AI Infrastructure, Data Centers Drive New Deals

Marvell Technology has completed its acquisition of Celestial AI, adding optical interconnect technology to its Data Center Group. The deal is aimed at expanding Marvell's connectivity offerings for next‑generation AI and cloud data center architectures, where high‑bandwidth links between accelerators are a bottleneck.

Marvell, which trades around $80 a share, has delivered a 52.2% return over five years but is down 27.2% over the past year. The company and its investors are watching how effectively Celestial AI’s optical interconnects can be integrated into Marvell’s portfolio and converted into design wins and deeper relationships with large cloud and AI customers.

The acquisition aligns with existing investor narratives that center on Marvell growing through custom AI silicon and high‑speed interconnects for hyperscale data centers. Both consensus and more bullish views see Marvell leaning into AI data center demand, with Celestial AI’s technology positioned at the intersection of custom chips, optical links and long‑term cloud contracts.

Salesforce has been awarded a 10‑year, US$5.6 billion contract with the U.S. Army to deliver secure data and AI technologies. The agreement is described as one of the largest government cloud and AI deals to date and is focused on federal digital modernization across Army systems.

In parallel, Salesforce launched EMBERPOINT, an AI‑driven public safety alliance with Lockheed Martin, PG&E and Wells Fargo. EMBERPOINT is intended to apply Salesforce’s AI agent platforms, autonomous technology and data tools to wildfire prevention and critical infrastructure use cases.

These developments highlight how Salesforce’s core strengths in cloud software, data management and AI are being applied beyond customer relationship management, into long‑duration government projects and regulated industries. At a share price of about US$191, Salesforce is trading well below some analyst target ranges while pursuing these large AI‑related opportunities.

Caterpillar has partnered with American Intelligence & Power Corporation and Boyd CAT to provide 2 GW of dedicated onsite power for hyperscale AI data centers at the Monarch Compute Campus in West Virginia. The project will use fast‑response natural gas generator sets and battery energy storage to support high‑density, variable AI workloads without relying on public grids.

Brookfield Renewable already has large agreements with Microsoft and Alphabet to support their data center build‑outs. The company operates a global portfolio of hydroelectric, solar and wind assets, with storage and nuclear power services, and offers clean energy to enterprises looking to power AI infrastructure while meeting sustainability goals.

Eaton, an industrial focused on electrical products, has seen record backlog as data centers become an increasingly important customer category. Its backlog was up 34% versus 2024 as 2026 began, reflecting strong demand for power control equipment from AI‑driven facilities.

Texas Instruments has created a new data center sales category for its analog and power‑management chips, which are used to control energy flow and convert physical events into digital signals. Data center sales at Texas Instruments rose 64% in 2025, even as the broader business worked through a demand lull.

Brookfield Renewable’s role as a clean‑energy supplier to major technology companies positions it at the intersection of data center expansion and decarbonization. Its partnership structures include both a partnership share class, with a yield around 5.1%, and a corporate share class with a lower yield but identical underlying economics.

Separately, analysts have noted that AI infrastructure spending could reach US$500 billion or more in 2026, with companies such as Eaton, Texas Instruments and Brookfield Renewable cited as potential beneficiaries. These firms provide power systems, analog chips, and renewable generation that are critical for the build‑out of AI data centers.