inference cost

NVIDIA’s Rubin platform — unveiled at CES 2026 — is being pitched as a generational leap in rack‑scale AI computing: a six‑chip, tightly co‑designed system that promises dramatically lower inference token costs, exaflops‑scale rack throughput, and a reimagined storage layer for long‑context...

The parallels between the dot‑com boom of the late 1990s and today’s AI surge are unmistakable: breathless narratives, new vanity metrics, and money piling into infrastructure and market share long before sustainable profits appear — but the differences matter just as much, and they determine...

Microsoft used the Goldman Sachs Communicopia + Technology Conference to lay down a clear, product‑level road map for how it expects AI to reshape the enterprise — centering that plan on Microsoft 365 Copilot, a multi‑model infrastructure called Azure AI Foundry, and a “front end as platform”...

Microsoft’s AI unit has publicly launched two in‑house models — MAI‑Voice‑1 and MAI‑1‑preview — signaling a deliberate shift from purely integrating third‑party frontier models toward building product‑focused models Microsoft can own, tune, and route inside Copilot and Azure. Background...

Microsoft’s message to founders is simple and forward‑looking: GPT‑5 is now part of Azure’s production stack, and Azure AI Foundry packages the model family, routing, safety controls and deployment plumbing startups need to move from experiment to revenue‑grade product quickly. The announcement...

Microsoft’s announcement that it has deployed two first‑party models — MAI‑Voice‑1 for speech generation and MAI‑1‑preview as a consumer‑focused foundation model — marks a deliberate strategic shift toward productized, in‑house AI and a clear attempt to reduce operational dependence on...

Microsoft’s move to ship MAI‑Voice‑1 and MAI‑1‑preview marks a clear strategic inflection: the company is no longer only a buyer and integrator of frontier models but a serious producer of first‑party models engineered to run inside Copilot and across Microsoft’s consumer surfaces. Microsoft...

Microsoft has quietly shipped its first fully in‑house AI models — MAI‑Voice‑1 and MAI‑1‑preview — marking a deliberate shift in strategy that reduces dependence on OpenAI’s stack and accelerates Microsoft’s plan to own more of the compute, models, and product surface area that power Copilot...

Microsoft has quietly moved from partner-dependent experimentation to deploying its own, production‑focused models with the public debut of MAI‑Voice‑1 (a high‑throughput speech generator) and MAI‑1‑preview (an in‑house mixture‑of‑experts language model), rolling both into Copilot experiences...

OpenAI's recent decision to rent Google's Tensor Processing Units (TPUs) to power ChatGPT and other AI products marks a significant shift in the AI infrastructure landscape. This move not only diversifies OpenAI's hardware dependencies but also sends a clear signal to Microsoft, its largest...