AI This Week
AI audio startup ElevenLabs has launched a new marketplace for iconic voices. The platform allows companies to license AI-replicated voices of famous figures, both living and dead, for advertisements and other content. ElevenLabs presents this as a solution to ethical concerns surrounding AI voice cloning, stating it operates on consent, transparency, and fair compensation for performers or their estates. The initial list of 28 voices is striking. It includes living stars like Michael Caine, Matthew McConaughey, and Art Garfunkel alongside historical figures such as Maya Angelou, Alan Turing, and J. Robert Oppenheimer. Some voices are cloned directly, while others are synthetically recreated from archival audio.
Anthropic, a key rival to OpenAI, appears on a faster path to profitability despite generating less than half the revenue. OpenAI recently saw its annualized revenue soar to $3.4 billion. Anthropic’s has also exploded, hitting $1.6 billion. Yet the company's different business model might give it an edge in reaching profitability first. The situation highlights the industry's central struggle: balancing staggering growth with the immense costs of computing power and talent needed to build advanced AI. This financial showdown stresses the high-stakes bet on sustainable dominance.
OpenAI is releasing GPT-5.1. This upgrade aims to make its flagship model smarter and more enjoyable after the previous version failed to impress. The original GPT-5 launch was so underwhelming that it prompted partner Microsoft to explore models from competitor Anthropic. This new version directly addresses user experience. It introduces two specialized models, Instant and Thinking, designed for different tasks. OpenAI is also expanding its personality presets to include options like Quirky, Nerdy, and Cynical, moving beyond a "one-size-fits-all" approach. The move signals a clear response to recent criticism and a renewed focus on what users want from their AI.
SoftBank has sold its entire stake in Nvidia. The massive sell-off generated $5.83 billion. The Japanese firm is directing that capital toward its “all in” bet on ChatGPT maker OpenAI, planning a $22.5 billion investment that will increase its ownership from 4% to 11%. This is not the first time SoftBank has cashed out of the chipmaker, having sold a $4 billion stake back in 2019. This latest move for “asset monetization” helped its Vision Fund post a blowout $19 billion gain. A source close to the matter stated the decision was not related to concerns over AI valuations but was a strategic play for capital.
Salesforce is acquiring Spindle AI. The deal injects the analytics startup’s forecasting technology directly into Salesforce’s Agentforce platform. Spindle AI specializes in agentic AI, building systems that model complex business scenarios and predict outcomes. This is a major step in Salesforce's strategy to create an "agentic enterprise." The company envisions a future where specialized AI agents collaborate to manage autonomous, end-to-end operations for customer experience. Spindle's team will work on making these agents smarter and more reliable through advanced observability and self-improvement functions. This purchase builds on other recent acquisitions, showing Salesforce's intense focus on owning every layer of the multi-agent AI stack.
A new legal battle is erupting over AI. Amazon has sued Perplexity, demanding the AI startup stop its Comet browser from shopping on its site. The lawsuit alleges Perplexity’s AI agent accesses customer accounts without consent, committing computer fraud and degrading the shopping experience Amazon built. Perplexity responded sharply, accusing Amazon of bullying and trying to stifle innovation that makes shopping easier for customers. This conflict exposes the high-stakes tension between established tech giants and the new wave of autonomous AI agents that act on behalf of users. As this AI browser war escalates, some competitors are embracing similar technology.
Apple is preparing a massive upgrade for Siri. The company will reportedly use a custom version of Google’s Gemini AI to drive its new Apple Intelligence system, according to a new report. This partnership will cost Apple around $1 billion annually and gives it access to a powerful 1.2 trillion parameter model for generating summaries and performing planning tasks. The model will run on Apple's Private Cloud Compute servers, a key detail for privacy. The move comes after Apple delayed its own AI rollout and considered other partners. While Apple continues developing its own technology, this deal highlights the intense competition in the AI space. The new Siri is expected next spring.
Google is studying “Project Suncatcher,” an initiative to move data center workloads off Earth and into orbit. The concept taps abundant solar power and the cold of space to run and cool AI-scale compute. It also promises new capacity as demand for training and inference surges. The plan raises hard questions: launch cost, serviceability, orbital debris, latency, and regulation. It draws clear comparisons to earlier radical infrastructure bets, such as underwater server farms and satellite networks. The idea signals how far cloud providers may go to meet AI’s hunger for energy and chips.
Nvidia and Deutsche Telekom are launching a one-billion-euro industrial artificial intelligence hub in Germany, a significant move in the global AI race. The project represents Europe's latest bid to make up lost ground against the United States and China. Based in Munich and set to go live in early 2026, the center will offer companies access to powerful AI for processes from design to robotics, all operating on secure IT infrastructure. A key focus is "data sovereignty," keeping sensitive industrial data secure on the continent. German industrial giants SAP and Siemens are also partners, aiming to enhance their own AI capabilities and client services.
In a stunning breakthrough, artificial intelligence has uncovered 303 new geoglyphs hidden in southern Peru's desert, doubling the number of known Nazca Lines. An international team from Japan's Yamagata University and IBM accomplished the feat in just six months, completing work that once took years in a fraction of the time. The ancient figures, dating from 200 B.C. to 650 A.D., depict felines, fish, and human forms. The AI system identified the patterns by analyzing vast amounts of satellite and drone imagery, spotting lines nearly invisible to the human eye. Archaeologists later confirmed each finding on the ground. This technology is accelerating archaeological discovery worldwide.