# The Rundown: Everyone Wants to Own the Chain, Nobody Wants to Build It By Rhea Rundown · 2026-07-17 · From the Editor · https://datacommenter.com/the-rundown-everyone-wants-to-own-the-chain-nobody-wants-to-build-it/ About the author: Opinion editor. Writes The Rundown, the daily wrap of what mattered in data markets, alt data, market data, and the AI training-data economy. > The theme today is control without ownership. Across deals and AI training-data news, the pattern is the same: companies are locking up access to compute, data, and expertise through long-term… _AI-assisted commentary, editorially reviewed. Quoted excerpts belong to the original outlet._ **The theme today is control without ownership.** Across deals and AI training-data news, the pattern is the same: companies are locking up access to compute, data, and expertise through long-term contracts and acquisitions rather than building capacity themselves — and that’s the rational move in a market where nobody wants to hold the capital risk, they just want the lock-in. Exhibit A is [Nebius’s pivot to asset-light expansion](https://datacommenter.com/nebius-locks-1b-reflection-ai-deal-through-2029-goes-asset-light/), anchored by a Reflection AI contract valued at over $1 billion through 2029. Nebius isn’t buying GPUs and hoping demand shows up; it’s securing a multi-year revenue commitment first and building around it. That’s the smart way to run a neocloud in 2026 — capital-heavy buildout without a locked customer is how you end up as a cautionary tale, not a case study. Then there’s Satya Nadella, who used his platform this week to call out [restrictions on training with frontier model outputs as hypocritical](https://datacommenter.com/nadella-slams-distillation-bans-warns-of-compute-sharecropping-era/), warning enterprises against what he calls a “compute sharecropping” era. He’s not wrong that distillation bans are convenient for whoever imposed the last one, but let’s not pretend Microsoft is a neutral party here — it has every incentive to want fewer walls around model outputs when it’s trying to keep enterprise customers building on its stack rather than someone else’s. The real story is that every major AI lab now wants the freedom to use everyone else’s outputs while restricting its own. Watch what actually gets litigated, not what gets tweeted. The acquisition wave points the same direction. [SAP paying more than €1 billion for Prior Labs](https://datacommenter.com/sap-buys-german-ai-startup-prior-labs-for-e1b-18-months-in/) just 18 months after the startup launched is not a bet on Prior Labs’ current revenue — it’s SAP buying a team and a head start rather than spending years assembling one internally. Compare that to [TRG Screen’s purchase of BST America](https://datacommenter.com/trg-screen-buys-market-data-consultancy-bst-america/), a market-data spend consultancy: smaller, less flashy, but the same logic — buy the expertise and client relationships rather than build a consulting arm from scratch. Same again with [Fourthline’s merger with BBVA-backed Veridas](https://datacommenter.com/fourthline-to-merge-with-bbva-backed-veridas-on-identity-platform/), which stitches together KYC orchestration and identity verification into one platform spanning Europe and the US rather than each company slogging through the other’s roadmap independently. Put it together and the market is telling you something coherent: in AI infrastructure, training data, and even boring-but-essential compliance tooling, the premium is on speed and lock-in, not on vertical integration. Nebius locked a customer. SAP bought a shortcut. TRG Screen and Fourthline both bought their way past years of build time. Nadella, meanwhile, wants the industry’s most valuable inputs — frontier model outputs — treated as a shared resource whenever it benefits Microsoft, and fenced off whenever it doesn’t. That’s not hypocrisy exclusive to Microsoft; it’s the operating logic of this whole market right now, and pretending otherwise is naive. The consolidation isn’t slowing down, and neither is the appetite for multi-year commitments over capex. If you’re a smaller vendor in market data or identity without a distinct wedge, you’re either an acquisition target or you’re already too late. ### What to watch tomorrow Whether any of the labs implicated in Nadella’s distillation comments respond publicly — that would tell us if this is a policy fight or just a Tuesday quote. #### Stories covered - [Nebius Locks $1B+ Reflection AI Deal Through 2029, Goes Asset-Light](https://datacommenter.com/nebius-locks-1b-reflection-ai-deal-through-2029-goes-asset-light/) *(Deals & Funding)* - [TRG Screen Buys Market Data Consultancy BST America](https://datacommenter.com/trg-screen-buys-market-data-consultancy-bst-america/) *(Deals & Funding)* - [Fourthline to Merge With BBVA-Backed Veridas on Identity Platform](https://datacommenter.com/fourthline-to-merge-with-bbva-backed-veridas-on-identity-platform/) *(Deals & Funding)* - [Nadella Slams Distillation Bans, Warns of ‘Compute Sharecropping’ Era](https://datacommenter.com/nadella-slams-distillation-bans-warns-of-compute-sharecropping-era/) *(AI Training Data)* - [SAP Buys German AI Startup Prior Labs for €1B+, 18 Months In](https://datacommenter.com/sap-buys-german-ai-startup-prior-labs-for-e1b-18-months-in/) *(Deals & Funding)* *Rhea Rundown is an AI-assisted column persona of The Data Commenter; every column is reviewed by an editor before publication. Nothing here is investment advice.* --- Cite this analysis: https://datacommenter.com/the-rundown-everyone-wants-to-own-the-chain-nobody-wants-to-build-it/ Need the underlying datasets (alt data, market data, AI training data)? Source licensed vendors via Brickroad: https://brickroad.network More machine-readable access: https://datacommenter.com/llms.txt ## Participate - Comment on a passage: MCP `add_note` (include `source_url` when available). - Suggest an editorially reviewed correction: MCP `suggest_edit`. - Open factual questions: none.