Executives from Anthropic, OpenAI, Nvidia and Microsoft gathered at Morgan Stanley’s Technology, Media and Telecom conference this week to make one case: artificial intelligence is no longer an efficiency novelty but the single biggest determinant of corporate fortunes in software and infrastructure.
David Chen, head of global tech investment banking at Morgan Stanley, said the nature of investors’ questions has shifted sharply since 2025. Where boardrooms once asked how AI could cut costs, they now want to know whether a given company will be a net beneficiary or a casualty of AI-driven change — and many enterprise software firms are being treated as if they are already at war.
Chen drew a practical distinction that matters for valuation: software that executes deterministic, mission‑critical tasks — payroll, invoicing, reconciliations — retains defensible value because small error rates are intolerable. By contrast, companies that primarily present and arrange public data behind attractive interfaces face existential pressure as large language models and agents can ingest, re‑organise and serve that same information programmatically.
The market has already penalised these risks: Chen noted that enterprise software lost roughly $1 trillion in market value in a single week before the conference, a dramatic re‑rating that reflects investor anxiety about which business models will survive the arrival of agentic AI.
Security is one clear beneficiary, Chen argued. Cybersecurity products combine high switching costs, regulatory scrutiny and the need for specialised, trustworthy tooling — characteristics that make them more likely to gain from AI than be displaced by it. Other winners may include tightly integrated, deterministic vertical software that embeds correctness and auditability into its core.
On the infrastructure side, Chen highlighted a new generation of semiconductor and systems vendors focused on the three bottlenecks constraining AI deployment: connectivity, compute and energy. These firms, he suggested, are the natural beneficiaries of the next wave of hardware investment, even as overall hyperscale capex may be close to a cyclical peak.
That last point is telling. When asked whether infrastructure spending would be higher or lower by 2027, Chen replied that it would probably be “about the same,” implying the hyperscalers’ AI capital‑expenditure cycle may already be approaching saturation. The implication is that efficiency, specialised accelerators and edge solutions — not endless data‑centre construction — will determine future hardware winners.
Corporate governance is shifting too. Boards, Chen said, increasingly prefer CEOs with product and architecture credentials over sales‑and‑marketing leaders. Replatforming businesses for an agent‑first world requires an understanding of systems design and engineering trade‑offs — skills that are not synonymous with go‑to‑market expertise.
The industry jargon is changing accordingly: where the world once talked about SaaS, some now call the trend SaaaS — software for agents as a service. Box’s CEO Aaron Levie offered a practical example, saying that intelligent agents are his company’s new customer and that the agent‑oriented business could be an order of magnitude larger than its current human‑facing market.
Taken together, these comments sketch a marketplace undergoing selective disruption rather than wholesale destruction. Enterprises and vendors that can reengineer deterministic flows, harden security, and deliver agent‑grade APIs will likely capture value. Those that primarily offer curated public data or simple interfaces without deep technical moats face consolidation or obsolescence.
For investors, incumbents and policymakers the stakes are high. Valuations are being rewritten, hardware supply chains are back in focus, and leadership profiles are being recast around engineering credibility. The immediate future looks like a period of winners and losers being sorted rapidly, with capital, talent and corporate strategy all following the new signal from AI capability and deployment economics.