Capital & Industrial Strategy

Anthropic has received multiple inbound investment offers at valuations of $800 billion or above, according to Bloomberg, though the company has so far declined to accept them. For context, that would place Anthropic at roughly 2.1x its last confirmed $380 billion valuation and within striking distance of OpenAI's current implied valuation — which one investor who has backed both companies told the Financial Times via TechCrunch already requires assuming an IPO valuation of $1.2 trillion or more to justify. The investor characterised Anthropic's $380 billion mark as the relative bargain, a striking framing that suggests capital is rotating toward Anthropic as OpenAI's risk-adjusted return profile deteriorates. Separately, The Wall Street Journal reports Anthropic has added Novartis CEO Vas Narasimhan to its board and is eyeing a potential IPO as soon as this year.

The geopolitical dimension of Anthropic's positioning is accelerating in parallel. U.S. Treasury Secretary Scott Bessent publicly endorsed Mythos as a breakthrough that will keep America ahead of China, per Bloomberg, and the Treasury is actively seeking access to the model — framed as a vulnerability assessment exercise, per Semafor. Anthropic co-founder Jack Clark confirmed the company briefed the Trump administration on Mythos despite ongoing litigation against the government, per TechCrunch. Canada's AI minister has also publicly praised Anthropic's staged release approach, per Bloomberg. This multi-government embrace of Anthropic — including by an administration it is suing — reflects how frontier model capability has become a statecraft asset, and how that status translates directly into investor demand and implied valuation support.

SoftBank's lenders are actively recruiting additional banks to join the $40 billion loan facility backing its OpenAI investment, per Bloomberg. This is one of the largest single debt structures assembled for an AI investment and represents a significant test of whether institutional credit markets are willing to extend leverage at scale to a sector characterised by negative free cash flow, escalating compute costs, and valuation multiples that require heroic IPO assumptions to underwrite. The loan's viability hinges on SoftBank's ability to syndicate widely enough to distribute risk — the fact that additional banks are being recruited rather than commitments already being closed suggests some friction in the process, though Bloomberg does not describe this as a failure of syndication.

The broader context is that OpenAI's implied valuation of over $1.2 trillion — the figure investors cite as the floor needed to justify the last equity round — now sits in tension with its unresolved corporate restructuring (converting from capped-profit to for-profit), regulatory scrutiny, and competition from a re-rated Anthropic. Private credit markets are beginning to flag concern: Reuters reports Wall Street is monitoring private credit exposure to AI as disruption risks and fund outflows create concern. The SoftBank facility is the highest-profile test case of whether that concern translates into actual credit rationing.

Meta and Broadcom have extended and expanded their multibillion-dollar custom chip partnership, per Bloomberg and Reuters. The strategic logic is straightforward: Meta is deepening its ability to design application-specific integrated circuits (ASICs) tuned to its own AI workloads, reducing dependence on general-purpose Nvidia GPUs and giving it cost and performance advantages at the scale it operates. Broadcom's role as a custom silicon design and packaging partner — rather than a commodity vendor — positions it favourably against other chip suppliers. Notably, Broadcom CEO Hock Tan departed Meta's board concurrent with the deal announcement, suggesting a clean separation of governance and commercial relationship, though Bloomberg does not elaborate on the precise reason for the departure.

This deal fits a clear industry pattern: the hyperscalers most committed to AI at scale (Meta, Google, Amazon, Microsoft) are all pursuing custom silicon strategies to escape Nvidia's pricing power and supply constraints. The Meta-Broadcom expansion confirms that this is not a transitional hedge but a permanent architectural shift. For investors tracking the AI compute stack, it reinforces Broadcom's structural position as the preferred custom silicon partner for hyperscalers that lack in-house fab relationships with TSMC — a role that generates high-margin, multi-year contracted revenue.

ASML beat Q1 2026 revenue and profit consensus and raised its full-year sales guidance, per Bloomberg and CNBC, citing sustained AI-driven semiconductor production demand. ASML is the sole supplier of extreme ultraviolet lithography equipment, making its order book a reliable leading indicator of where chipmakers expect AI compute demand to flow over the next 12 to 24 months. A raised guidance at this stage of the year, after a period of macro uncertainty driven by U.S. tariff policy, is a meaningful signal that TSMC, Samsung, and Intel are not pausing their advanced node capacity expansion. TSMC's own demand signals, reported by the Wall Street Journal, corroborate this — AI-related wafer demand is tracking ahead of earlier projections.

On the government industrial strategy side, Italy's €211 million grant to 2D Photonics, per Bloomberg, is notable as a targeted bet on photonic interconnect technology for AI data centers — a layer of the stack that currently represents a chokepoint for throughput and energy efficiency. This is not a generalist AI fund but a specific technology subsidy aimed at a domain where European startups have credible IP. Italy is effectively using state capital to create a nationally anchored champion in a supply chain segment that is otherwise dominated by U.S. and Asian firms, consistent with the EU's broader AI Factories initiative.

Two enterprise adoption signals stand out today. Novo Nordisk struck a partnership with OpenAI to integrate its models across drug discovery and clinical data analysis operations, per The Wall Street Journal and Reuters — Novo Nordisk shares gained more than 4% on the announcement, an unusually large single-day move for a pharma major on a technology partnership, suggesting investors are pricing in material operational impact rather than treating this as a pilot. Amazon separately launched an AI research tool targeting early-stage drug discovery, per Reuters, positioning AWS as a vertical pharma AI platform rather than simply a compute provider.

In financial services, American Express released an agent payments infrastructure toolkit and — critically — announced it would indemnify merchants for losses caused by AI agent errors, per Fortune. The liability assumption is the analytically significant element: it transforms AI agent payments from a technology experiment into a commercially underwritten product, removing a key adoption barrier for enterprise merchants who have been unwilling to expose themselves to settlement risk from autonomous transactions. This is the kind of institutional backstop that accelerates deployment curves in regulated industries.