· 个股论点
简单定性回复：$IREN 确实前置了连接容量(Connected Capacity)，若从该角度（例如 2026 年上半年 1+ GW）对比 $NBIS 2026 年下半年的 1 GW，$IREN 胜出，且应利用这段时间签署更多合同。长期来看，Nebius 在两者间实现容量持平，且凭借全栈(Full-Stack)而非纯 IaaS 拥有更高利润率，这在 $MSFT 合同的每兆瓦营收中已有所体现。我记得曾对电力成本与利润率做过定量拆解，发现在 Oracle TI 事件中，利用率/编排(Utilization/Orchestration)比廉价电力更重要。关于 $NBIS 的趣闻：他们在 NJ 300 MW 设施的新绿地(Greenfield)站点与 Dataone 合作采用新的表后(Behind the Meter)方案，但短期可能落后于 $IREN。我不是能源专家，所以不个人预测天然气管道改造估算，仅采用管理层对连接/签约容量的预测。关于 $GLXY，同意加密货币会计处理的观点。这在加权时并非考量因素。
So just a quick qualitative reply, $IREN does frontload connected capacity and if you go from that standpoint (eg. 1+ GW H1 2026) vs. 1 GW from $NBIS H2 2026, $IREN wins and should be using that time to sign more contracts. Longer term, Nebius achieves capacity parity between the two and has higher margins from full-stack vs. pure iaas and you saw this from the revenue per mw from the $MSFT contract. I remember doing a quantitative breakdown on power costs with margins and found that utilization/orchestration mattered vs. cheap power during that Oracle TI fiasco. Fun fact about $NBIS, they have new behind the meter approaches with dataone on their greenfield sites with power generation on their 300 MW facility in NJ., but probably lags near term to $IREN. I'm not an energy expert so I won't personally go into projections regarding natural gas pipelines conversion estimates, so just going with connected/contracted capacity projections from management. For $GLXY agreed about the crypto accounting. That wasn't a factor when doing weighting.