Capital & Industrial Strategy

The US Commerce Department issued an export control directive on June 12 restricting Anthropic from offering its Fable and Mythos models to foreign nationals, citing risk of diversion to foreign military intelligence operations. Anthropic responded by disabling access globally — including for its own foreign-national employees — citing the impracticality of case-by-case enforcement. As of Tuesday, high-level talks between Anthropic leadership and White House officials have failed to produce a resolution, with one official acknowledging it will 'probably take more than a few days.' Both parties have stated publicly they are working toward a deal, but no terms are confirmed. Sources across Reuters, TechCrunch, and Politico suggest the directive may be partly retaliatory — a signal that the White House's stated laissez-faire AI posture has a short half-life when geopolitical or political interests intersect.

The collateral damage is significant and multi-directional. Dozens of cybersecurity professionals have formally protested the ban, arguing it handicaps defenders more than attackers, since offensive actors will find alternative tools while legitimate enterprise security teams lose access to state-of-the-art vulnerability detection. Meanwhile, the FT notes the controls are effectively a gift to Beijing, strengthening China's pitch that its own models offer more reliable access than US alternatives subject to sudden government shutdown. Cohere's chief AI officer confirmed to Bloomberg that the Canadian firm has seen a surge of inbound interest from governments and enterprises seeking non-US frontier model exposure. Separately, Upstage's CEO told Bloomberg the episode vindicates South Korea's sovereign AI investment thesis.

Nvidia priced a $25 billion investment-grade bond offering — its first debt issuance since 2021 — with demand described as strong across multiple tranches. The deal closed at the upper end of initial guidance, which Bloomberg and Reuters both confirmed. The FT notes the sale tested investor appetite for AI-sector exposure at a moment of intense debt issuance from tech heavyweights. Nvidia's decision to raise debt despite generating substantial free cash flow is strategically deliberate: it preserves equity while locking in credit at current rates ahead of what the company anticipates will be sustained multi-year capital expenditure on next-generation chip fabrication partnerships and data center buildout.

Nuveen's private credit chief Laura Parrott told Bloomberg she is actively positioning in the 'boring, plain vanilla' AI infrastructure layer — power, cooling, physical data center plant — rather than hyperscaler or model-layer plays. This is a meaningful signal: institutional fixed-income capital is now moving into AI infrastructure at scale, not just equity markets, suggesting the asset class is maturing beyond venture and growth equity into broader institutional allocations.

Salesforce has confirmed a $3.6 billion acquisition of Fin, an AI customer service platform whose flagship product handles customer queries across chat, email, WhatsApp, SMS, phone, and Slack. The deal is confirmed closed based on reporting from TechCrunch, Bloomberg, and CNBC, though full regulatory completion details are not specified. The strategic rationale is explicit: Salesforce intends to integrate Fin's team and technology directly into Agentforce, its enterprise agent-building platform, to accelerate capability depth in customer service automation — the highest-volume agentic use case in its installed base.

The $3.6 billion price tag for a customer-service AI platform reflects the premium Salesforce is willing to pay to avoid ceding the agentic layer to new entrants. The acquisition also signals that the transition from piloting to scaling AI agents in enterprise customer operations is now far enough advanced that incumbents feel urgency: building organically would take too long. Bloomberg Intelligence analyst Anurag Rana flagged the deal as central to Salesforce's strategy to win new enterprise AI business rather than just defend existing CRM contracts.

Sarvam AI, a Bengaluru-based startup focused on Indian-language AI models, closed a $234 million funding round led by HCLTech, which is taking a 10.5% stake and valuing the company at $1.5 billion. HCLTech's $150 million anchor commitment is strategically significant: one of India's largest IT services firms is not merely a passive financial investor but is explicitly positioning to own a piece of the domestic AI model stack it will sell to enterprise clients. The round, confirmed by TechCrunch and Reuters, coincides with parallel infrastructure moves: Jabil and Adani announced a partnership to build AI data center infrastructure in India, per Reuters.

The Anthropic export control episode has sharpened the sovereign AI argument in ways that abstract policy debate never could. South Korea's Upstage CEO told Bloomberg that the restrictions validate the case for domestic model development as a matter of strategic autonomy, not just commercial preference. The pattern across India, South Korea, and the EU is consistent: governments and domestic champions are treating the Anthropic episode as a proof point to accelerate investment in model and infrastructure layers that are not subject to US government override.

Audited financials reported by the FT show OpenAI spent $34 billion in 2025, driven by model development, compute infrastructure, and rapid organisational expansion. The figures are audited and therefore confirmed rather than estimated. The scale of expenditure — against revenue that, while growing, has not been publicly quantified at a matching level — frames the planned IPO as a test of whether public equity markets will value frontier AI labs as platform infrastructure businesses with durable moats or as R&D-intensive cost centres with uncertain monetisation timelines.

The FT also notes separately that AI giants are confronting a hard lesson about pricing power: model commoditisation is compressing the premium that frontier labs can charge, with Anthropic cited as one of the more rationally valued companies in its peer group until the White House intervention introduced a new category of regulatory risk premium. The combination of OpenAI's cost trajectory and Anthropic's regulatory exposure creates a challenging valuation environment for any frontier model lab approaching public markets.