Orange Polska SA (STU:TPA1) Q1 2026 Earnings Call Highlights: Strong Start with Robust EBITDA Growth

Orange Polska reported a strong start to the year, with EBITDA up 10% and core telecom revenue growth of nearly 5% year-on-year, supported by a 10% increase in the fiber customer base and market leadership in mobile number portability. Management expects Q2 2026 core telco revenue growth of about 4.8% and 5% to 7% CAGR in ICT/B2B through 2028, while noting weather-related CapEx delays of roughly 70 million zloty and ongoing volatility in memory-chip supply. The tone is constructive but tempered by slower prepaid trends, legacy declines, and supply-chain risks in ICT.

Analysis

This reads less like a one-off hardware pop and more like an early evidence point that AI capex is broadening beyond GPUs into the CPU layer and adjacent infrastructure. If enterprise and cloud buyers are pulling forward general-purpose compute refreshes, the first beneficiaries are not just the obvious accelerators but also memory, networking, and system integrators that ride a wider server bill of materials. The second-order effect is that the market may be underestimating how quickly OEM order books can re-rate when AI workloads start stressing CPU-side throughput, cache, and memory bandwidth rather than only accelerator demand.

The key near-term risk is that the narrative can outrun actual unit economics. A CPU-led AI cycle usually carries lower ASP uplift than GPU-led demand, so the upside to revenue may lag the sentiment move unless it comes with tighter mix and stronger foundry execution. If the rally is driven by optimism about a multi-quarter demand inflection, the reversal trigger is simple: any indication that the AI server build is still concentrated in a narrow set of hyperscaler programs rather than broad enterprise deployment over the next 1-2 quarters.

From a competitive standpoint, this is potentially constructive for the broader semiconductor ecosystem, but not uniformly. Suppliers with exposure to memory-intensive configurations and server platforms should see the best spillover, while names with limited AI content risk missing the multiple expansion even if they benefit modestly operationally. The contrarian view is that the market may be confusing cyclical restocking with a durable AI-driven step-up; if so, the move in the most levered names could be overdone relative to the pace of end-demand normalization.

Key Decisions for Investors

• Long INTC on a 3-6 month horizon only if the stock can hold post-gap levels for 2-3 sessions; use a call spread to cap downside, as the trade is currently sentiment-extended but could re-rate further if server demand broadens.
• Pair long INTC / short a high-multiple AI beneficiary with little CPU leverage over the next 4-8 weeks; the goal is to isolate a broadening-AI-compute theme while reducing exposure to pure multiple compression if the market cools.
• Add a tactical long in memory-exposed semiconductor names on any pullback over the next 1-2 weeks, because CPU-led server refreshes typically pull DRAM/NAND demand higher before it shows up in headline revenue.
• If you own expensive AI infrastructure names, trim 20-30% into strength; the risk-reward deteriorates if the market is simply pricing a faster but not larger capex cycle.
• Set a catalyst watch for the next two earnings cycles: if management teams cite AI-related CPU orders but do not raise full-year guidance, fade the move via short-dated call overwrites or put spreads.