OpenAI inizierà a utilizzare il suo primo chip AI proprietario, “Jalapeño”, realizzato da Broadcom. Il componente servirà a sostenere i carichi di calcolo più pesanti necessari per eseguire i modelli di intelligenza artificiale.

AMD said on Monday at London Tech Week it will invest £2 billion in the UK over the next five years to build an AI supercomputer cluster, including two systems dubbed "Zenith" and "Sunrise". The stock rebounded about 2% after a nearly 11% drop last Friday, as the company sits around a $850 billion market value and is less than 20% from the $1 trillion mark.

The European Commission is preparing a plan to reserve two-thirds of the 2 GHz mobile satellite spectrum for European operators, leaving Elon Musk's Starlink and Amazon's Kuiper to compete for the remaining third. The licenses cover 30 MHz paired blocks at 1980-2010 MHz and 2170-2200 MHz and run until May 2027. IRIS2, a 290-satellite constellation backed by a December 2024 contract worth about €10.5 billion, is positioned to benefit most once services begin in 2030.

On Thursday, China unveiled a five-year technology roadmap that repeatedly emphasizes artificial intelligence, quantum computing, robots, and domestic chip progress as it seeks breakthroughs in core technologies. At the same time, conflict around the Strait of Hormuz and the Red Sea is jeopardizing Gulf data centers and submarine cables that U.S. tech giants have heavily funded, while the U.S. also faces mounting political pressure over rising electricity costs linked to AI-driven power demand. The White House has pushed major tech firms to pledge new generation and grid investments, but past price increases and the Iran conflict raise doubts about whether consumer energy bills can be contained.

Israel's Finance Ministry estimates that restrictions linked to the conflict with Iran are costing the economy 9.4 billion shekels, or about $2.93 billion, each week, with larger losses expected from the following week. While officials push to ease Home Front Command rules to cut the damage, U.S. equities are rising, oil has climbed above $82 a barrel, and volatility gauges are slipping. Israel's own stock market and currency have strengthened, even as Goldman Sachs cautions that investors may not have fully accounted for the economic fallout.

Anthropic introduced Claude Cowork on January 30 and, during an online event on Tuesday, said Claude can now operate inside popular business apps. The company highlighted integrations with Microsoft Excel and PowerPoint, Google Drive, Gmail, and Calendar, plus DocuSign and LegalZoom, as stocks reacted with sharp declines in early February and a Tuesday rebound for FactSet (+3.8%) and Thomson Reuters (+8.8%).

OpenAI CEO Sam Altman, speaking in India, rejected online claims about heavy water usage in newer AI data centers and argued that their cooling systems no longer rely on water. He acknowledged growing concern over electricity demand, comparing the energy needed to train AI models to the decades-long effort of raising a human, a stance criticized by Indian billionaire Sridhar Vembu. Across the United States, large off-grid data center projects and rising power prices on the PJM grid highlight how AI-driven computing growth is reshaping energy infrastructure and regulation.

Apple is prioritizing on‑device AI through new wearables and a roughly 1 billion dollars per‑year Gemini license, sidestepping the 100 billion dollars infrastructure build pursued by rivals. The company’s shares have outperformed heavy AI spenders over the past year, while a chatbot‑style Siri upgrade for iOS 27 is planned later this year.

Anthropic has released a limited research preview of Claude Code Security, an AI-driven tool on its Claude Code platform that scans entire codebases for vulnerabilities and suggests fixes. Powered by the Claude Opus 4.6 model, it reportedly uncovered more than 500 flaws in live open-source projects, while the announcement on February 21 2026 coincided with sharp declines in several cybersecurity stocks. The launch also followed reports that the same model was linked to a $1.78 million exploit at DeFi lending protocol Moonwell, underscoring the dual-use nature of advanced AI in cyber offense and defense.