phase – driven by ChatGPT and the rerating of semiconductor and platform leaders – gives way to harder questions.
Are current valuations still justified by earnings? Does the AI build-out reflect genuine demand rather than over-enthusiasm? How durable is hyperscaler spending? And, following the sharp divergence between semiconductors and software, where should investors now position themselves?
Sign up for our newsletter
These questions framed our webinar The AI cycle: Earnings reality, bubble risks, and preferred positioning, held on 11 March 2026. Moderated by John Woods, CIO and Head of Investment Solutions for Lombard Odier Asia, the discussion brought together Corinne de Boursetty, Investment Strategist – Sustainability and Research, and Jian Bo Gan, Equity Research Team Head, Asia.
In his opening remarks, John Woods explained that the session was designed to explore “the insights and opportunities shaping today’s AI-driven markets”. The task was not to amplify the noise around AI, but to test the theme against the variables that matter most to investors: valuation, adoption, capex, funding quality, regional differentiation, and portfolio positioning. Together, the speakers revealed why AI is “not a bubble (yet)”, why China deserves closer attention relative to US technology, and how to invest across the AI value chain.
The AI ecosystem, is not confined to technology alone. It spans information technology, communication services, and parts of consumer discretionary, together accounting for roughly 35% to 40% of the global equity market
A broader market story
Corinne de Boursetty began by widening the frame. The AI ecosystem, she argued, is not confined to technology alone. It spans information technology, communication services, and parts of consumer discretionary, together accounting for roughly 35% to 40% of the global equity market. As a result, AI is already influencing several of the market’s most important profit pools, as the MSCI All Country World Index (ACWI) shows. Here, communication services returned 30% and information technology 26% since January 2025, versus 25% for the MSCI ACWI overall. Meanwhile, consumer discretionary lagged at 4%, held back by tariff concerns and slower-growth fears.
While that divergence has helped fuel bubble concerns, Corinne de Boursetty’s central point was that strong price performance alone doesn’t prove speculative excess. Instead, to account for earnings, end-demand, and balance-sheet quality, bubble risk should be assessed across five main categories: valuation, the AI adoption rate, capital expenditure on hyperscalers, funding quality, and energy availability.
On valuation, the tone was measured. The presentation showed the MSCI US IT Index trading at 23 times 12-month forward earnings – well below the levels reached during the dot-com era. At the same time, global technology earnings are expected to grow by 39% in 2026 compared with 15% for the broader market, while the sector’s PEG ratio stands at 1.1 versus 1.4 for the wider equity market. In other words: valuations are elevated, but remain supported by superior growth.
Concentration remains high, and the market is still heavily dependent on a small group of leaders such as Microsoft, Alphabet, Amazon and Meta. But the evidence suggests today’s AI cycle still looks materially more grounded than a classic speculative blow-off
The comparison with the late 1990s was central to Corinne de Boursetty’s argument. Unlike during the dotcom period, earnings per share have, so far, broadly tracked the rise in prices. As she put it, “There is a bit of a dislocation between the recent pullback and fundamentals that remain very strong,” and current conditions remain “well below the dot-com bubble peak”. The implication was not that risk is absent. Concentration remains high, and the market is still heavily dependent on a small group of leaders such as Microsoft, Alphabet, Amazon and Meta. But the evidence suggests today’s AI cycle still looks materially more grounded than a classic speculative blow-off.
The second pillar of Corinne de Boursetty’s argument was adoption. AI, she explained, is moving through successive phases – from generative AI to more agentic and physical applications – and its core functions can be defined as automation, innovation, and productivity. What matters for investors is that adoption is no longer confined to infrastructure providers or a handful of platform companies. Now, sectors from finance and telecoms to healthcare and retail are finding ways to create value using AI.
The underlying data support that point. Corinne De Boursetty noted that, in February 2026, ChatGPT reached around 800 million weekly active users, and that “white-collar” sectors – including information, professional and scientific services, education, and finance – are “where we see the fastest adoption rate”. This suggests the AI cycle is not just a story of chips and cloud infrastructure, but increasingly one of enterprise integration and productivity gains across the real economy.
The next test is spending, with a major market concern over the past year being whether hyperscaler capex can sustain its current pace. Here, Corinne De Boursetty was clear: AI hyperscaler capex reached USD 455 billion in 2025, and the consensus is that it will rise to USD 712 billion in 2026 and USD 834 billion in 2027. That still points to a very large infrastructure build-out, even if growth is likely to moderate over time.
