Clare Pleydell-Bouverie

Lumentum is a pure-play AI optical exposure with higher beta to the optical sector than Coherent, due to its focus on EML lasers, OCS, and CPO technologies, making it the best choice for investors seeking concentrated AI optical exposure.

Hyperscaler capital expenditure growth is significantly outpacing estimates (67% actual vs. 19% estimated). Additionally, "Physical AI" (robotics, autonomous driving) is emerging as a new demand layer not yet priced in. The market views NVDA primarily as a data center stock, ignoring the "Physical AI" cycle (robots/cars needing silicon). As autonomous driving (e.g., Wayve) and robotics scale, NVDA's addressable market expands beyond just LLM training. LONG. 2026 will be a year of accelerating "beat and raise" cycles for NVDA. Supply chain constraints (sold out) or a faster-than-expected drop in software ROI.

Dell's Q1 FY27 revenue surged 88% YoY, AI server revenue alone hit $16.1B, backlog of $51.3B ensures near-term visibility. The company raised full-year guidance significantly, confirming that AI server demand is real and rapidly expanding. Dell is now a core AI infrastructure supplier.

Snowflake's product revenue grew 34% YoY, large customers increased 29%, and remaining performance obligations rose 38% to $9.21B. The company is finally monetizing AI through data consumption, and its partnership with AWS and acquisition of Natoma position it as an AI data platform. The SaaS apocalypse narrative is reversing.

Astera Labs benefits from increasing AI server internal connectivity demand: PCIe retimers, CXL switches, and the new Scorpio X-series fabric switch for GPU clusters, with strong earnings guidance and early sampling.

ARM's CPU architecture is expanding from mobile into data centers and AI servers, driven by AI agentic AI requiring more CPU coordination, NVIDIA's Grace and Vera CPUs being ARM-based, and ARM's own AI CPU strategy targeting $15B revenue by 2031.

UBS raised its worst-case default scenario for private credit to 15%. Approximately one-third of private credit exposure is to the software sector. If AI disrupts legacy software cash flows (as per the SaaS thesis above), the lenders to these software companies (Private Credit/BDCs) will face massive impairments. Software is "asset-light," meaning recovery rates in bankruptcy will be near zero. SHORT. The "wheels are coming off" the asset class due to its exposure to disrupted tech. M&A activity picks up, allowing distressed software firms to be acquired rather than defaulting.

AI allows for "Software 2.0" (accelerated compute) to displace "Software 1.0" (legacy code). New AI-native startups can compete with lower costs. Legacy SaaS companies historically commanded premium valuations (10x revenue) due to "recurring revenue" safety. AI destroys this moat, leading to margin compression (pricing power loss) and multiple compression (re-rating to 3-4x revenue). SHORT/AVOID. 90% of the software sector faces further de-rating. Indiscriminate selling creates value traps where quality proprietary data companies are oversold.