Capital & Industrial Strategy

Oracle and OpenAI have abandoned plans to expand a flagship AI data centre in Abilene, Texas after prolonged negotiations over financing and OpenAI's evolving infrastructure requirements, according to Bloomberg and the Financial Times. The collapse created an immediate opening for Meta, which is now in talks with developer Crusoe to lease the planned expansion capacity, with Nvidia facilitating the discussions. The failed partnership underscores the scale of capital required for frontier AI infrastructure and the operational complexity of coordinating multi-billion-dollar commitments between hyperscalers, AI labs, and specialised data centre developers.

The timing is particularly significant given Oracle's concurrent plans to cut thousands of jobs as it manages a cash crunch driven by massive AI data centre expansion commitments. The Oracle-OpenAI breakdown suggests either a mismatch in risk appetite — Oracle may have sought more binding commitments before deploying capital — or fundamental disagreement over the economics of long-term capacity contracts. Meta's willingness to step in immediately indicates continued aggressive infrastructure buildout by established hyperscalers, who may have stronger balance sheets and clearer unit economics than AI-native labs relying on venture backing.

The Department of Defense has officially designated Anthropic a supply-chain risk following the collapse of a $200 million contract over disagreements on military control of AI models, particularly for use in autonomous weapons and mass domestic surveillance, Bloomberg reports. This designation — historically reserved for adversary nation companies like Huawei — could cut Anthropic off from a broad range of US government contracts beyond the DoD. The Pentagon has turned to OpenAI instead, which accepted similar terms but subsequently saw ChatGPT uninstalls surge 295%, according to TechCrunch. Microsoft, Google, and Amazon have confirmed that Claude remains available to non-defense enterprise customers through their cloud platforms.

The US Commerce Department has reportedly drafted regulations requiring permits for AI chip shipments anywhere in the world without American approval, according to Bloomberg, and the Financial Times reports draft rules for civilian government contracts mandate that models be available for 'any lawful use' — a policy framework that would structurally disadvantage AI companies that impose safety restrictions on model access. This creates a strategic bind: companies that adopt restrictive acceptable use policies to differentiate on safety may be systematically excluded from government procurement, while those that accept broad government access face consumer backlash and potential employee attrition.

Blue Owl Capital led a $750 million debt financing for Vista Equity Partners' acquisition of Nexthink, a workplace IT analytics firm, demonstrating that private credit providers remain willing to back software deals despite mounting concerns over AI-driven disruption and liquidity pressures in the sector, Bloomberg reports. The deal is notable because it comes as traditional lenders have pulled back from software lending amid uncertainty about which business models remain defensible against generative AI substitution. Separately, JPMorgan is shifting more of the record Electronic Arts LBO financing toward junk bonds rather than leveraged loans, according to Bloomberg, suggesting that even blue-chip software assets are facing structural re-pricing of credit risk.

The divergence in financing approaches — with private credit backing vertical software in IT operations while gaming assets shift to bond markets — suggests lenders are making sector-specific bets on AI resilience. Nexthink's focus on workplace analytics may be perceived as AI-enhanced rather than AI-displaced, whereas gaming studios face uncertainty about generative AI's impact on content production costs and creative labour models. This is not a blanket retreat from software lending but a re-underwriting of which categories justify leverage in an AI-disrupted market.

AI chipmaker Cerebras has selected Morgan Stanley to lead a renewed initial public offering attempt, Bloomberg reports, marking a potential reopening of the IPO window for AI infrastructure companies after months of market volatility. Cerebras previously attempted to go public but timing was disrupted by macroeconomic conditions. The company's differentiated approach — building wafer-scale AI processors that compete with Nvidia's GPUs — positions it as a strategic alternative for customers seeking supply chain diversification, though it remains a fraction of Nvidia's scale. The IPO attempt comes as Korean chipmaker Rebellions also eyes competition with Nvidia and AMD, according to Bloomberg, signalling sustained investor interest in challenging the incumbents despite Nvidia's dominance.

The broader context is critical: the US Commerce Department's draft regulations restricting AI chip shipments globally without American approval would structurally advantage US-based chip designers like Cerebras while constraining Asian competitors. This creates a policy-driven tailwind for American AI semiconductor companies seeking to differentiate on supply chain security and regulatory compliance. However, the viability of these challengers depends on whether customers prioritise vendor diversification enough to absorb the switching costs and potential performance trade-offs versus Nvidia's mature ecosystem.

Google awarded CEO Sundar Pichai a compensation package worth up to $692 million in stock options explicitly tied to Alphabet's share price performance and the growth of autonomous vehicle unit Waymo over the next three years, the Financial Times reports. The structure signals the board's confidence in AI-driven revenue growth despite investor pressure to rein in capital spending. The package design — linking executive wealth creation directly to specific business unit performance — reflects a broader trend of boards using compensation to enforce strategic accountability as AI investments cross $100 billion annually across major tech companies. The explicit inclusion of Waymo, an autonomous vehicle business with unclear near-term profitability, indicates the board views physical AI applications as core to Alphabet's long-term value rather than ancillary bets.