普通外文研报
Top 10 themes from the 2026 RBC Capital Markets Canadian Private Technology Conference
研报英文原文证据摘录
Top 10 themes from the 2026 RBC Capital Markets Canadian Private Technology Conference
sed products sold on business
outcomes rather than traditional SaaS subscriptions. Across the conference, AI is no longer a
peripheral tool but a mandatory operational layer. One panelist indicated that AI adoption is
"non-negotiable" internally, with token usage actively measured and employees expected to use
AI agents as a core part of their workflow.
Theme 5: Hardware, power, and physical constraints on AI scaling. Several sessions revealed
that AI's continued scaling faces physical constraints fundamentally different from software
challenges. The AI hardware stack has three layers that are improving at different rates:
processing (growing exponentially), memory (growing much more slowly), and interconnect
(the critical bottleneck). In some cases, expensive GPUs are sitting idle waiting for data
because interconnect cannot keep pace. Power, not capital, was identified as a key fundamental
constraint. Tech companies have historically underestimated how slow it is to deploy power
infrastructure globally and are now coming to terms with power as the binding constraint.
Regarding interconnects, optical delivers the equivalent performance at a fraction of the power
consumption of copper-based solutions. Additionally, optics extend connectivity to tens of
thousands of GPUs, enabling unified compute at scale. For physical AI use cases, edge hardware
with embedded AI chips enables on-device inference without cloud dependency, limiting only a
tiny fraction of data flowing through expensive frontier models.
Theme 6: AI natives claim wins against incumbents. Multiple sessions suggested that AI native
companies are winning competitive bake-offs against established players. For example, an AI
本摘录由系统从所标注的 PDF 证据页直接提取并保留英文原文,不做批量翻译;登录后在阅读器切换中文时才按需翻译。
打开研报阅读器