Capital & Industrial Strategy

OpenAI has confirmed a $122 billion raise at an $852 billion valuation, the largest private funding round on record by a significant margin, per Bloomberg. Khosla Ventures founder Vinod Khosla publicly defended the valuation as not excessive, framing OpenAI's trajectory as commensurate with an industrial-revolution-scale technology shift — a view consistent with a16z's Martin Casado, who simultaneously cautioned that the advantage of raising cheap money at the frontier is time-limited, given that building competitive AI models is not structurally prohibitive for well-resourced entrants, per the Financial Times.

Anthropic's decision to rebuff offers that would have valued it above $800 billion is strategically distinct: it suggests the company is either managing dilution ahead of an IPO, holding out for better terms, or prioritising governance constraints on its investor base. Per Bloomberg, the rebuff is confirmed as a deliberate choice rather than a failed raise. Meanwhile, CNBC flags that negative public opinion on AI is hardening — citing polling data and growing political friction around data centres — which could create headwinds for both companies' IPO timelines and valuations in public markets.

Jane Street has executed two simultaneous transactions with CoreWeave: a $1 billion equity investment confirmed as a closed deal, and a $6 billion cloud services contract, per the Wall Street Journal and Reuters. The structure is notable: Jane Street simultaneously de-risks CoreWeave's revenue base through a long-term offtake commitment while gaining equity upside — a private credit-style construction applied to AI cloud infrastructure. This is not a passive financial bet; it embeds one of the world's largest proprietary trading firms as a captive compute customer, suggesting that quantitative finance's AI workloads have scaled to the point where securing dedicated GPU capacity is a competitive necessity.

The deal also reinforces CoreWeave's post-IPO position as a credible alternative to hyperscaler cloud for latency-sensitive, high-throughput workloads. Financial services firms have historically been reluctant to route sensitive workloads through shared public cloud environments, and a dedicated relationship of this scale with CoreWeave could catalyse similar structures among other institutional trading desks and hedge funds.

TSMC reported a 58% year-on-year profit increase, beating analyst estimates and demonstrating that AI chip demand has absorbed geopolitical shock from the ongoing Middle East conflict without measurable impact on order flows, per Bloomberg and Reuters. Simultaneously, ASML raised its full-year 2026 revenue outlook, citing customers accelerating expansion plans, per the Financial Times. Together, these results validate the thesis that the AI infrastructure build cycle remains in its capital-expansion phase rather than approaching a demand inflection point.

These results also have implications for the US-China chip restriction regime. ASML's raised outlook reflects demand concentrated among leading-edge customers — overwhelmingly TSMC, Samsung, and Intel — which benefits from export controls that limit Chinese fabs' access to EUV tooling. The profitability of the current cycle is therefore partly a function of the restrictions themselves, creating a structural incentive for both companies to lobby for the maintenance of current controls.

Northern Spain has attracted $90 billion in committed data centre investment from Amazon Web Services, Microsoft, and other hyperscalers, and Big Tech is positioning the region as a replicable model for EU AI infrastructure strategy, per Bloomberg. Separately, Google-linked data centres are in the market with a $5.7 billion junk bond issuance — confirmed as a live transaction — which would be the largest high-yield deal ever used to finance AI infrastructure, per Bloomberg. These two developments together reveal the financing architecture underpinning the buildout: sovereign and regional governments provide permitting, energy, and subsidy frameworks; hyperscalers commit anchor tenancy; and the resulting contracted revenue streams are securitised into capital markets instruments that transfer infrastructure risk to yield-seeking institutional investors.

The Spain corridor also illustrates a geopolitical dynamic: EU member states competing to host hyperscaler infrastructure are effectively embedding US technology dependencies into national industrial strategy, a trade-off that the European Commission has not yet formally resolved. Local community opposition reported by Bloomberg suggests the political economy of siting is becoming a constraint on deployment velocity independent of capital availability.

Allbirds, the former $4 billion wool sneaker brand that was weeks away from liquidation, has rebranded as NewBird AI and secured a $50 million convertible financing facility to enter the GPU-as-a-Service market, per the Wall Street Journal, TechCrunch, and Bloomberg. Shares surged over 400-580% on the announcement — a reaction that reflects less about NewBird AI's operational prospects and more about the market's willingness to assign speculative value to any entity claiming AI infrastructure exposure. The $50 million convertible facility is a confirmed closed financing, though the business model — competing with CoreWeave, Lambda Labs, and hyperscalers in GPU rental — faces severe capital intensity and customer acquisition barriers that the announcement does not address.

The episode is analytically significant not for Allbirds specifically, but as a data point in the broader pattern of distressed or repositioning public companies attempting to capture AI valuation premiums through pivots. The market's receptiveness to such pivots — even from operationally unrelated businesses — indicates that retail and speculative capital remains highly sensitised to AI infrastructure exposure, a dynamic that historically precedes crowded trade unwinds.