Daily AI Pulse — November 4, 2026

THOUGHT OF THE DAY

Telecom networks are emerging as a new layer of AI infrastructure. Fujitsu’s 1Finity platform combines NVIDIA AI Aerial with unified GPU infrastructure, positioning mobile network operators to offer AI services rather than simply transport data. If operators can turn existing network assets into deployable compute and AI platforms, they could become distribution and infrastructure partners—not just connectivity providers.

The AI buildout is creating investable demand in power conversion and optical connectivity. Flex’s planned separation of its Cloud and Power Infrastructure business, combined with its acquisition of EPC Power, highlights the value of power-conversion and grid-forming capabilities; reported moves in optical interconnects point to a parallel need for higher-bandwidth data-center fabrics. The opportunity is broadening from accelerators to the systems that move power and data around them, though project execution and customer demand will determine which suppliers capture durable returns.

Update: political resistance to data-center expansion is shifting from grid access to project approval. Reports of a Texas election-driven pivot toward a moratorium on new data-center projects extend the permitting and connection risks already facing developers. A moratorium would be a more direct constraint than higher costs or delayed grid connections, putting a premium on projects with secured sites, power, and community support.

COMPUTE & SEMICONDUCTORS

  • Celestica is projected to deliver 85% revenue growth in AI hardware, underscoring continued demand for rack-scale systems. The key question is whether that growth converts into sustained margins as customers scale deployments.
  • Astera Labs is positioning around AI-fabric switching, while Marvell’s reported move involving Lumentum reflects the strategic importance of optical interconnects. Bandwidth and power efficiency between accelerators are becoming as important as accelerator supply as clusters scale.
  • Oracle’s reported cancellation of a $7 billion deal with Tencent amid tighter U.S. export controls is a reminder that geopolitical restrictions can disrupt large compute and cloud commitments.

DATA CENTERS & INFRASTRUCTURE

  • Flex plans to spin off its Cloud and Power Infrastructure business as Axiom Solutions; the acquisition of EPC Power adds power-conversion and grid-forming capabilities. That combination targets a critical deployment need: integrating large, variable data-center loads with available power.
  • Schneider Electric, NVIDIA, and HPE introduced a modular data-center reference design focused on faster deployment and energy efficiency. Modular designs may help shorten build timelines, but they do not remove constraints around power availability, permitting, or project economics.
  • 3M’s expanded work with AWS on Expanded Beam Optical connectivity highlights demand for reliable fiber connections in high-density data centers.

ADOPTION & MONETIZATION

  • Palantir is reported to have grown revenue 96%, a strong signal that enterprise AI demand can translate into commercial growth. The result supports the case for platforms that connect AI capabilities to deployed customer workflows, though the summary provides no detail on growth drivers or profitability.
  • Apple is reportedly capping some employees’ AI usage at $25 per day while relying on external models including Claude, ChatGPT, and Gemini. The combination raises questions about internal adoption and in-house capability, but the reported usage limits alone do not establish a broader change in product strategy.

POSITIONING IDEAS

Bullish

  • Celestica — projected AI-hardware revenue growth supports exposure to rack-scale systems, where demand is extending beyond chips.
  • Power-conversion and optical-interconnect suppliers — the reported Flex-Axiom move and the focus on optical connectivity support demand for infrastructure that enables AI clusters to operate and scale.

Bearish

  • Apple — reported internal AI usage caps and reliance on external models point to execution and adoption risks relative to AI leaders.
  • Data-center developers with Texas exposure — a reported push toward a project moratorium would add approval risk on top of existing grid and permitting constraints.

This content is for informational purposes only and does not constitute financial, investment, or trading advice. Always consult a qualified financial professional before making any investment decisions.