Capital & Industrial Strategy

SpaceX priced at $135 per share, raising $75 billion in what the Financial Times called the world's biggest IPO, against more than $70 billion in retail orders alone. Shadow markets indicated a 35%+ first-day pop, per Bloomberg. The deal is structurally significant beyond its size: SpaceX is being positioned to investors not merely as a launch and satellite company but as a rockets-to-AI conglomerate, with enterprise AI cited as a $26 trillion addressable opportunity. Gopuff's adoption of xAI's Grok for its shopping assistant — publicly disclosed on the eve of pricing — is one of the few confirmed enterprise deployments and serves transparently as demand validation for that thesis.

The wider consequence, as Alexis Ohanian told Bloomberg at SuperReturn in Berlin, is that the IPO unlocks a pipeline: OpenAI, Anthropic, and a queue of AI-adjacent companies are now being priced against a live public benchmark. The FT characterises the sums Wall Street is being asked to absorb as 'only a down payment.' About three dozen investors — spanning Sequoia to a Tampa family office — hold positions across SpaceX, OpenAI, and Anthropic simultaneously, per WSJ, meaning the IPO wave represents liquidity events for a tightly networked capital cluster that has been accumulating positions for years.

Prometheus, the industrial AI startup Bezos co-founded in November 2025 with an initial $6.2 billion, has now raised a further $12 billion — $18.2 billion in total — at a $41 billion post-money valuation, with JPMorgan and BlackRock leading the financing, per TechCrunch and Semafor. The company's stated aim is an 'artificial general engineer' capable of automating complex physical product design and drug development. Bezos, who has returned to a CEO role for the venture, is publicly dismissive of AI-driven mass unemployment fears — a positioning that is partly strategic, intended to manage regulatory and public-opinion risk around a company whose explicit mission is automating engineering work.

The Prometheus raise, read alongside Neura Robotics' $1.4 billion round backed by Nvidia and Amazon (WSJ), and RLWRLD's Nvidia partnership to develop dexterous robotics benchmarks (Bloomberg), marks a structural shift in where frontier AI capital is concentrating. The software model layer is commoditising rapidly; the physical-world AI stack — sensors, manipulation, industrial simulation, drug design — is emerging as the next high-multiple frontier. Semafor's observation that Bezos and Musk are both building AI-anchored conglomerates (Semafor) points to a model where AI becomes the connective tissue of diversified industrial empires rather than a standalone product category.

KKR has formally launched Helix Digital Infrastructure, a $10 billion vehicle structured to function as a single coordination point for hyperscaler data centre, power, and connectivity procurement, with Nvidia providing chip supply certainty, Vistra providing power capacity, and Kuwait's sovereign wealth fund providing Gulf capital, per WSJ and Semafor. The structure is notable: rather than a fund investing in infrastructure assets, Helix is an operating company — a vertically integrated procurement and development vehicle that collapses the fragmented supply chain between GPU allocation, land, power, and network into a single counterparty for hyperscalers.

Kuwait's participation is strategically significant beyond its dollar commitment. Gulf sovereign wealth funds have so far been primarily passive LP investors in AI funds; Helix represents an operational stake in AI infrastructure delivery, suggesting Gulf states are seeking to control productive assets in the AI build-out rather than merely fund them. This mirrors the pattern of Gulf capital in energy infrastructure over the prior decade. Concurrently, the US insurance regulator's reported probe into credit risks tied to data centre financing (FT) introduces a systemic risk signal that institutional capital allocators should monitor: insurance sector exposure to AI infrastructure debt is becoming large enough to attract prudential scrutiny.

OpenAI is actively weighing significant price cuts for its API and enterprise services in anticipation of matching moves from Anthropic, per Bloomberg. The timing is explicitly tied to both companies' anticipated public offerings: each needs to demonstrate enterprise revenue scale before hitting public markets, but cutting prices to win market share compresses the margins that IPO investors will scrutinise. This is a classic pre-IPO dilemma — growth versus unit economics — playing out in real time. The competitive dynamic is reinforced by WSJ reporting that enterprises are now mixing and matching models to avoid paying premium prices, eroding both companies' pricing power simultaneously.

OpenAI's acquisition of Ona to extend Codex's capability for longer-running agentic coding tasks (CNBC) is a complementary move: as price becomes a weaker differentiator, capability depth — particularly in high-value developer and enterprise workflows — becomes the defensible moat. Anthropic, meanwhile, is reported to be pursuing new data centre leases and seeking additional financial backing from Google (Reuters), indicating that its $965 billion implied valuation has not translated into self-sufficient infrastructure financing — it still depends on its largest strategic investor for compute capacity.

Infineon is opening its €5 billion semiconductor fab in Germany — its largest single investment — with EU subsidy support, as the bloc pursues chip production sovereignty, per Bloomberg. The facility is a direct product of the EU Chips Act framework, which committed public funds to attract and retain semiconductor manufacturing capacity in Europe. Infineon's focus on power semiconductors and automotive chips rather than leading-edge logic means the fab does not directly address Europe's dependency on TSMC for advanced AI processors — but it does reduce vulnerability in the adjacent industrial and automotive semiconductor markets that underpin European manufacturing competitiveness.

The Infineon opening arrives against a backdrop of stark warnings from AI policy analysts that Europe's AI buildout remains orders of magnitude too small to remain competitive, per Semafor. The divergence between the EU's chip sovereignty ambitions — substantial in capital terms — and its AI model and infrastructure investment — still dwarfed by US and increasingly Chinese and Gulf capital — is the defining tension in European AI industrial strategy. Public opinion provides additional political friction: a Reuters/Ipsos poll found Americans broadly wary of the data centre build-out, a sentiment that likely maps to European electorates and complicates the permitting and public financing of large-scale AI infrastructure.