The Physical AI Layer: Analogue and Edge Take Centre Stage
Last month's earnings round was the clearest validation yet that the AI infrastructure cycle is real and working. The major cloud platforms all reaffirmed, and in places raised, their AI capital expenditure plans alongside Q1 prints, with combined commitments now exceeding $650bn for 2026. The market rewarded operators that are converting AI capex into visible cloud revenue at scale, while punishing those where the link between spend and returns remains unclear.
For the next leg of the cycle, attention is shifting one layer below the hyperscalers, to the two places where AI physically shows up.
AI in Everything Physical
AI does not replace the physical world; it amplifies it. Every unit of intelligence deployed requires a multiple of analogue, power, and control chips to make it real. Twenty years ago, a single car used a handful of chips. Today it uses hundreds, potentially thousands, and that is before the robotics and automation wave fully arrives. Factory automation, warehouse logistics, EV charging, and grid digitisation are all pulling analogue content sharply higher per unit of output. European analogue chip manufacturing holds a structural position in physical-AI silicon comparable to what GPU makers hold in compute, yet the market still prices it as cyclical automotive exposure. In the community's words: "AI might be the brain, but analogue is the oxygen and the lungs."
Edge AI Is Arriving
Running AI models on the device rather than round-tripping to the cloud is becoming essential as agentic AI assistants require continuous background processing, multi-sensor fusion, and on-device security. None of today's installed base of phones, cars, or wearables was designed for this. That sets up a structural upgrade cycle. Recent quarterly results confirmed the thesis: custom silicon deals with hyperscalers are moving to shipment, automotive chip revenue is growing at nearly 40% year on year, and at 12–14x forward earnings versus 30–40x for GPU leaders, edge AI silicon remains significantly underpriced relative to its expanding addressable market.
What Drove the Collective Ideas This Week
WoW: +5%
The Conviction ideas express a concentrated set of investment themes: AI infrastructure and its second-order demand pull through power, grids, and industrial capacity; power scarcity and reliability; a turning capex cycle; real assets and commodity-linked inflation hedges; security and resilience spend; and the institutionalisation of digital assets.
This week's +5% advance was broad-based, with the strongest contributions coming from edge AI, peptide manufacturing, and European critical minerals. Edge AI led as new mass-market chip platforms extended on-device intelligence from flagship handsets into mid-tier and entry-level segments, confirming the structural upgrade cycle is diffusing across every price point. Peptide manufacturing re-rated as the GLP-1 demand engine strengthened further, with oral formulations generating billions in their opening commercial quarter and weekly prescriptions exceeding 200,000, all flowing through contracted pipelines guided for 35–45% revenue growth this year. European copper supply security advanced on a new feedstock agreement that reinforced the conviction read on domestically sourced critical minerals at a moment when copper is structurally bid by the AI grid buildout, electrification capex, and a hardening EU policy framework.
On the weaker side, geoscience and seismic data gave back ground as a softening oil tape and the US-Iran diplomatic rapprochement triggered profit-taking, despite constructive operational newsflow underneath. The clean-power infrastructure complex continued to grind higher across the board: behind-the-meter fuel cell licensing, geothermal-plus-storage inflection, and nuclear data centre partnerships all advanced, confirming that dispatchable, decarbonised baseload is becoming the binding constraint for hyperscaler expansion.
Best Content Shared This Week
📚 AI Is Showing Up in the P&L
Why Read? The first real wave of AI-driven earnings is beginning to flow through public software, and the results challenge the idea that AI destroys the application layer. Twilio, Datadog and Atlassian all reported accelerating growth, rising usage and stronger guidance as AI workloads increased demand across communications, observability and knowledge-work platforms. The piece argues the real winners are not the companies fighting AI, but the incumbents whose infrastructure, data and distribution AI increasingly depends on. Read more
🎧 Agentic AI and the Next Tech Monopoly
Why Listen? Ben Rogoff breaks down how the AI cycle is shifting from copilots to fully agentic systems, and what that means for market structure, competition and valuations across technology. The discussion explores whether AI could create new natural monopolies, how to identify the likely winners and losers, and why the current phase of the cycle may look very different from previous software booms. Listen here
Stock of the Week
SpaceTech Gets Fresh Firepower
Seraphim Space has raised approximately £137m through a C Share issue, the largest fundraising by a UK investment company since 2023. The raise brought in support from existing shareholders, new institutions and retail investors, giving the trust fresh capital to deploy into a pre-identified pipeline of SpaceTech opportunities.
The timing matters. Space infrastructure is increasingly relevant across defence, climate, communications, cyber security and mobility, and Seraphim remains the only listed UK vehicle dedicated exclusively to the theme. With a debt-free balance sheet and new capital behind it, the trust is better positioned to back early-stage and growth-stage companies as the space economy matures.
The key risk is still valuation volatility, as returns depend heavily on private portfolio marks rather than steady cash flows. But if demand for space-based infrastructure continues to accelerate, Seraphim now has the firepower to lean into that opportunity.
