Signals

Spain has blacklisted Palantir from public and private state-controlled companies, citing concerns about the potential misuse of classified national-security information. It is a European state moving from rhetoric to procurement exclusion of a US data-analytics vendor — a hard enforcement data point in a debate that has mostly been conducted in principle. It compounds the sovereign kill-switch logic that has been building for weeks: dependency on US-controlled AI and data infrastructure is now being actively unwound at the national level. In parallel, Germany's Infineon opened its Dresden "Smart Power Fab" three months early as a symbol of EU component autonomy, and Brookfield told CNBC it sees the UK and Europe as "the middle ground between the United States and China" for data-centre expansion. For investors, the neutral-jurisdiction and sovereign-compute wedge is firming into a coherent lane rather than a slogan. The corollary risk is concentration exposure: US vendors whose value assumes unimpeded access to allied public-sector markets now face a demonstrated precedent for exclusion.

SAP said it will restrict hiring and travel to redirect spending into a significant AI push. It is a modest-sounding operational note that carries a large concession: a core-ERP incumbent is constraining headcount growth to fund the transition, rather than treating AI as an add-on to seat-based licences. Mark's framing is the signal within the signal — "SAP has now had their penny-drop moment, three years after Salesforce." The lag is the point. The recognition that per-seat enterprise-software value is under pressure reaches the slower incumbents years after the faster movers have priced it in, which is exactly how long adoption curves tend to run before they force a response. For investors tracking the software transition, incumbent cost-base reallocation is a more reliable tell than product announcements. When a company reorganises its spending away from people and toward tokens, it is revealing what it now believes about where the value went.

Microsoft is committing US$2.5bn and moving around 6,000 engineers and salespeople into a new subsidiary, Microsoft Frontier Co., to work directly with clients on AI implementation. John's read: "their own KPMG-style org, like OpenAI and Anthropic have." It follows AWS standing up a US$1bn Forward Deployed Engineering unit the week before, and mirrors the outcome-oriented services motions the frontier labs have already built. The convergence is the story. The platform vendors and the model vendors are arriving at the same conclusion — that margin now sits in the human-plus-agent work of making the technology deliver a client outcome, not in selling model access as a self-serve product. That is value pooling at the services end as raw capability commoditises. For investors it reframes the incumbents' defensibility: Microsoft is reaching for a services moat rather than a model moat. It also complicates the tidiest labour-displacement narratives — the labs are hiring humans to close the last mile even as they automate the middle.