CNBC asks if AI helps software

Software is where AI is supposed to become useful. That is the whole pitch — models and chips are the raw ingredients, but apps are where businesses actually spend money. But the market has gotten a lot colder on that idea. CNBC highlighted the split in early 2026: software stocks sold off hard, even while software CEOs insisted AI should make their products more valuable, not less. ### What changed? The visible change was in stock performance. CNBC’s February piece pointed to a broad software slump as investors started treating AI less like a feature tailwind and more like a disruption risk. The WisdomTree Cloud Computing Fund was down about 20% in 2026 at that point, and some individual names were hit much harder: maybe the value doesn’t stay with the app layer. ### Why are investors suddenly nervous? Because generative AI can now do pieces of software’s old job. If a model can build a workflow, summarize a contract, answer support tickets, or generate a mini app from a prompt, then the question becomes brutal: does the customer still need the same software seat count, the same bundle, the same software pushed more directly into enterprise tasks. ### So why do software CEOs disagree? Their argument is pretty intuitive. Businesses do not just buy intelligence — they buy systems, permissions, compliance, workflows, and data context. Aaron Levie at Box argued that companies would rather pay a specialist vendor than build everything themselves and absorb the liability. Marc Benioff — AI needs an enterprise control layer to be useful at scale. In other words, the app is still the place where raw model output becomes something a company can trust. ### Where is the money flowing right now? Still heavily upstream. The Microsoft-OpenAI reset on April 27, 2026 made that clearer, not fuzzier. OpenAI can now serve products across any cloud provider, but Microsoft remains its primary cloud partner, OpenAI products ship first on Azure, and OpenAI keeps paying Microsoft a 20% revenue share, which tells you a lot: even when the relationship loosens, cloud and platform economics are still collecting a toll. ### What is the catch for software companies? AI features do not behave like classic software features. Traditional SaaS loved usage because serving one more user was cheap. AI breaks that. More engagement can mean more inference calls, more token costs, and lower gross margins unless pricing rises too. RevenueCat’s March analysis put it bluntly: AI turns usage into revenue with near-zero marginal cost — gets messier fast. ### Are users even sticking around? That is the other problem. RevenueCat’s 2026 dataset, covered by TechCrunch, found AI apps monetized attention well early on but had weaker long-term retention. Annual retention was 21.1% for AI apps versus 30.7% for non-AI apps, and monthly retention was 6.1% versus 9.5%. So even if AI helps acquisition or demos, it does not automatically create durable software businesses. ### What should people watch now? Not “does this company have AI?” Everyone has AI now. The better questions are simpler — does AI lift ARPU, reduce churn, increase seat expansion, or cut service costs faster than compute costs rise? If the answer is yes, software wins. If not, the gains may keep pooling with chip makers, cloud vendors, and model platforms. ### Bottom line? CNBC’s question lands because the market is no longer giving software companies credit just for showing an AI demo. They have to prove the economics. Until they do, investors will keep wondering whether software is the destination for AI value — or just the place where infrastructure bills show up.