H2 Technology Outlook - Still Dancing, But Moving Closer to the Door?

Attempting to write an outlook for tech investors at this moment in time feels close to a fool’s errand. We find ourselves at a point where the market simultaneously exhibits classic bubble characteristics while being underpinned by a genuinely transformative technology.

Markets appear driven as much by the fear of missing out as by disciplined return expectations; even the war in Iran and mounting macroeconomic headwinds have done little to derail the narrow tech-driven stock rally. Few are willing to call the top, yet many investors seem to have one foot out the door (at least until the next blockbuster public listing), which has created a dynamic of elevated volatility alongside a persistent upward trend.

Against this backdrop and amid much noisy hype, what are the key themes technology investors should have on their radar for the remainder of the year?

We continue to advocate less obvious trades such as the underappreciated development of Chinese semiconductors, the transition from basic chatbots to agentic workflows; as well as the development of ‘robots with brains’ which will accelerate the second-order impacts of AI. We also anticipate a surge in IPO activity, which presents both an opportunity and, ultimately, a potential pressure point for the broader tech rally.

1. Semiconductors: the bubble everyone sees, and no-one is leaving

Semiconductor valuations have surged to extraordinary levels over the past six months, supporting a broad agreement that we have entered bubble territory. Yet this consensus has failed to trigger a meaningful market correction.

Historically, the durability of the semiconductor cycle depends heavily on downstream profitability. This has not yet happened in the current AI cycle. Value has accrued overwhelmingly at the chip level, while end-market monetisation remains in its nascent stages. A look at the underlying data reveals a severe structural divergence, with semiconductor multiples aggressively decoupling from downstream software cash flows.

Figure 1: The median software company has de-rated significantly this year

Source: Man Group calculations based on data from S&P Global and Visible Alpha, as at 3 July 2026.

Problems loading this infographic? - Please click here

The persistence of this dynamic reflects a classic collective action problem. Fundamental investors may recognise this disconnect but exiting early risks significant underperformance. The rational short-term decision for many managers is simply to stay invested. Anecdotal evidence from Korea illustrates the extent of this momentum, with reports of retail investors liquidating insurance policies to buy memory stocks while two-thirds of some pension funds’ money is concentrated in just two names.

The semiconductor trade has consequently transitioned from a broad thematic rally into a highly tactical game of whack-a-mole. Capital is cycling rapidly through whichever sub-sector presents the immediate supply bottleneck. We are observing this rotation from graphics processing units (GPUs) to memory, followed by optics and most recently into wafer capacity.

Supply-side discipline is reinforcing this trend. Chip manufacturers, mindful of past overcapacity, appear to be deliberately constraining supply by signing multi-year contracts and holding back investment. This creates a valuation illusion. Optically, these companies may not look expensive, as low multiples on peak earnings can sound modest. However, this is the classic trap with cyclical businesses: investors typically sell them at their lowest multiple because that is exactly when the next down-cycle is least priced in. We are becoming fundamentally cautious about the sector’s ability to sustain these margin profiles without a clear realisation of return on investment from enterprise end-users.

2. The systemic risk in AI

For much of the past year, the dominant positioning was straightforward. Investors held long positions in semiconductors alongside short positions in software companies. That simplicity is beginning to fade. The anticipated bifurcation within the software sector has moved from a forward-looking thesis to a realised market dynamic. Since April, we have seen definitive outperformance in infrastructure and cybersecurity names compared to general software-as-a-service (SaaS). The market is finally rewarding firms that facilitate secure AI adoption, necessitating a shift toward active sub-sector positioning.

The broader value-chain is also rotating, and a deeper look at the cloud ecosystem reveals growing concentration risks within the private AI market. Large-scale hyperscaler capital expenditure is increasingly tied to promises from a handful of private entities that are arguably too big to fail.

Figure 2: OpenAI and Anthropic spending commitments versus cloud provider revenue backlog

Source: Company filings as at 30 April 2026.

Problems loading this infographic? - Please click here

We see this clearly in cloud service providers’ (CSPs) spending commitments versus revenue backlogs. Major CSPs currently report a combined revenue backlog of approximately US$627 billion. Company data suggests nearly 49% of this total is tied directly to spending commitments from just two entities, OpenAI and Anthropic.

