To GZC Investors
As we enter the first quarter of 2026, I want to provide a comprehensive update on our portfolio positioning, the macro environment as we read it, and the specific thesis developments that have informed our allocation decisions over the past several months. This letter covers both pools — Technology and Commodities — and includes our current thinking on the structural forces that will define the investment environment through mid-2026 and beyond.
The headline observation is that the AI infrastructure cycle has not peaked — it has broadened. What began as a semiconductor and GPU bottleneck story has evolved into a full-stack infrastructure cycle that now encompasses power delivery, thermal management, data center construction, renewable energy procurement, and the physical materials required for every layer of buildout. This broadening is consistent with our Bottleneck-to-Ticker framework's prediction: structural demand cycles migrate through supply chains as each layer solves the constraint at its node, revealing the next layer beneath it.
Technology Pool: Deepening Conviction, Selective Additions
Within the Technology pool, the highest-conviction names remain NVDA and VRT — our two positions that represent the most acute supply chain bottlenecks in the current AI infrastructure cycle. NVDA's GPU architecture dominance and CUDA ecosystem lock-in continue to translate into pricing power that hyperscaler capex budgets have proven willing to absorb. VRT's power and thermal management position has been reinforced by every data center operator's acknowledgment that cooling and power density are the binding constraints on AI compute deployment pace.
We have made selective additions to two newer coverage names. Applied Digital (APLD) has reached a stage in its HPC campus buildout where the tenant pipeline provides sufficient revenue visibility to justify a position at current prices. Nebius Group (NBIS) has demonstrated operational execution in its European GPU cloud infrastructure that validates the European AI sovereignty thesis we outlined last year. Both positions are sized to reflect their earlier stage versus our more established holdings, but the conviction behind both thesis has strengthened materially over the past quarter.
AMD remains a monitored position where the thesis is intact — MI300X datacenter traction is real — but competitive dynamics with NVDA in the highest-margin AI training segment warrant ongoing scrutiny. We have not reduced our position but we are monitoring the gross margin trajectory of the data center GPU business as an indicator of whether the CUDA moat is eroding at the pace AMD bulls anticipate.
Commodities Pool: Energy Discipline and Critical Minerals
The Commodities pool has undergone the most significant evolution in our coverage over the past several months. We have exited several legacy energy positions that were primarily cycle-exposure plays without structural bottleneck characteristics, and concentrated the pool in names where the thesis is structural rather than cyclical.
The additions that define the current Commodities pool: Energy Fuels (UUUU) for its domestic uranium and rare earth processing position; Brookfield Renewable (BEP) for its hyperscaler clean energy procurement relationships; and the industrial metals exposure through ICOP and SLV, which we believe represents a non-discretionary demand theme from the physical materials required for AI infrastructure and energy transition buildout.
Diamondback Energy (FANG) and EOG Resources (EOG) anchor our E&P exposure — not as cyclical oil bets but as the structurally superior operators in a basin consolidation cycle where cost and inventory quality increasingly determine who compounds and who survives. The energy transition thesis does not imply zero oil demand in the near term; it implies that the operators with the lowest cost and highest capital discipline will capture disproportionate value from the remaining conventional energy cycle.
Macro Observations: AI Capex Visibility Extended
The most important macro observation from the Q4 2025 and Q1 2026 hyperscaler earnings season is that AI capex guidance was either maintained or increased across every major hyperscaler. This is not routine — it reflects genuine demand visibility from enterprise AI adoption reaching production deployment at a pace that is straining existing compute capacity. The demand signal is not just from hyperscaler-internal AI development; it is from enterprise customers consuming AI infrastructure services at a scale that justifies continued supply expansion.
We note that the power constraint thesis has become increasingly explicit in hyperscaler commentary. Multiple management teams have cited power availability — not GPU availability — as the primary limitation on their ability to deploy additional AI compute capacity. This is the VRT thesis expressed directly by the market participants who control the procurement budget.
Looking Ahead
Our focus for the balance of 2026 is the maturation of our HPC infrastructure positions (APLD, NBIS) from thesis validation to revenue execution, the nuclear energy timeline as reactor procurement decisions crystallize into signed contracts for UUUU, and the copper supply-demand dynamics as new mine development timelines become clearer relative to the pace of data center and EV buildout.
As always, we are available to discuss any aspect of the portfolio or our research process directly. We appreciate the trust you place in GZC's management of your capital.
— Ahijah Ireland, Founder & CIO, Green Zone Capital
