Microsoft

Bending Spoons, an Italian holding company, acquires struggling software firms like Airtable and Miro at steep discounts, leveraging Berkshire Hathaway’s $900M+ investment. The company, inspired by Berkshire’s long-term strategy, prioritizes operational efficiency and profitability over resale, unlike traditional private equity. Recent layoffs at acquired firms reflect its cost-cutting approach amid the AI-driven 'SaaSpocalypse,' where legacy software faces declining valuations. Revenue surged to $700M in Q2 2026, driven by acquisitions, though long-term customer retention remains critical as AI reshapes the sector.

Oct 2026 · CNBC TechNews article

ServiceNow Inc. is making one of the more consequential and visible bets in enterprise AI. The company says AI Control Tower can discover and govern any agent, work across models and clouds, track cost and value, and shut down behavior that violates policy. That sales pitch plays directly to Service

Oct 2026 · SiliconANGLENews article

Paysend CEO Ben Chisell limits AI tools to high-performing employees, requiring them to demonstrate tangible contributions before access. The company achieves **25% more engineering output** with the same team size but rejects blanket AI distribution, citing risks like magnified inefficiencies. Instead, Paysend tests tools—such as Claude—by assigning them to new hires for a month, leading to tailored solutions like a dedicated content-brief agent. Unlike competitors rolling out AI companywide, Paysend focuses on **customer problem-solving** and employee upskilling, avoiding cost-cutting automation. Research shows **77% of AI-using UK firms see no revenue change**, despite productivity gains, underscoring Chisell’s cautious approach.

Oct 2026 · CNBC TechNews article

Microsoft’s AI research group introduces **MAI-Base-2**, a 2-trillion-parameter base model that outperforms the leading open-source alternative, **DeepSeek-V4-Pro-Base**, in science, programming, and math benchmarks. The model uses a **mixture-of-experts (MoE) architecture**, activating only 80 billion parameters per prompt to improve hardware efficiency by up to 12x. Trained over 38 days using **16,000 Nvidia GB200 chips** and **50 trillion tokens**, it integrates a vision module for tasks like scientific chart analysis. Microsoft plans to license the model to **Azure partners** and use it to develop future LLMs.

Oct 2026 · SiliconANGLENews article