AI News Roundup — May 7, 2026

Previously: AI & Tech News Roundup — April 29

Microsoft: AI Adoption Hits 17.8% of Working-Age Population

Microsoft published its latest Global AI Diffusion Report today, showing that AI usage among the world's working-age population climbed from 16.3% to 17.8% in Q1 2026. The report tracks actual AI tool usage across markets — not consumer sentiment or stated intent — making it one of the more grounded measures of real-world adoption.

The 1.5 percentage point quarter-over-quarter gain suggests the diffusion curve is still accelerating, not plateauing. That's notable given that adoption was already at a significant base — early-stage S-curves tend to accelerate, and the data points in that direction.

Google, Microsoft, and xAI Agree to Pre-Release Government Testing

The Center for AI Standards and Innovation (CAISI) announced agreements with Google DeepMind, Microsoft, and Elon Musk's xAI allowing U.S. government evaluators to test AI models before they are publicly released. The deal mirrors earlier agreements CAISI reached with OpenAI and Anthropic in 2024, effectively bringing all the major frontier labs into the framework.

The arrangement — voluntary for now — represents the Trump administration's preferred approach to AI oversight: industry collaboration over mandated regulation. Whether pre-release government access translates into meaningful safety intervention or functions primarily as optics is the core debate among AI policy watchers. What's clear is that regulators are no longer asking to be looped in after deployment.

Anthropic ARR Eclipses OpenAI

A significant competitive milestone landed this week: Anthropic's annual recurring revenue has reportedly surpassed OpenAI's for the first time. Anthropic is tracking toward a $30 billion ARR, while OpenAI trails at approximately $24 billion.

The reversal reflects Anthropic's strong enterprise traction — particularly in regulated industries (finance, healthcare, legal) where the company's safety-first positioning resonates with procurement teams. OpenAI still holds the larger consumer brand and the broader developer ecosystem, but the revenue gap now tells a different story about where enterprise AI spend is actually going.

Anthropic hits $30B ARR, overtaking OpenAI's $24B for the first time.

OpenAI Releases GPT-5.5 Instant

OpenAI released GPT-5.5 Instant on May 5, positioning it as a more capable agentic model built specifically for multi-step task completion. The release notes emphasize the model's ability to handle complex, ambiguous instructions — checking its own work and carrying tasks through to completion rather than stopping at a single output. It builds on the GPT-5 architecture with improved tool use and reasoning performance across extended task chains.

DeepSeek V4-Flash: 284B Parameters, 13B Active

DeepSeek's latest entry is V4-Flash — a 284-billion parameter Mixture-of-Experts model that activates only 13 billion parameters per token during inference. Released in late April and refined through this week, the model delivers frontier-level performance at a fraction of the compute cost of dense architectures.

DeepSeek continues to demonstrate that Chinese AI labs can match Western frontier performance while undercutting on efficiency by a significant margin. V4-Flash specifically targets inference cost — the expense that compounds at scale — and positions itself as the go-to model for high-volume API use cases where per-token pricing matters.

Oracle Plans $50 Billion Data Center Expansion

Oracle announced plans to raise up to $50 billion to fund a sweeping expansion of its global data center network, with the build-out specifically targeted at supporting generative AI workloads. The scale of the commitment underscores the infrastructure arms race currently underway: AI compute demand is growing faster than existing capacity, and cloud providers are treating data center construction as a durable strategic moat.

Oracle's positioning here is deliberate. As AWS, Azure, and Google Cloud compete at the top of the market, Oracle has carved a niche as an enterprise-friendly alternative — and is now betting heavily that its existing enterprise relationships translate into AI infrastructure contracts.

SpaceX and xAI Announce Merger Plans

Elon Musk announced plans to merge SpaceX and his AI venture xAI, with the stated goal of embedding Grok models directly into SpaceX's operations. The vision: accelerate the development of autonomous spacecraft and robotic Mars colony infrastructure using xAI's models.

The announcement is characteristic of Musk's cross-company integration approach. The practical mechanics — how language models contribute to orbital mechanics and spacecraft autonomy — remain underspecified. What's clear is that the merger creates a single entity combining xAI's AI capabilities with SpaceX's hardware and deployment infrastructure, which has significant implications for both companies' government contracts.

Snap Cuts 1,000 Jobs, Cites AI

Snap announced layoffs affecting approximately 1,000 employees alongside the closure of over 300 open roles, with CEO Evan Spiegel citing "rapid advancements in artificial intelligence" as a primary driver. The announcement follows a pattern now visible across consumer tech: AI capability improvements are reducing the headcount required to ship products, with effects now showing up in workforce numbers rather than just executive commentary about efficiency.

Snap's case is particularly pointed given the company's scale — the cuts represent a meaningful portion of its workforce and signal that AI-driven headcount reduction isn't confined to struggling companies. Even platforms with large active user bases are finding that fewer engineers are needed to maintain and improve products at the current pace of AI tooling advancement.

The Week in Context

For more on the AI-crypto infrastructure crossover, see: Bitcoin Miners Are Becoming AI Infrastructure Companies.

The through-line this week: AI is simultaneously becoming more capable (GPT-5.5, DeepSeek V4-Flash), more embedded in government oversight frameworks (CAISI agreements), more concentrated in revenue terms (Anthropic overtaking OpenAI), and more consequential for employment (Snap, a series of similar announcements across the industry).

The Microsoft adoption data is the most important single data point of the week — 17.8% of the global working-age population actively using AI tools is a number that compounds. The downstream effects on labor markets, productivity, and competitive dynamics are still early innings.