AI Newsletter

Mistral has launched European Compute Units and regional inference hosting – including support for third-party models like GLM 5.2. Combined with the multi-billion dollar Microsoft deal for thousands of Nvidia Vera Rubin GPUs, Mistral is positioning itself as a sovereign infrastructure layer for regulated European industries. However, the question remains whether Mistral can technologically keep pace with OpenAI, Anthropic, and Google, jeopardizing its strategic credibility.

Intel established itself in 2026 as an unexpected AI beneficiary: Custom Silicon grew to approximately 2 billion dollars in annual revenue – triple year-over-year. Xeon CPUs and Advanced Packaging directly benefited from the AI infrastructure boom. The finding signals that the AI hardware market is broader than expected and not dominated by Nvidia alone.

At AMD's Advancing AI 2026 event, CEO Lisa Su presented a historic turning point: inference demand exceeds training for the first time as the primary AI compute requirement – driven by agentic AI systems with ~35 quintillion tokens monthly. This shifts competitive focus away from training clusters toward real-time inference infrastructure. For chip makers and cloud providers, this means a fundamental realignment of their product roadmaps.

Current forecasts predict that 40% of all enterprise applications will integrate native AI agents by end of 2026 – yet 93% of companies expect measurable ROI only by 2028. The biggest bottlenecks: weak internal processes and missing governance structures. The result is a dangerous gap between deployment pace and organizational maturity.

Prediction markets rate Alibaba with 95% probability as the leading Chinese AI lab – while Z.ai gained 29.6 percentage points over the month and is considered an emerging challenger. In global comparison, Google leads with 74% ahead of Meta as second-best AI lab behind Anthropic. The data shows increasing consolidation around a few dominant players in each region.