Corinne De Boursetty also emphasised that, with capex as a share of operating cash flow approaching 97%, funding quality is becoming one of the most important factors to monitor. While that doesn’t mean the model is broken, investors do need to watch balance-sheet discipline more closely.
Evidence from the AI supply chain also remains positive, with a strong upcycle in memory – particularly High Bandwidth Memory (HBM) – showing AI demand continues to encounter real capacity constraints rather than surplus. This suggests the spending cycle is still responding to genuine shortages rather than mere exuberance.
“China tech is attractively priced” thanks to four key strengths: valuation advantage, superior earnings growth, competitive edge & innovation, and AI adoption & monetisation
“China tech is attractively priced”
Turning to regional positioning, Jian Bo Gan argued that “China tech is attractively priced” thanks to four key strengths: valuation advantage, superior earnings growth, competitive edge & innovation, and AI adoption & monetisation.
The earnings comparison between China and the US was particularly striking. For 2026, Jian Bo Gan showed that, for MSCI China versus MSCI USA, expected EPS growth was 39% versus 27% in IT, 34% versus 31% in communication services, and 29% versus 9% in consumer discretionary. Yet, Chinese communication services and consumer discretionary still trade at materially lower valuations than their US peers.
China is also leading in adoption and execution. As Jian Bo Gan pointed out, by 2023, China was ahead of Germany, Japan, and the US in terms of the number of factory robots per 10,000 workers, and installed-capacity growth in Chinese data centres is expected to be strong. “China’s greatest strength is in its adoption of advanced tech in everyday life,” argued Jian Bo Gan, before adding that “faster adoption and commercialisation would mean faster monetisation opportunities”. In an AI cycle increasingly constrained by both compute and electricity supply, China’s power availability and deployment speed are becoming strategically important.
Market anxiety has become most acute around software, as investors worry that existing products could be displaced by AI agents. Jian Bo Gan, however, took a more nuanced view. “Cybersecurity, critical and data infrastructure software,” he said, are less at risk, while “basic applications and content” are more vulnerable. Jian Bo Gan’s suggestion, therefore, was not that investors should avoid the software sector entirely, but that they engage with it selectively.
With software having already been heavily derated, that distinction matters. As Corinne De Boursetty argued, “Software is really the avenue of AI rollout and adoption.” For investors, conclusion was equally explicit: following the recent sell-off, there are attractive opportunities in MSCI World Software. In other words, the market may have become too indiscriminate in pricing disruption risk, creating a more interesting entry point in higher-quality software names tied to enterprise adoption.
When asked whether earnings can really hold up if capex growth slows, Corinne de Boursetty accepted that AI capex cannot sustain annual growth of 60% indefinitely.
On China, Jian Bo Gan acknowledged that regulation remains an important factor for investors to consider. But he also suggested the current mood is materially more supportive than during the country’s earlier regulatory crackdown, with China now positioning technology as a clear strategic priority.
The broader investment conclusion remained constructive, but selective. In developed markets, the focus was on “leading AI beneficiaries across the full value chain,” with a preference for enablers such as advanced semiconductors and lithography equipment, as well as adopters including IT platforms, enterprise software, and cybersecurity. Meanwhile, in Asia, the emphasis was on AI enablers and adopters in China, Korea, and Taiwan.
Investors should remain disciplined on valuation, alert to concentration and funding risks, and more selective across software and regional exposures
A noisy and polarised debate, rethought
John Woods, Corinne de Boursetty, and Jian Bo Gan showed that the question of whether to invest in AI is not a binary one. Instead, a better approach to deciding whether and how to invest entails breaking the cycle into its investable components: valuations, earnings, adoption, capex discipline, supply bottlenecks, sector differentiation, and regional opportunity.
Viewing the AI cycle in this way reveals a solid foundation of strong earnings growth, rapid adoption, and real infrastructure demand. At the same time, though, investors should remain disciplined on valuation, alert to concentration and funding risks, and more selective across software and regional exposures.
At Lombard Odier, our role is to bring perspective as well as conviction. Amidst a noisy and, often, polarised debate around investing in the AI cycle, it is essential that we continue to step back and rethink the narrative, so we can help our clients identify the most compelling risk-adjusted opportunities.
important information
This is a marketing communication issued by Bank Lombard Odier & Co Ltd (hereinafter “Lombard Odier”).
It is not intended for distribution, publication, or use in any jurisdiction where such distribution, publication, or use would be unlawful, nor is it aimed at any person or entity to whom it would be unlawful to address such a marketing communication.
share.