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:
Google has launched a more powerful version of its Nano Banana image generation tool that is capable of producing highly realistic images and coherent text, even to the point where it can generate infographic explainers. This has obvious implications for identity security, and should raise alarms even as Google offers a new watermarking system that is meant to definitively verify whether a given image was AI-generated. In any case, Nano Banana Pro appears to have genuinely compelling reasoning capabilities under the hood.
It was launched just days after Gemini 3, Google’s latest flagship reasoning model, which attained very impressive scores on a number of benchmarks, and reaffirmed the steady rate of AI progress that has been seen over the last couple of years. Google’s recent AI progress has been so impressive that it has even somewhat unsettled rival OpenAI, whose CEO, Sam Altman, issued an internal memo acknowledging that Google may have “leapfrogged” OpenAI in some areas, and that there could be some “rough vibes” ahead.
Nevertheless, OpenAI this week showcased its GPT-5 models’ applications in scientific research, releasing a long paper, Early science acceleration experiments with GPT-5, co-written with researchers at Berkeley, Oxford, Cambridge, Vanderbilt, Lawrence Livermore and others. The paper details a number of use cases in which GPT-5 AI helped human researchers to do things like solve a long-standing Erdős number-theory problem. It’s part of the firm’s “OpenAI for Science” program, which aims to get “fully automated AI research tools” in use by around 2028.
San Francisco-based Physical Intelligence has raised about $600 million in new funding at a roughly $5.6 billion valuation, in a round led by Alphabet’s growth fund CapitalG with participation from Jeff Bezos and a long list of top-tier VCs. The startup is focused on building “brains for robots” – foundation models that can control many different robot types and learn complex tasks in messy, real-world environments. Physical Intelligence recently launched a new “vision-language-action” (VLA) AI model called π*0.6 (pi-star-zero-point-six), which is designed to learn from both human demonstrations and its own trial-and-error experience, via a reinforcement learning method the company calls Recap.
SoftBank is planning to invest up to $3 billion into a factory that will make modular data center units. The Lordstown, Ohio, facility, which previously churned out electric vehicles, will be repurposed to make pre-assembled, portable data center modules that can be shipped and installed at OpenAI sites, starting with a huge facility in Milam County, Texas, and potentially expanding to other locations in the “Stargate” network of huge AI data centers. This is part of SoftBank’s broader $18 billion commitment to the Stargate Project, a planned $500 billion AI infrastructure joint venture between OpenAI, SoftBank, Oracle and MGX.
Saudi Arabia is leading a $900 million Series C funding round for US video-generation startup Luma AI. The round is being led by HUMAIN, an AI company backed by Saudi Arabia’s Public Investment Fund (PIF), and values Luma at over $4 billion. As part of the agreement, Luma has agreed to work with HUMAIN on “Project Halo”, a planned 2-gigawatt AI supercluster in Saudi Arabia that will provide the compute backbone for Luma’s push into multimodal AGI and “world models” that simulate the physical world beyond text interaction.
Nvidia delivered its much-anticipated (feared?) quarterly report this week, and massively outperformed, with FY26 Q3 revenues up 62 percent year-over-year, at $57 billion, and net income up 65 percent, at $31.9 billion. The results arrived amid serious market volatility, and in interviews and on Nvidia’s earnings call, CEO Jensen Huang explicitly pushed back on “AI bubble” talk. “Blackwell sales are off the charts, and cloud GPUs are sold out,” he said in a press release, referring to the firm’s latest data center chip family.
The chatbot’s take: We’re not really math people here at ID Tech.
–
By Alex Perala
