Azure’s $100B Year and OpenAI’s Efficiency Push

Compact Conversations for 2026-07-29: 5 AI stories, ai news worth knowing in just 5 minutes.

[Audio embed placeholder]

The Lead: Microsoft’s Azure Hits $100 Billion in Annual Revenue

Microsoft reported Azure revenue exceeded $100 billion for the full fiscal year 2026, a 41% increase. In the fourth quarter, Azure grew 43% year-over-year, beating expectations. The company also noted strong AI product adoption, with Microsoft 365 Copilot reaching over 30 million paid seats and GitHub Copilot hitting 50 million users.

Why it matters: Azure’s scale and accelerating growth signal the massive infrastructure demand driven by enterprise AI workloads, making it a key indicator for cloud spending and AI adoption trends.

Source: CNBC

Number to Know: OpenAI’s GPT-5.6 Model Family Aims for Frontier Efficiency

OpenAI detailed how its new GPT-5.6 model family balances capability and cost. The Luna model is priced 80% less than the flagship Sol model. Gains come from optimizations across the inference stack and agentic harness, with GPT-5.6 Sol itself helping to autonomously rewrite production code, cutting serving costs by 20%.

Why it matters: For teams scaling AI applications, these efficiency improvements directly impact the bottom line, making advanced model capabilities more accessible and cost-predictable.

Source: OpenAI Announcements

The Feed

OpenAI Introduces ‘Presence’ for Enterprise AI Agents

OpenAI launched Presence, a product for deploying trusted AI agents in production environments like customer support. It pairs model reasoning with company policies, guardrails, and escalation rules. OpenAI uses it for its own phone support, resolving 75% of issues without human help.

Why it matters: This represents a move from proof-of-concept to production-ready AI agents, offering a framework for enterprises to deploy AI in controlled, high-value workflows.

Source: OpenAI Announcements

Onyx Security Raises $113M to Safeguard AI Agents

Onyx Security raised $113 million in funding. The company focuses on monitoring and controlling AI agent behavior in production to ensure they operate within approved boundaries.

Why it matters: The significant funding round highlights investor focus on the growing need for governance and security tools as enterprises operationalize AI agents.

Source: Axios AI+

Report: One in Four Dollars Spent on AI is Wasted

A Harness report finds 25% of AI spending is wasted, with more than half of businesses lacking a dedicated owner for AI costs, leading to overspend and inefficient resource allocation.

Why it matters: As AI budgets expand, this governance gap becomes a critical operational and financial risk, emphasizing the need for cost accountability and management practices.

Source: CIO Dive

One Thing to Try

A simple first step to control AI spending is to designate a person or team to own AI-related cloud costs. Review your last quarter’s bill, flagging line items for model inference, vector databases, or AI-specific services. Monthly reviews by a single owner can surface waste and create accountability.

Sources

Transcript

Host A: Welcome to Compact Conversations, the show that compresses the day’s AI news into 5 minutes.

Host A: [curious] Today’s lead is Microsoft’s quarterly earnings. Azure revenue hit 100 billion dollars for the full fiscal year 2026, up 41 percent. In the most recent quarter, Azure grew 43 percent year-over-year, beating analyst expectations of 40 percent. Microsoft’s overall revenue came in at 90 billion dollars, also ahead of forecast, and shares jumped about 8 percent after hours.

Host B: The company also reported strong adoption of its AI products. Microsoft 365 Copilot now has over 30 million paid seats, up from 20 million just three months ago. CEO Satya Nadella said GitHub Copilot has reached 50 million users. [with emphasis] The finance chief projected Azure will grow 45 percent in the current quarter, so the momentum appears to be accelerating as enterprise AI workloads scale.

Host A: One number to know today is 80 percent. That’s how much cheaper OpenAI says its new GPT-5.6 Luna model is compared to its flagship Sol model. [thoughtful] OpenAI released a technical post today explaining how it’s balancing capability and cost across its model family. Terra performs as well as GPT-5.5 at half the price, and Luna is the fastest and most affordable option for developers trying to reduce inference costs.

Host B: The company says these gains come from years of optimization work in how it runs models—things like load balancing and speculative decoding—and in its agentic harness that orchestrates tools for Codex and ChatGPT Work. OpenAI claims GPT-5.6 Sol itself helped autonomously rewrite production code, cutting end-to-end serving costs by 20 percent. It’s part of a broader push to make frontier models more accessible to cost-conscious enterprises.

Host A: OpenAI also introduced a new product today called OpenAI Presence. [conversational] It’s designed for deploying AI agents in enterprise production environments—things like customer support or resolving internal IT requests. The product pairs model reasoning with company-set policies, guardrails, and escalation rules, so enterprises keep control over how agents behave in sensitive workflows.

Host B: OpenAI says Presence is already powering its own English-language phone support line, resolving 75 percent of inbound issues without human intervention. [with a small lift] That’s a concrete proof point. The company also named pilot customers: BBVA is exploring voice support in Mexico, SoftBank is testing Japanese conversations, and IAG is looking at support during severe weather. The product is available through a limited general availability program led by OpenAI’s forward-deployed engineers and select systems integrators.

Host A: In funding news, Axios AI Plus reports that Onyx Security has raised 113 million dollars. The company focuses on monitoring and controlling AI agent behavior in production—essentially helping enterprises verify that agents stay within approved boundaries and escalate when they should. The funding size, according to the report, signals growing investor confidence in the AI governance space as enterprises scale agent deployments.

Host B: And a report from CIO Dive, citing data from Harness, finds that 1 in 4 dollars spent on AI goes to waste. The report says more than half of businesses lack a dedicated owner for AI costs, which can lead to overspend and inefficient resource allocation. [skeptical] That governance gap is becoming a real operational concern as AI budgets grow and cloud bills climb. The report specifically notes this lack of ownership is a primary driver of wasted spending.

Host A: One thing to try is to audit your team’s AI cost allocation. [conversational] The Harness report suggests a simple first step: designate one person or a small team to own AI-related cloud spending. Pull the last quarter’s bill and flag any line items tied to model inference, vector databases, or AI-specific services. Having one owner review those costs monthly can surface waste and create accountability before budgets scale further.

Host B: You don’t need a new tool or a complex audit to start. Just visibility and ownership.

Host A: That’s Compact Conversations for Wednesday. More AI news tomorrow. Until then, happy prompting.