A dominant market theme is that increased enterprise spending on AI is directly reducing budgets for other areas. This "crowding out" effect is impacting traditional software, IT services, and even non-AI hardware, creating a tough environment for incumbent vendors not central to the AI stack.
Despite public narratives from tech CEOs about data security, enterprise IT executives are less concerned about frontier models stealing IP. Their primary, immediate worry is the practical problem of AI compute and token costs far exceeding budgets, forcing them to throttle usage and re-evaluate their AI strategy.
To optimize AI costs and sustainability, UBS employs a "model garden" with various frontier and smaller models. An internal AI system then routes employee questions to the most appropriate, cost-effective model, preventing the wasteful use of powerful, expensive LLMs for simple, non-frontier problems.
For critical processes in regulated industries, standard AI model evaluations ("evals") are insufficient. Enterprises like UBS are pushing for research into mathematical proofs to formally verify that AI agents behave correctly across multiple tasks, establishing a much higher standard of trust and safety.
When implementing AI for business use, the knowledge needed to evaluate models resides in subjective human experience. A key bottleneck is converting this domain-specific expertise into a machine-readable format that can be used to reliably assess AI performance against real-world business needs.
The traditional moat of enterprise software—high switching costs—is eroding. For the first time, customers can harness powerful AI models to custom-build their own alternatives, particularly for ancillary applications. This threatens the growth models of incumbent SaaS companies that rely on upselling.
For physical AI, the primary constraint is not the cost of data but its fundamental non-existence. Unlike software AI, you can't advance without deploying robots "in the wild" to capture edge cases—a classic chicken-and-egg problem that simulation alone cannot solve and capital cannot easily buy.
The acquisition of Cursor by SpaceX's AI division was driven as much by Cursor's strong enterprise brand and go-to-market team as by its coding data. This highlights the strategic value of commercial infrastructure, like Fortune 500 relationships and sales teams, in AI M&A.
The AI compute crunch isn't only about GPU scarcity. Startups are choosing smaller cloud providers ("neoclouds") over AWS because they offer more flexible terms. They can avoid the large, long-term, and expensive commitments that incumbents often require for high-demand NVIDIA chips.