Susana Schwartz
Nvidia-Groq deal first, but not last, pivot to inference
Arista Networks’ Q4 earnings indicate its revenue may double to $3.25 billion this year — the latest company to benefit from the current “inference boom.” According to Deloitte’s TMT Predictions for 2026, inference will be the primary revenue driver in AI data centers this year, with enterprises moving from pilot projects to full-scale production. As a result, inference workloads are expected to account for about two-thirds of all compute, with predictions that the market for inference-optimized chips to grow to over $50 billion. Following all of the big announcements last week at the India AI Impact Summit, it’s clear that the biggest hyperscalers — Amazon, Google, Meta, Microsoft, and Oracle — are set to spend approximately $700 billion in capex, with approximately $525 billion of that dedicated to AI infrastructure that supports inference. In addition to the high-profile Groq -Nvidia deal, others worth watching are the Nvidia-Meta alliance, AMD’s acquisition of Untether AI’s engineering team, and Intel’s pursuit of SambaNova. This highlights the role chip startups will play, as they are expected to grow in size as enterprise demand for inference expands. RCRTech will be reporting on the growth in inference and the increasing role of start-ups as trends evolve.
Susana Schwartz
Technology Editor
RCRTech
AI Infrastructure Top Stories
AI for DSS: Dynamic Spectrum Sharing solves data transmission constraints so AI-driven networks can operate with lower latency and greater efficiency. Symbiotically, can AI dramatically change how DSS works? RCR Wireless takes a closer look.
Tata partners with OpenAI: OpenAI will be the 1st customer of Tata Consultancy’s “HyperVault” data center, and ChatGPT Enterprise will be deployed across the Tata workforce — part of a deal to deploy enterprise tools and upskill talent in India.
AI Today: What You Need to Know
Tariffs and AI: The Supreme Court’s tariff ruling could impact AI-heavy firms like Nvidia, AMD, and Alphabet, which may be eligible for significant refunds, though impact of Trump’s retaliatory 15% global tariff is still unknown in the AI realm.
Micron accelerates DRAM fab: Micron is expected to spend upwards of $200 billion on new memory fabs in the U.S., with tens-of-billions slated to be spent overseas this decade, including new fabs in Idaho and a megafab in New York.
Samsung to overtake Nvidia?: Samsung Electronics is forecast to regain the world’s top semiconductor revenue ranking in 2026, with DRAM revenue expected to be up 170% YoY and NAND up more than 90%.
Pure Storage rebrands: Pure Storage has rebranded to Everpure, and will acquire data intelligence and orchestration company 1touch to extend its data management capabilities, adding data discovery and semantic context to its Everpure Platform.
PPL seeks $1B: PPL Corp has seen its data center pipeline in Pennsylvania swell by 23% compared last quarter, with around 25.2 GW of projects in advanced stages, but the company is now seeking $1 billion in equity to pay off short-term debt.
