The past year has been characterised by extraordinary advances in artificial intelligence (AI), but over recent weeks, we’ve reached an inflection point. Sentiment has turned dramatically for software names, and we cannot ignore the dispersion between AI’s impacts on software and hardware companies.
A sector in capitulation: the software rout
This year, we have seen excessive selling at levels far beyond anything we saw during Covid, with the degree of bearishness in the software sector now comparable to the dot-com unwind and the Global Financial Crisis.
What’s unusual is not the magnitude of selling, but its speed and uniformity. Although it is not the only sector to have been hit, every corner of the software ecosystem is feeling the effects of AI, including platforms, cybersecurity, enterprise applications and internet services. Year to date, software indices have undergone enormous markdowns, while semiconductors and cloud infrastructure have powered ahead in the opposite direction.
This is no longer rational de-risking. It’s mass capitulation driven by the fact that investors are finding it challenging to model what AI will mean for software business models.
Why investors cannot model AI’s impact on software
Software companies are growth companies, so their worth is heavily tied to their terminal value – the cash flows expected from the company far into the future. A software company’s valuation depends critically on the assumption that it will continue to compound growth at attractive rates for many years.
When investors look at AI, particularly enterprise-ready agents from companies like Anthropic, which are already eroding parts of the software value chain, many are making the leap to “What if long-term growth is no longer 10-15%, but closer to zero? Or negative?” If you move that growth assumption just a few percentage points, a company’s terminal value contracts sharply. And with such sensitivity to long-term growth assumptions, you can justify almost any price to the downside.
Enterprise-ready agents are clearly reducing the barriers to entry for software development, including reducing the cost, which makes any assumptions of future growth very difficult to model and raises an unquantifiable threat in terms of future competition. Once fear sets in and a story gathers momentum, it becomes self-reinforcing, which is what we are currently seeing. The challenge for investors in terms of modelling AI’s impact on software names is understanding just how easy it will become to develop software in the future and what this means for software companies’ growth.
The subscription model under pressure
Another part of the problem is the subscription-based nature of most software businesses. Customers pay for a licence, per user, and traditionally the business enjoys high stickiness and very high gross margins. But if an AI model can produce a comparable design in seconds at a fraction of the cost, or if a firm no longer needs the same number of designers, the threat to the seat becomes real.
We’re moving into a space where you will pay for the outcome – what the software produces – rather than the license – the permission to use the software. This marks a shift in how value is captured. In this model, vendors will have to prove that their tools deliver a high standard of output alongside tangible productivity gains. For many software businesses, this raises uncomfortable questions about pricing power, user retention and the durability of recurring revenue, particularly if AI drives both a reduction in seats and a revaluation of what customers are actually willing to pay for.
Over the last two weeks, we’ve seen accelerating layoffs across companies exposed to knowledge-worker productivity or automation pressures. These job cuts feed a narrative that software seats will be cut, licences will be cancelled and revenue will weaken. This chain reaction is what is driving the current sell-off.
The other side of the story: hardware’s exceptional strength
While software is suffering, the hardware ecosystem – data centres, semiconductors, cloud infrastructure – is experiencing the opposite. Hyperscaler investment is exploding. Across the big four (Amazon, Alphabet, Microsoft, Meta), capital expenditure is projected to reach nearly $700 billion in 20261.
Hardware names have seen remarkably strong year-to-date performance driven by insatiable demand for compute; multi-year AI infrastructure buildouts; widening performance gap between models, requiring ever more specialised chips; and robust cloud revenue inflections, particularly in Alphabet and Amazon’s cloud businesses. While software grapples with questions around company valuations, hardware remains in a period of structural optimism.
Is there a turning point for software?
The honest answer is the catalyst is not yet clear, although several conditions could stabilise the sector. These include:
- Signs that the speed of adoption in AI is slower than expected, allowing enterprise software providers to monetise AI rather than be rapidly displaced by it, keeping switching costs higher.
- Signs SaaS incumbents can minimise customer churn while pivoting their business models, despite this competitive threat, keeping some network effects.
- Mergers and acquisitions, whether that be private equity or trade buyers, suggesting a floor has been found under valuations.
At present, however, the market is not in the mood to discriminate. In this first phase – the “sell everything” phase – investors clearly do not feel they have enough information to begin distinguishing the opportunity set.
This is precisely when disciplined active managers need to begin their search, differentiating between the companies whose long-term fundamentals remain intact or are now mispriced due to emotion rather than genuine structural decline.
The Sustainable Global Equity Fund’s positioning: remaining selective, appraising fundamentals
Despite the prevailing sector mood, we continue to hold selective positions within software where the underlying business quality, switching costs and platform advantage remain robust. Specifically, we would consider Microsoft and cybersecurity company Palo Alto as demonstrating these attributes. These businesses may face near-term narrative pressure but still retain characteristics that can translate into long-term shareholder value.
At the same time, we are continuing to evaluate whether the current environment is beginning to offer early-stage entry points into oversold software names. But for that to be the case, we must see a durable competitive moat that AI enhances rather than disrupts; a path to monetising AI plugins, agents or automation layers; and a business model resilient enough to withstand potential short-term subscription contractions.
The task now is to stay analytical and continue to discern where absolute value appears and understand what, if any catalyst, would trigger a reappraisal of fundamentals.
Capital at risk. The value of an investment and the income from it may go down as well as up and the investor may not get back the amount initially invested.
This document has been prepared by EdenTree Investment Management Limited for Financial Advisors, other intermediaries and other investment professionals only. It is not suitable for private individuals. This document has been produced for information purposes only and as such the views contained herein are not to be taken as advice or recommendation to buy or sell any investment or interest thereto.
This financial promotion is issued by EdenTree Investment Management Limited (EdenTree) Reg. No. 2519319. Registered in England at Benefact House, 2000, Pioneer Avenue, Gloucester Business Park, Brockworth, Gloucester, GL3 4AW, United Kingdom. EdenTree is authorised and regulated by the Financial Conduct Authority and is a member of the Investment Association.
Firm Reference Number 527473.
