Capital & Industrial Strategy

Meta's announcement of roughly 8,000 layoffs — its largest reduction since 2023 — lands alongside a confirmed $135 billion data centre spending plan for 2026. Microsoft simultaneously offered voluntary buyouts to 7% of US staff, its first such programme, even as it prepares to deploy $140 billion in AI infrastructure this year. The pattern is unambiguous: both companies are treating human labour as a variable cost to be optimised while treating compute as the primary fixed investment. WSJ FT FT

The strategic logic at Meta is explicit: Zuckerberg has stated he expects AI agents to take on the work of mid-level engineers, enabling smaller teams to deliver more output. Microsoft's restructuring is more surgical — it is changing bonus and equity structures to redirect compensation toward AI-centric roles rather than eliminating capability wholesale. For investors, the question is whether these labour savings will materialise before the capex cycle demands return justification. At $135 billion in annual data centre spend, Meta needs to demonstrate AI monetisation at a scale that has not yet been achieved.

Project Prometheus has confirmed closure of a $10 billion round at a $38 billion valuation — a deal notable both for its scale and its strategic positioning in physical AI, meaning models designed to interact with and control real-world systems. Bloomberg Meanwhile, Cognition AI, maker of the Devin software engineering agent, is reported to be in early-stage funding talks at a $25 billion valuation, more than double its prior mark, driven by enterprise demand for AI-native software development tooling. Bloomberg The Cognition figure is unconfirmed and should be treated as indicative of negotiating range rather than committed capital.

These two datapoints bracket a significant dynamic: the market is bifurcating between foundation model labs commanding sovereign-scale valuations and application-layer companies targeting specific high-value workflows — legal, software development, physical operations — where AI delivers measurable productivity gains. Anthropic's parallel moves reinforce this: the confirmed deal with Freshfields to co-develop legal AI tools taps domain expertise the lab cannot replicate internally, while its hiring for a Europe data centre dealmaker signals operational scaling ahead of commercial expansion. FT CNBC

DeepSeek has released preview versions of its V4 model, billed as the most powerful open-source AI platform currently available — a direct challenge to GPT-4-class and Claude-class systems. The timing is deliberate: it arrives roughly a year after the V3 release that shocked Western markets with its compute efficiency claims. Simultaneously, the FT reports DeepSeek is raising external funding for the first time, targeting a $20 billion valuation, primarily to fund retention packages after multiple senior researchers defected to rivals. Bloomberg FT WSJ

Tencent's parallel unveiling of a flagship model upgrade — its first significant test for an OpenAI researcher it recruited — indicates Chinese frontier AI development is accelerating across multiple organisations simultaneously, not just DeepSeek. Bloomberg The White House's simultaneous accusation of industrial-scale AI IP theft by Chinese entities adds a geopolitical overlay: the Trump administration is hardening its IP enforcement rhetoric even as it softens its posture toward domestic AI labs like Anthropic. FT

Intel's 20% share price jump on Q1 earnings — revenue of $13.6 billion, materially ahead of consensus — driven by data centre CPU demand from AI agent workloads, validates the thesis that non-GPU compute will capture significant AI infrastructure spend. Texas Instruments posted its best single-day gain since 2000 on similar demand signals. Applied Digital confirmed a $7.5 billion data centre lease with an unnamed US hyperscaler, one of the largest single data centre transactions on record. Bloomberg Reuters

However, the infrastructure buildout faces structural friction. The WSJ reports Oracle is pushing Wall Street debt capacity to its limits as it finances AI data centre expansion through bond markets, compounded by power constraints and local opposition to data centre siting. Axios flags supply chain cracks — specific components, particularly networking equipment and specialised cooling systems, are creating bottlenecks that capex commitments alone cannot resolve. SoftBank's move to convert an Osaka factory into battery production for its own AI data centres illustrates how hyperscalers are beginning to vertically integrate energy storage to work around grid limitations. WSJ Bloomberg

Three government-level AI strategy signals emerged in parallel. First, the Trump administration is walking back its adversarial posture toward Anthropic — lobbyists and policy officials confirm the stance is softening despite Pentagon restrictions that remain formally in place — indicating the administration recognises it cannot simultaneously treat Anthropic as a national security risk and maintain US AI leadership. Politico Second, the UK government is in confirmed talks with Anthropic to provide British banks access to the Mythos model for cybersecurity applications, a significant signal that the UK is using procurement relationships with US AI labs as its primary industrial strategy tool rather than funding domestic development. FT

Nokia's CEO publicly warned that Europe risks falling materially behind the US and China on data centre build-out, adding a corporate voice to what has been primarily an academic and think-tank concern. Reuters Microsoft's $18 billion Australia commitment underscores how US hyperscalers are using bilateral infrastructure investment to lock in allied-nation government relationships, a strategy that simultaneously expands commercial reach and deepens geopolitical alignment.