Hotard预计半导体和内存供应将持续紧张到2028年，新内存晶圆厂要到2029年才投产。
限制来自台积电面临的需求（这点常被谈及）以及半导体供应链的其他环节。内存方面，现实是晶圆厂要到2029年才能建成投产。原文 (en)
"It's constraints in terms of the demand that TSMC has, which gets talked about quite a bit, and in other parts of the semiconductor supply chain," said Hotard. Hotard's expectation is that supply will remain "tight" into next year and even 2028 before there is much relief. "For things like memory, the reality is the fabs are not going to be completed and online until 2029."
Hotard 称 AI 与云的连接需求巨大且无法满足，因为算力受限时连接可提升算力利用率。
尽管市场上有不少情绪，但 AI 与云对连接的需求是巨大且难以满足的。原因在于算力受限时，连接成为提升算力资产可及性和利用率的另一种方式。原文 (en)
Despite a lot of the sentiment in the market, the demand for connectivity in AI and cloud is massive and it's insatiable. And it's driven by the fact that if compute is constrained, connectivity becomes another way to maximize the access to the assets or the utilization of the compute assets.
Hotard 预计诺基亚光网络与 IP 网络 2027 年合计增长 25% 至 30%。
对于 2027 年，我们现在预计这些业务合计增长 25% 至 30%，反映出底层市场的强劲以及我们抓住这一需求的地位。原文 (en)
For 2027, we now expect these businesses to grow between 25% and 30% on a combined basis, reflecting the strength of the underlying market and our position to capture that demand.
Hotard 警告称，AI 需求带来的供应链紧张范围广泛，并造成先进制程、内存和材料的结构性成本上涨。
这一次更加广泛，涉及所有先进制程节点、内存，也涉及部分采购材料如 PCB 等。这不只是短缺，也是我们在供应链中看到的结构性成本上涨。原文 (en)
This is much more broad-based, touches all the leading-edge nodes, touches memory, also touches some of the sourced materials, PCBs, et cetera. And it's not just a shortage. It's also a structural cost increase that we're seeing in the supply chain.
Hotard 认为电信市场没有即将到来的投资周期，将保持平稳，但成本通胀将推高价格。
我没有看到迫在眉睫的投资周期。我认为市场未来仍将持平。不过鉴于成本通胀，我们预计价格会上涨。原文 (en)
I don't see an imminent investment cycle. I think the market continues to be flat as we look ahead, and we've been pretty vocal and transparent about this. However, we expect -- given the cost inflation, we expect pricing to go up.
Hotard 称 RAN 正从专有芯片转向通用计算平台，诺基亚计划与英伟达合作引领。
科技行业长期以来不断将工作负载从专有芯片迁移到通用计算平台。RAN 正在进入同样的转变，由服务智能的独特需求驱动。我们正与英伟达合作，定位引领这一转变。原文 (en)
The technology industry has a long history of moving workloads from proprietary silicon to general-purpose computing platforms. RAN is entering the same transition driven by the unique requirements of serving intelligence. We are positioning Nokia to lead this shift in partnership with NVIDIA.