The moment when share prices last came under serious pressure was nothing like a classic “shock” in the old sense: no profit warning, no geopolitical upheaval, no sudden interest rate stimulus. Rather, it was the release of a tool. A new plug-in from the ecosystem of Claude, the language model of the US group Anthropic, gave an impressive demonstration of the abilities that large language models now offer. Where legal work, sales, marketing and data analysis had previously been dominated by high-margin software packages, specialist databases and tried-and-tested workflows, in future an AI agent will likely suffice. This sparked the price volatility that is now affecting a number of industries such as software, logistics and financial services.
The mechanism was initially felt most strongly in the software industry. The market has already coined a term for it: “software Armageddon”. Within just a handful of trading days, hundreds of billions of dollars were wiped off the market value of the global software and services segment as investors suddenly began looking at survivability alongside growth. The fact that the sell-off is occurring in a sector that had for years been considered a bulwark of predictable, recurring sales is what lies at the heart of the current nervousness. That is because artificial intelligence is threatening not only products, but also price logic. The classic “per user” licensing model comes under pressure when AI automates the work of multiple employees, but the customer continues to need just a single licence or simply uses an AI-supported alternative product. Suddenly, black clouds are beginning to hover over the margins that had previously been the pride of the industry.
Despite these fears, plenty of money continues to pour into AI. And that is the third source of tension that is worrying investors: high capital spending, unclear monetisation, rising indebtedness. Alphabet, for example, is looking to pull in billions through the bond market, while Oracle has announced capital measures and additional financing. This indicates that they want to embrace the AI trend rather than deny it. Hyperscalers and major platform providers are building data centres and infrastructure at a rate that defies any traditional IT cycle. This is creating something of a dilemma for investors: without massive AI investment there is a risk of strategic irrelevance, but with it there is a risk of a prolonged period of depressed cash flows and margins. Oracle is the prime example of this dichotomy. The group is aggressively expanding capacities and AI abilities, even though its debt mountain is already high and cloud growth is no longer meeting expectations. Investors see this as a risk, but at the same time it is a signal that even “old” software companies are willing to reinvent themselves in AI.
Backing for the sector comes from an influential quarter: Nvidia. CEO Jensen Huang described the notion that AI would “replace” software as illogical. The product cycles, in his view, will actually result in more software rather than less. Analysts echo these arguments: large language models are powerful, but often they are too generic. According to Mark Murphy, head of software research at JPMorgan, it “feels like an illogical leap” to say a new plug-in would “replace every layer of mission-critical enterprise software.” Chief market strategist Talley Leger of Wealth Consulting Group is singing from the same hymn sheet: “I think the software sell-off is getting overdone and the underlying logic seems flawed.” He can imagine that improving AI tools will make it easier to create new and better software applications at lower prices, therefore improving software company margins.