Verified material developments: proposed consolidation of AI optical-connectivity supply; OpenAI’s announced grid-flexibility approach; and a proposed UK data-center campus tied to an existing power-station site. No additional material, verified changes were included for Compute, Platforms, Applications, Robotics or Cyber in this reporting window.
Coverage: verified announcements published November 2–3, 2026. Only material developments are included.
Power / Cooling / Networking — Marvell to acquire Lumentum
Marvell and Lumentum announced an all-stock acquisition valued at an implied $115 billion for Lumentum. The companies say combining Marvell’s connectivity products with Lumentum’s optical technologies and manufacturing capabilities will support higher-bandwidth, more power-efficient AI data-center connections. The deal is subject to shareholder and regulatory approvals and is expected to close in the first half of 2028. Why it matters economically: It signals that optical interconnects are becoming strategically important as AI clusters grow, and would consolidate more of the connectivity supply chain under one company. The announced synergies and market opportunity are company projections, not guaranteed outcomes. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1633978/000119312526442758/d163169dex991.htm?utm_source=openai))
Power / Cooling / Networking — OpenAI proposes flexible electricity demand
OpenAI announced its Grid Protectors program, proposing that future compute campuses reduce electricity use when local grids are strained and, where available, use on-site generation or storage. For its planned Effingham County, Georgia campus, OpenAI says its agreement with Georgia Power includes up to 1,000 MW of flexible demand response and that it will pay the project’s required infrastructure and electric-service costs. The campus is still in development. Why it matters economically: If implemented as described, flexible demand could help limit peak-grid stress and associated infrastructure costs—an increasingly important factor in data-center economics and in who bears the cost of new power infrastructure. The claimed benefits remain unproven at this project’s operating scale. ([openai.com](https://openai.com/index/grid-protectors-program/?utm_source=openai))
Infrastructure — Drax, Brookfield and Radiant plan UK data-center campus
Drax, Brookfield and Radiant announced a partnership to develop data-center capacity at Drax Power Station in North Yorkshire. The plan targets an initial phase of up to 100 MW, with an ambition to reach 500 MW between 2028 and 2031; the partners say total investment could exceed £20 billion. Progress depends on planning and regulatory and network approvals, as well as securing a tenant. Why it matters economically: Pairing a proposed data-center campus with an existing power-station site could help address the land-and-power constraints that shape where large AI infrastructure can be built. The announced investment and capacity are conditional plans, not committed operational capacity. ([drax.com](https://www.drax.com/news/drax-brookfield-and-radiant-announce-20bn-partnership-to-deliver-gw-scale-data-centre-capacity-at-drax-power-station/?utm_source=openai))
The merger remains subject to approvals and is expected to close in 2028. OpenAI’s campus and demand-response commitments are not yet demonstrated at operating scale. Drax’s capacity and investment figures are plans conditional on approvals, grid arrangements and a tenant. Announced targets and potential benefits should not be treated as realized results.