Capital & Industrial Strategy

Anthropic has confirmed a $35 billion cloud computing agreement with Lambda, a data centre operator backed by Nvidia, which will supply the chips and hold the lease on the Texas facility. The deal is confirmed as closed, per reporting from Bloomberg, WSJ, and Reuters. The scale dwarfs conventional cloud procurement and is better understood as a balance-sheet manoeuvre: by having Nvidia hold the lease, Anthropic secures massive compute without the capital liability appearing on its own books, while Nvidia deepens its strategic entrenchment in the frontier model ecosystem.

This deal illustrates a structural pattern now visible across the AI stack: chip companies are becoming infrastructure financiers. Nvidia's willingness to backstop Lambda's lease is functionally a subsidy to Anthropic's scaling ambitions, tying a major model developer to Nvidia silicon for years. For competitors using AMD or custom silicon, this kind of financially engineered lock-in is difficult to replicate. The arrangement also raises questions about what obligations Anthropic carries — any shift in its compute strategy or a hyperscaler partnership expansion would now have to navigate a $35 billion anchor contract.

Nvidia's $3.5 billion investment in Taiwan's MediaTek — its largest-ever direct investment outside the US — is a direct response to the threat of Big Tech building its own AI silicon. As TechCrunch reports, the strategic logic is clear: as Google (TPUs), Amazon (Trainium/Inferentia), and Microsoft (Maia) reduce their dependence on Nvidia for training and inference, Nvidia needs to extend its architectural influence into edge AI, consumer devices, and emerging markets where MediaTek has dominant share. Bloomberg reported MediaTek shares soared 10% on the news.

The partnership signals Nvidia's intent to make its CUDA ecosystem and AI software stack the default layer across a much broader set of silicon, not just its own GPUs. MediaTek's strength in mobile, automotive, and IoT chipsets gives Nvidia a route into form factors and supply chains it does not currently control. This is less a financial return play — the $3.5 billion is an investment, terms of which remain partially unconfirmed in detail — and more a strategic hedge against a world where the data centre GPU market faces margin compression from hyperscaler in-house chips.

OpenAI's advertising business reached a $1 billion annualised revenue run rate in under a year, per CNBC and Axios. This is not a trivial line item: it represents a fundamentally different monetisation architecture than the API-plus-subscription model that has defined AI company revenue to date, and it materially changes OpenAI's path to profitability. Anthropic's Super Bowl ad mocking the move now looks strategically costly — Anthropic has no equivalent revenue diversification.

The velocity of growth matters as much as the absolute figure. Reaching $1 billion ARR in under a year in a new business line suggests either exceptional advertiser demand for access to ChatGPT's user base, favourable CPM dynamics due to high-intent query context, or both. For investors tracking the pre-IPO landscape flagged by Madrona's Matt McIlwain — who noted on Bloomberg that an Anthropic IPO could open AI listings in 2027 — this ad revenue milestone makes OpenAI's revenue profile considerably more attractive to public market investors than a pure SaaS multiple story.

South Korea has proposed a record $597 billion national budget for 2027 with AI investment as a headline priority, per Reuters. Separately, the EU has ordered an AI supercomputer from Bull, the European compute infrastructure company, as demand has exceeded existing capacity, also per Reuters. These are distinct actions but together represent a coherent signal: sovereign governments are treating AI compute capacity as critical national infrastructure, allocating capital through direct procurement and budget mandates rather than waiting for private markets.

The UK is taking a complementary approach: the government is offering £100 million to homegrown AI startups to improve public services, per FT, partly motivated by political backlash against the Palantir-NHS contract. Adobe's $4 billion deal to offer free AI tool access in Saudi Arabia, per Reuters, sits in this same category: sovereign wealth and national AI strategies are driving commercial deal structures that would not exist in pure market conditions. The Trump administration's posture — warning communities opposing data centres they risk being 'backwards and poor' per CNBC — reflects a US industrial policy stance that is directionally similar, even if the mechanism is political pressure rather than direct subsidy.

Three distinct but structurally connected stories signal an accelerating enterprise AI adoption wave that is re-pricing professional services. Goldman Sachs, Morgan Stanley, and Citi are formally demanding lower legal fees from Big Law partners on the basis that AI has made routine work faster, per FT. Separately, Accenture, Capgemini, and the Big Four are facing client demands for savings as companies route AI transformation work around their traditional advisers, per FT. And activist investor Engine Capital is pressing EPAM Systems — a software services firm — for buybacks as markets price in AI-driven displacement of its model, per FT.

The common thread is that enterprise clients are now operationalising AI productivity gains as direct cost reduction on third-party services, not just internalising them. This is the mechanism by which AI transitions from pilot to structural market impact. A WSJ opinion piece WSJ framing AI as a defensive necessity — invest or lose competitive parity, but don't expect excess returns — captures the enterprise logic precisely. For PE-backed software companies, the FT's report FT on a $40 billion maturity wall of Covid-era buyout debt coming due in 2028 compounds the pressure: these assets were valued on growth assumptions that AI disruption is now undermining, and refinancing at current rates requires demonstrating AI resilience they may not have.