Capital & Industrial Strategy

The Musk v. OpenAI trial is now in its final phase, with closing arguments centring on whether Sam Altman misled early co-founders about OpenAI's transition from nonprofit to capped-profit structure. As the Financial Times reports, Musk's legal team is arguing that Altman's representations to early donors and collaborators — including Musk himself — were fundamentally dishonest, a claim that, if accepted by the jury, could expose OpenAI to damages or injunctive relief that would complicate or delay its commercial restructuring and IPO plans. TechCrunch notes that trust in Altman as a steward of what was originally a public-benefit mission has become the dominant narrative in the trial's final days.

The strategic stakes extend well beyond the litigation itself. OpenAI's ability to close its for-profit conversion and proceed toward a public offering depends on demonstrating clear legal title to its governance structure. A jury finding against Altman on trustworthiness grounds — even without a damages award — would hand ammunition to regulators, state attorneys general, and institutional investors conducting pre-IPO due diligence. CNBC contextualises the relationship breakdown as an 11-year arc from co-founders to adversaries, underscoring that this is not a manufactured dispute but a genuine conflict over who controls the trajectory of the most valuable AI company in the world.

The Wall Street Journal provides the most detailed account to date of how AWS has repositioned itself in the AI infrastructure race. The strategy rests on three pillars: a $200 billion capital expenditure programme concentrated in data centre build-out, a custom silicon roadmap through Trainium and Inferentia chips designed to reduce dependence on Nvidia and offer enterprise customers price-performance advantages, and a series of strategic investments and partnerships — most notably the $4 billion stake in Anthropic — that give AWS an anchor AI model tenant and a differentiated model-as-a-service offering. The cumulative effect is that AWS is now competing on infrastructure depth, not just API access.

The Anthropic relationship is particularly significant from a capital strategy perspective. By holding a large equity stake while also being Anthropic's primary cloud provider, Amazon has effectively vertically integrated across compute, model, and distribution layers — a structure that mirrors Microsoft's OpenAI relationship but with a more explicit infrastructure dependency built in. This week's news that Anthropic will brief the Financial Stability Board on the cyber capabilities of its Mythos model adds a regulatory dimension: as Anthropic engages systemic financial regulators, AWS's position as its infrastructure backbone becomes more strategically entrenched.

Publicis Groupe has confirmed a $2.2 billion deal to acquire LiveRamp, a US-based data connectivity and identity resolution company, according to the Financial Times. The deal is a confirmed announced intention; regulatory clearance has not yet been obtained. The strategic logic is straightforward: as AI-powered creative and media execution becomes commoditised, the durable competitive moat in advertising shifts to proprietary first-party data infrastructure. LiveRamp's core product — enabling brands to connect and activate customer data across platforms without sharing raw identifiers — is precisely the capability that becomes more valuable as cookie deprecation advances and privacy regulation tightens.

This acquisition continues a consolidation pattern in marketing technology where the large holding companies are acquiring data infrastructure rather than creative agencies. Publicis already operates the Epsilon data platform acquired for $4.4 billion in 2019; LiveRamp extends that stack into data clean room and identity graph territory. The combined entity would control one of the largest privacy-compliant customer data networks outside the walled gardens of Google and Meta, a position that becomes a direct bargaining chip in AI-driven programmatic negotiations.

Industrial electricity prices across core European economies remain 2-3x higher than US equivalents on a purchasing-power-adjusted basis, and CNBC reports this is now materialising as a hard constraint on data centre siting decisions. The continent is bifurcating: Nordic countries benefiting from hydropower, and Iberian markets with surplus solar capacity, are actively winning hyperscaler investment, while Germany, France, and the Benelux — where industrial demand competes with data centre load — are losing ground. This is not a temporary price spike but a structural consequence of deindustrialisation policy, carbon pricing, and insufficient grid investment.

The China comparison is pointed. Bloomberg highlights former Treasury Secretary Paulson's assessment that China's coordinated investments in renewable generation, battery storage, and transmission infrastructure are creating a durable energy cost advantage that directly supports AI compute build-out. Former Ambassador Burns adds that China's grid investments are already reshaping global supply chains for the components that power data centres. The implication for European industrial strategy is that energy policy and AI competitiveness are inseparable — and current EU frameworks are not treating them as such.