Investors had a tough start to 2026, with software stocks falling more that 20% YTD on fears that agentic AI could disrupt software companies’ business and revenue models. We have seen this before, when investors feared cloud/software-as-a-service (SaaS) trends would disrupt pure licensing, maintenance, and service revenue models.
As chatbots and agentic tools (systems capable of taking autonomous, goal‑directed action) get closer to the user and automate tasks once handled by enterprise applications, software suddenly looks more vulnerable than hardware. Chips remain the picks and shovels of this cycle, but large parts of the software layer are being repriced as ‘optional’ rather than ‘necessary’.
That does not mean that all software is dead. If history is any guide, the next phase will be about separation. Some vendors are likely to be disintermediated – features turned into commodities, pricing power erased, renewal cycles questioned. Others are so deeply embedded in client workflows, compliance frameworks, and data plumbing that they can pivot towards bundling AI, rewriting products, and defending distribution. The market is trying to guess who is who – and it is doing so with a blunt instrument.
What will survive the AI disruption?
The sell-off in technology stocks is an unforgiving inflection spawning from AI’s ruthless frontier model improvements. AI has marked a hiatus across many value chains; those expecting valuation mean reversion may instead witness a decaying moat. Moats, not valuations, form the base of our AI survivability framework.
The sell‑off reflects how uncertain investors have become about weakening competitive advantages and long‑term company value. Some high-growth software stories are compromised as agentic platforms automate workflows previously requiring expensive per-seat licenses. Advertising and consultancy firms are, apparently, not invited to the party either. This shift has created a dire scenario for software-as-a-service, where legacy incumbents see their competitive advantages liquefied.
We shy away from workflow-only incumbents whose business models rely on fragmented interfaces, as AI agents now navigate desktop layers directly, rendering intermediary seats obsolete. Advertising and consultancy platforms see a similar erosion from a more attractive client approach: build and not buy.
The reality is also that some hyperscalers will continue offering enhanced software verticals, taking further away from smaller players. Hyperscalers are actually monetising AI, which is not a broad-based case; in turn, they have upped their capex for the year ahead, to around USD 685 billion. We see some upside risk on speculative AI application startups and frontier model firms that offer high optionality but suffer from high inference costs and thin distribution.
We prefer the physical gatekeepers – those who have the cash and the hardware. Rallies will hinge around hyperscalers’ capital deployment. Memory is the key constraint, with supply lacking across high-bandwidth memory, flash memory, and storage. These firms secure the terminal value that the application layer is forfeiting.
While there is indiscriminate selling in the equity market today, we believe ultimately there will be a distinction between fully disrupted business models and those that face margin pressure as coding is commoditised. As we stated following DeepSeek’s five seconds of fame, doomsday talk is exaggerated, as switching costs and retraining remain high, especially for enterprise software.
What’s the bottom line for investors?
The sell-off in technology stocks reflects investors’ recognition of the fact that artificial intelligence erodes software's competitive advantages faster than expected. Workflow incumbents and legacy software models face structural risk as AI agents bypass interfaces. We remain constructive on Cloud Computing & AI, with a clear preference for hardware and hyperscalers, where proprietary compute and data remain the clearly defensible scarcity.
Outside the ‘new economy’, the macroeconomic backdrop looks oddly supportive. Risk assets have rebounded globally, helped by political relief rallies in Japan and Thailand and by a stabilising oil price as US-Iran talks defuse tail risks. More importantly, inflation appears to be behaving on both sides of the Atlantic, raising the prospect of a Goldilocks stretch characterised by moderate growth, cooling prices, and policy flexibility. Even the US consumer, grumpy for much of last year, is showing signs of life as Michigan sentiment hits a six-month high.
Into this mix, we would stay selective: cautious on software until the dust settles, constructive on semiconductors and on ‘reinventors’, and open to cyclicals if inflation prints cooperate.