Capital & Industrial Strategy

Anthropic is actively fielding unsolicited investment approaches that could value the company at close to $1 trillion, according to Financial Times, which would make it the most valuable private AI lab and overtake OpenAI's implied valuation. This is an announced intention, not a closed deal — no terms have been confirmed. The valuation thesis rests on accelerating revenue, with Claude's enterprise and API adoption driving the growth that earlier investors in the $7.3 billion Series E underwrote.

The strategic complexity deepens considerably with the confirmed Anthropic-SpaceX compute agreement, reported by Bloomberg and analysed by Reuters Breakingviews. Anthropic is paying an estimated $4 billion to access compute from SpaceX's Stargate-adjacent infrastructure — this despite Musk publicly labelling Anthropic 'evil' as recently as February, per Fortune. The arrangement illustrates how acute compute scarcity is forcing strategic compromises across competitive and ideological lines. Reuters Breakingviews notes this deal also complicates SpaceX's own AI valuation narrative, as Anthropic's capacity spend may be framed as validation of SpaceX infrastructure but simultaneously reduces Anthropic's margin profile heading into any IPO or secondary transaction.

Separately, WSJ reports that Anthropic's Mythos model launch triggered internal chaos in the White House AI strategy team, with the Trump administration now considering a framework for federal oversight of frontier model rollouts — a development that could insert regulatory risk into Anthropic's growth trajectory and any near-term valuation round.

Nvidia has committed up to $2.1 billion in data center developer IREN as part of a broader infrastructure partnership, confirmed by Bloomberg and Reuters. The deal goes beyond a passive equity stake: Nvidia gains preferred access to IREN's data center capacity at a time when GPU deployment bottlenecks are limiting the ability of hyperscalers and model developers to absorb chip supply. This is confirmed capital deployment, not a letter of intent.

The strategic logic is straightforward — Nvidia is converting chip revenue into infrastructure equity, securing downstream demand certainty and reducing its exposure to the cyclical capex decisions of hyperscaler customers. Combined with Jensen Huang's push to join Trump's China delegation (Bloomberg, Semafor) and the Corning manufacturing partnership (CNBC), Huang is simultaneously pursuing infrastructure lock-in domestically and market access diplomacy internationally. SoftBank's reported exploration of homegrown AI servers with Nvidia and Foxconn (Reuters adds another dimension — Nvidia is embedding itself into national AI hardware strategies across Japan, the US, and the Middle East simultaneously.

Big Tech's combined AI infrastructure spend has driven free cash flow to a decade low, per Financial Times, as the five largest US players commit an estimated $725 billion in aggregate capex. TSMC's confirmed 17.5% April revenue growth (Bloomberg) validates that semiconductor demand from hyperscalers remains robust and is not yet decelerating. Principal Financial Group is seeking to raise $3 billion across two data center funds for US and European markets (Bloomberg) — this is a fundraising target, not committed capital — reflecting institutional appetite to access AI infrastructure returns through private credit and equity structures rather than direct equity.

However, CoreWeave's Q2 guidance miss and widening losses (Bloomberg) introduce an important counterpoint. CoreWeave's model — borrow heavily, build capacity, sell GPU-hours — is being squeezed as hyperscalers build proprietary capacity and spot GPU pricing remains volatile. By contrast, Datadog's 31% stock surge on AI-driven observability demand (CNBC) illustrates where software-layer returns are materialising: companies that sell into the AI workflow rather than rent raw compute are demonstrating superior unit economics. Microsoft's Three Mile Island nuclear restart (Bloomberg) underscores that the energy constraint — not chips — is increasingly the binding limit on hyperscaler expansion, driving 20-year power offtake agreements that were unthinkable three years ago.

Moonshot AI has closed a $2 billion funding round at a $20 billion valuation, with annualised recurring revenue confirmed at over $200 million as of April 2026, according to TechCrunch. The round reflects genuine commercial traction — paid subscriptions and API monetisation — rather than speculative pre-revenue funding, which distinguishes it from the earlier wave of Chinese AI lab raises. The open-source positioning is deliberate: Chinese model developers are leveraging open weights as a distribution and ecosystem-building strategy that simultaneously accelerates domestic adoption and complicates US export control enforcement.

Baidu's chip subsidiary Kunlunxin is planning a dual IPO on Shanghai's STAR Market and Hong Kong (Bloomberg) — this is an announced plan, not a completed listing. The dual-market approach is strategic: STAR Market access targets domestic institutional and retail capital at AI-sector premium multiples, while Hong Kong provides international investor access. Baidu's motivation is capital extraction from a chip business that would otherwise remain buried in a conglomerate discount, while the listing also serves a national signalling function as China accelerates semiconductor self-sufficiency efforts. South Korea and the Netherlands' chip and AI cooperation discussions (Reuters) add to the picture of non-US chip alliances consolidating outside the Nvidia-dominated ecosystem.

Cloudflare's announcement that it is cutting 1,100 employees — approximately 20% of its global workforce — explicitly framed as a move to an 'agentic AI-first operating model' (Bloomberg, CNBC) marks a significant escalation in the rhetoric around AI-driven headcount reduction. The 18% stock decline on the announcement suggests markets are not rewarding the restructuring narrative at face value — investors appear concerned that the workforce reduction signals slowing organic demand rather than efficiency gains. Block's simultaneous upward revision to full-year profit guidance following AI-related job cuts (Bloomberg) provides the contrasting case: where restructuring is paired with credible margin improvement, the market is rewarding it.

The energy sector is emerging as an unexpected early adopter of industrial AI at scale. ADNOC Drilling is committing approximately $1 billion in AI and technology investment over the next year to reduce oil rig operating costs (Semafor), and Saudi Aramco is deploying a supercomputer with seven times its current compute capacity for seismic imaging and reservoir simulation (Semafor). These are confirmed capex commitments from state-owned energy majors — not pilots — representing AI moving into mission-critical industrial workflows with direct output implications.