Capital & Industrial Strategy

Broadcom delivered a strong forward earnings picture — guiding to over $30 per share in fiscal 2028, well above Wall Street estimates — anchored by an explicit forecast of surging AI chip sales over the next two years. CEO Hock Tan has tied this directly to deepening relationships with hyperscale AI labs building custom accelerators, a segment where Broadcom's ASIC business competes with Nvidia on total cost of ownership rather than raw performance. Bloomberg and Reuters both note the current-quarter revenue miss, which the market treated as a near-term execution issue rather than a structural problem given the 2028 guidance confidence.

The strategic significance is that Broadcom's forward bookings imply the hyperscalers — Google, Meta, and others — are continuing to write larger cheques for custom silicon rather than consolidating entirely on Nvidia's merchant GPU stack. This bifurcation of the AI chip market into Nvidia-led merchant silicon and Broadcom-led custom ASIC is now a confirmed structural feature, not a transitional phase. CNBC reported the current-quarter guidance disappointed on revenue, a tension investors should watch — near-term supply or customer timing risk does not negate the multi-year thesis but creates volatility around execution.

The Trump Justice Department filed an amicus brief in the New York Times v. OpenAI case arguing that restricting AI training on copyrighted material would undermine US competitiveness and create national security vulnerabilities. The brief explicitly frames fair use for AI training as a matter of national interest, stating the US has 'a strong interest in continuing to develop a robust and competitive artificial intelligence industry that sets the standard globally.' Reported consistently across Wired, WSJ, FT, and Reuters.

This is not merely a legal filing — it is industrial policy executed through litigation strategy. By intervening, the administration reduces the probability of an adverse precedent that would force AI companies to license training data at scale, a cost that would disproportionately burden startups and mid-tier players relative to incumbents with deep pockets. The simultaneous G20 push — where the US also urged member nations to allow AI training on creators' work — signals a coordinated international dimension to this strategy, not an isolated domestic legal move.

All G20 members unanimously adopted US-proposed guidelines at the Innovation Ministerial in Chapel Hill, North Carolina, calling for minimal regulatory intervention on AI and other emerging technologies. Commerce Secretary Lutnick used the forum to pitch adoption of US AI products and push data center buildout globally. Jensen Huang and Sam Altman attended alongside government ministers, with tech CEOs explicitly arguing that strict rules would 'strangle an industry capable of supercharging global growth.' Bloomberg, WSJ, and CNBC all confirm the unanimous adoption, though the guidelines are non-binding.

The strategic read here is that the US is attempting to export its permissive regulatory model before the EU's AI Act framework becomes the de facto global standard by virtue of market size. Non-binding guidelines are a weak instrument, but unanimous G20 adoption creates political cover for member governments resisting domestic calls for stricter AI rules, and it shifts the baseline of what 'responsible AI governance' means in multilateral settings. This is regulatory arbitrage executed at geopolitical scale.

Three enterprise infrastructure companies reported results or guidance this week that collectively argue AI spending is transitioning from discretionary experimentation to core operational investment. Snowflake raised its full-year revenue forecast and saw its stock surge 22%, with management citing an AI coding agent as a material adoption driver — five consecutive quarters of 29% Atlas growth at MongoDB points to the same structural demand. Reuters and CNBC covered Snowflake; Reuters confirmed Dell raised its annual forecast on AI server demand, with CNBC noting Dell has become a critical bellwether for the AI infrastructure trade.

HPE also raised forecasts on AI demand but saw shares fall on supply concerns — a nuance worth tracking. Reuters reported the split verdict, which suggests component constraints remain a live risk even as end-demand is validated. A trust-deployment gap is also emerging: a SAS survey cited by Semafor found 90% of enterprises have AI agents in decision-making roles but only 66% trust them — a gap that creates demand for AI security and governance tooling.

Two capital events this week confirm that AI security is consolidating into a distinct investment category. Palo Alto Networks paid approximately $500 million for Console, a Thrive Capital-backed AI IT service automation company — a closed deal with confirmed terms per sources cited by TechCrunch. Separately, HiddenLayer closed a $100 million round as enterprises seek tools to secure AI agents, their tool integrations, and agentic workflows — confirmed by TechCrunch. These are separate transactions targeting different parts of the stack: Console focuses on AI-driven IT automation, HiddenLayer on monitoring and securing AI model deployments.

The strategic logic of the Palo Alto acquisition is clear: as enterprises shift IT operations toward agentic AI, the attack surface expands and the buyer of security tooling converges with the buyer of automation tooling. Palo Alto is securing distribution into agentic IT workflows before the category matures. The residual competitive dynamic is notable — TechCrunch's sources identify Sequoia-backed Serval as now the de facto startup leader in AI IT service automation following Console's exit, meaning the race for the next acquisition in this space has effectively reset with one fewer credible target.