Welcome to the newest edition of ID Tech’s AI Update. Here’s the latest big news on the shifting landscape of AI and identity technology:
Anthropic plans to invest about $50 billion in building data centers in the US, starting with sites in New York and Texas. The work will start via a partnership with Fluidstack, a UK-based AI infrastructure firm. It is ultimately expected to lead to the creation of 800 permanent jobs and 2,400 construction jobs during the build-out. The data centers will come online in stages, starting in 2026.
Anthropic’s Claude Code AI tool was used by hackers linked to the Chinese state in a months-long cyber-espionage campaign, the company says. The campaign allegedly targeted about 30 entities globally, including technology firms, financial institutions, chemical companies, and government organizations. It succeeded in compromising a “small number” of targets before Anthropic’s engineers discovered what was happening and shut down the operation.
Cursor has closed a $2.3 billion Series D funding round at a $29.3 billion post-money valuation. That’s almost three times its roughly $9.9 billion valuation from a $900 million round in June, and about 12 times its value in January of this year. Built by Anysphere, Cursor is an AI-powered code editor, and it also has its own code-optimized LLM called Composer. The company says it will use its new funding to further scale up Composer, train on larger proprietary datasets, and buy more compute.
Thinking Machines Lab is in talks to raise funding in a round that would value it at about $50 billion. The startup is headed by former OpenAI CTO Mira Murati, and was founded in February of this year. For several months, it wasn’t clear to outsiders what exactly Thinking Machines is aiming to do, but it now has its first product – Tinker, a training API and Python SDK that is meant to let developers fine-tune and post-train their own large language models (LLMs) without having to manage their own GPU clusters.
SoftBank Group has sold its entire Nvidia stake, unloading about 32.1 million Nvidia shares last month for about $5.8 billion. But the firm insists it isn’t making an exit from the AI market; rather, SoftBank says it needs to reallocate resources more heavily to OpenAI. CFO Yoshimitsu Goto told reporters that SoftBank’s investment in OpenAI this year is “very substantial” and that “more than USD 30 billion needs to be made” in new OpenAI-related investments. The Nvidia sale was disclosed in a November 11 regulatory filing that also showed that SoftBank’s net profit had more than doubled, thanks in large part to its AI investments.
Meta’s chief AI scientist, Yann LeCun, is reportedly preparing to depart the firm for his own startup, though nothing has been formally announced yet. Considered one of the “godfathers of deep learning,” LeCun is said to be in early fundraising talks for his new venture, which will focus on “world models” – AI systems focused on building a visual and spatial representation of the world, rather than a text-based one. LeCun joined Facebook in 2013 to found FAIR (Facebook AI Research, now Meta AI) and has remained a part-time professor at NYU.
Google DeepMind has unveiled SIMA 2, a new version of its agentic system for 3D virtual worlds. SIMA stands for Scalable Instructable Multiworld Agent, and – to be a bit reductive – it’s meant to learn to play video games. SIMA 2, powered by Google’s Gemini AI models, is doing well; it has been trained on multiple major games and is able to set goals and reason about how to execute them in those virtual environments. It can also answer questions about its behaviours, such as, “Where are you and why are you doing this?”
Sam Altman says energy is the key bottleneck to rapid AI development. In a fairly wide-ranging interview, economist Tyler Cowen asked the OpenAI CEO, “If you could have more of one thing to have more compute, what would the one thing be?” Altman responded, “Electrons,” and went on to identify natural gas as the most immediate near-term solution to the problem, and fusion energy and solar as longer-term solutions. “You’re still bullish on fusion?” Cowen asked. “Very much, and solar,” Altman said.
The chatbot’s take: If AI can shoot someone in a video game…
–
By Alex Perala
