Top Tech News Today, August 25, 2026: Apple, Cisco, Infineon, OpenAI, Meta, Nvidia, SpaceX & More

It’s Tuesday, August 25, 2026, and the tech world just hit another acceleration point. Overnight, Apple quietly dropped a new Mac mini built for the on-device AI era, Nvidia pushed its specialized inference silicon into full production to make agents feel instantaneous, and a $150 million raise turned data centers into flexible grid partners. At the same time, Taiwanese prosecutors charged insiders over illegal AI server shipments to China, OpenAI shut down a Russian influence operation running through ChatGPT, and Xiaomi showed off its latest homegrown 3nm chip. Behind those headlines, the AI buildout itself is entering a more consequential phase. Nvidia is preparing for another crucial earnings test as questions grow around the economics of AI infrastructure, robotics startups raise hundreds of millions of dollars, Cisco pushes deeper into AI data centers, and Xpeng turns its EV expertise toward humanoid robots. Apple, meanwhile, is preparing for its first CEO transition in 15 years, while OpenAI is betting that AI agents will become part of everyday office work and governments from South Korea to the U.S. confront the economic, regulatory, and security consequences of increasingly autonomous technology. From Silicon Valley boardrooms and Asian fabs to neocloud funding rounds and physics-first models, here are the top tech news stories defining the day — the breakthroughs, power plays, and pressure points that matter most to founders, builders, and the global tech ecosystem. Technology News Today Nvidia’s Next AI Growth Test Arrives as Wall Street Questions the Economics Behind the Boom Nvidia heads into its second-quarter earnings report Wednesday with expectations that would have seemed extraordinary even a year ago. Analysts surveyed by LSEG expect revenue to nearly double from a year earlier to roughly $92.18 billion, driven largely by data-center sales projected to more than double. The company is also expected to guide toward third-quarter revenue of about $104.2 billion, an 82.8% year-over-year increase. Nvidia’s next-generation Vera Rubin processors are scheduled to begin shipping this autumn, putting the new architecture at the center of the company’s next growth phase. But investor attention is shifting beyond raw chip demand. Nvidia has become increasingly involved in financing the infrastructure that ultimately buys its hardware. The company recently helped arrange hundreds of billions of dollars in financing for AI customers and agreed to guarantee as much as $105 billion supporting OpenAI’s Ohio data-center lease. Critics worry these arrangements could blur the line between underlying demand and vendor-supported expansion. CEO Jensen Huang argues that Nvidia is deploying its enormous cash resources to accelerate infrastructure customers would eventually need anyway. The earnings call will therefore be about more than revenue. It will test whether the economics supporting the AI buildout remain convincing as spending moves into unprecedented territory. Why It Matters: Nvidia’s numbers have become one of the clearest indicators of whether the global AI infrastructure boom remains economically sustainable. Source: Reuters. Apple Unveils New Mac Mini with M6 Chip Ahead of iPhone Event Apple announced an updated Mac mini featuring its first 2nm M6 processor in the base model and an M5 Pro option in the higher tier, marking the first refresh of the compact desktop in nearly two years. Pre-orders opened today with shipping set for September 22; the M6 configuration starts at $899 with 16GB unified memory and 256GB storage, while the M5 Pro starts at $1,699. The M6 delivers a 12-core CPU, 12-core GPU, and dual 16-core Neural Engine optimized for on-device AI workloads, with memory bandwidth rising to 170GBps. The design retains the tiny chassis introduced in 2024 but adds faster Thunderbolt 5 ports and improved Ethernet. Demand for the prior M4 models had surged as users ran local AI agents, and memory shortages had constrained supply. This positions the Mac mini as a key local AI compute device just as Tim Cook prepares to step down, strengthening Apple’s hardware edge in the agentic AI era against cloud-heavy rivals. Why It Matters: The refresh accelerates affordable on-device AI for developers and creators while testing demand ahead of major iPhone launches. Source: The Verge. SpaceX Uses Grok AI for Thousands of Starlink Customer Support Interactions SpaceX revealed that its Grok-powered AI is already handling thousands of Starlink customer calls and chats for order placement and support. The system demonstrates early production use of advanced conversational agents in a high-volume satellite internet service. Vertical integration of AI support inside space and connectivity businesses accelerates operational efficiency while testing real-world agent reliability at scale. Why It Matters: Production AI customer service in critical infrastructure validates agentic systems beyond experimental pilots. Source: PCMag. AI Robotics Startup Generalist Raises Another $200 Million Just Two Months After $400 Million Round Investor appetite for physical AI is showing little sign of cooling. Generalist, an AI robotics startup founded by alumni of Google DeepMind and Boston Dynamics, has raised another $200 million, only about two months after securing a $400 million financing round. 8VC leads the new investment, with participation from existing backers, according to Axios. Unlike companies racing to manufacture humanoid robots themselves, Generalist is focused on the intelligence layer, developing general AI models capable of controlling different robotic systems. That distinction could become increasingly important. The robotics industry is beginning to resemble the early AI model market, with one group of companies building physical machines and another developing software that can operate across multiple hardware platforms. If general-purpose robot models improve quickly enough, manufacturers may eventually buy or license intelligence rather than developing every part of their AI stack internally. The latest Generalist financing also illustrates how aggressively capital is flowing toward companies positioned around embodied AI, where machines learn to perceive environments, reason about physical tasks, and execute actions. Robotics remains technically difficult, and commercial deployment at scale is far from assured. Yet investors increasingly see robot intelligence as a potential platform market rather than a collection of narrowly programmed automation systems. Why It Matters: Generalist’s rapid fundraising shows investors are betting that the next major AI platform could control machines, not just generate text or software. Source: Axios. Accelerated Understanding Launches Physics-First AI Model Handling 5 Trillion Data Points Former Project Prometheus candidates Anima Anandkumar and Benedikt Jenik unveiled Accelerated Understanding Inc. and its neural-operator-based AI that processed 5 trillion data points in a single prompt—orders of magnitude beyond current language models. Unlike text-trained transformers, the system predicts physical phenomena across space and time for applications in robotics, extreme weather, chip design, and energy exploration. The founders rejected a language-centric view of intelligence in favor of a nature-centric physics foundation. Enterprise demand is rising for domain-specific models that can simulate real-world systems far beyond generative chat, opening new frontiers for scientific and industrial AI startups. Why It Matters: Physics-native architectures could unlock reliable simulation at scales language models cannot reach. Source: TechStartups via Reuters. Cisco Expands Nvidia AI Alliance With Supermicro Servers as It Moves Deeper Into Data Centers Cisco is pushing beyond its traditional networking stronghold and deeper into the AI hardware stack. The company is working with Supermicro to add high-density AI s