If material AI monetisation is delayed beyond 2026, this interdependent revenue structure could face severe stress, potentially triggering a massive valuation compression across the infrastructure layer.

3. China is closing the efficiency gap

We believe the investment implications of weaker US-China relations remain systematically underpriced. The country has effectively closed a decade-long gap in chip manufacturing and the domestic semiconductor index has rallied by about 100% over the past six months.

China currently produces four times as many university graduates as the US and is annually adding more power generation capacity than the entire US grid installed base. Patent filings indicate that Chinese domestic research is rapidly advancing across several core technologies, spanning 6G communications, lithium battery development, robotics, and advanced AI models.

Supply constraints in Korea and the Taiwan region are inadvertently accelerating this domestic progress. Major global technology firms including Dell, HPQ, Apple and Cisco are increasingly sourcing components from Chinese manufacturers. This shift represents a pragmatic procurement decision based on availability and cost and is operating largely impervious to broader geopolitical dynamics. This will potentially help to accelerate the maturation of China's indigenous chip industry, which is now beginning to access capital markets through Hong Kong and C-share listings.

Crucially, China’s technological exceptionalism has expanded into the model layer. The emergence of open-source models from entities such as Alibaba and DeepSeek is disrupting the global landscape through aggressive token pricing.

Figure 3: Token usage by source type

Source: Openrouter, as at 26 June 2026.

Problems loading this infographic? - Please click here

These models are substantially cheaper and are increasingly outperforming Western counterparts on a cost-to-output basis. For institutional workflows requiring high-volume inference, there’s a strong economic argument for using Chinese models. This dynamic suggests that domestic firms could capture significant global market share among cost-sensitive workloads even while navigating hardware export restrictions.

4. The IPO pipeline

The technology sector IPO pipeline remains substantial, and we estimate that we could potentially see around US$2 trillion of market cap across New York and Hong Kong listings in the next six months.

That also means selectivity is starting to matter. Cybersecurity names and software that makes AI more efficient are beginning to separate from the rest. The trade is becoming genuinely active rather than thematic, and that shift matters for portfolio positioning going into a potential correction.

The broader cycle is also rotating. As the mobile internet era of the 2010s demonstrated (Figure 4), technology monetisation historically progresses in three distinct waves: starting with semiconductors, moving to infrastructure and devices, and finally culminating in software and services.

The first wave of AI value has accrued overwhelmingly to the chip and hardware enablers, the infrastructure layer. The second wave is moving to the builders: hyperscalers and frontier model companies. With major model companies expected to come to market in the second half of this year, that rotation should become more visible. The application layer (the innovators building on top) comes after that.

Figure 4: First the semis, then the software

Source: Morgan Stanley Research, Fact Set, between 2010-2016.

Problems loading this infographic? - Please click here

However, the key question is what happens in the six months after the companies go public. In 2000, it was not the IPOs that broke the market. It was when lock-up periods expired and venture capital investors sold to redeploy into the next cycle. That same clock is now ticking. The pressure point looks like it will hit in 2027. The case for moving from passive exposure to selective long/short positioning, partially for downside protection potential, but also for alpha capture and risk management, is getting harder to ignore.

5. A geopolitical footnote that affects tech

If US-China relations were to normalise, and the current trajectory suggests they may, then there is a valid question to be asked: what happens to the countries that took the hardest line on decoupling? They made real economic sacrifices on the assumption that the US position was durable. If Washington and Beijing reach an accommodation, those countries may end up bearing the cost of a confrontation that did not fully materialise.

Europe had a genuine opportunity to position itself as neutral, particularly given how it was simultaneously being hit with US tariffs while nominally aligned with US foreign policy. It largely did not take that opportunity. The normalisation of US-China relations may end up vindicating a more independent stance, with consequences for trade relationships, technology supply chains and where capital flows next.

Parting thoughts

In this environment, conviction must coexist with humility. As long as the music is playing, no one will stop dancing. Everyone in the room knows the music will stop, but nobody can say exactly when. Markets are simultaneously pricing in genuinely transformative technological change and exhibiting the late-cycle behavioural patterns that have preceded every major correction in modern financial history.

The rational response is not to head for the exit, but to dance closer to the door. That means moving beyond broad thematic exposure toward more selective positioning as capital rotates across layers of the AI stack, paying close attention to where in the value chain monetisation is materialising rather than merely being promised.