OpenAI Pays $116 Million Per Head, Google I/O 2025: Tokens, Tiers, and Vaporware

Google loves big numbers. Most of the TGIF meetings I attended at Google featured a BIG NUMBER designed to impress us. At this year’s I/O, Sundar Pichai proudly revealed that Google now processes 480 trillion tokens per month, up from just 9.7 trillion a year ago. Normal folks would say, “WTF does this mean?” A token is essentially a small piece of text, such as a word or part of a word, that AI models like ChatGPT process to understand and generate language, so Google is processing a lot of words.

Project Mariner caught attention by promising AI-powered agents to handle mundane tasks. But don't get too excited yet; like much of the event, specifics were sparse, and the rollout is incremental, bordering on vaporware territory.

AI assistants continue to be the holy grail for tech giants. Elon Musk teased his humanoid robots capable of simple tasks like dishwashing and tidying up, mundane chores that frankly everyone wants automated, for the love of god get me a bot that washes and folds laundry. Likewise, Google showcased Gemini Live, a real-time, interactive AI assistant. Unfortunately, the demo felt basic, solving simple tasks like bike repairs, something current models like ChatGPT already handle comfortably.

One truly valuable feature demoed, however, was Gemini's real-time translation during meetings. Imagine having a seamless conversation in multiple languages without delay so someone in SF could have a conference with AIPAC and fire folks in multiple languages at once or this could help remote work, enabling greater collaboration across language barriers.

Google also teased its "Gemini and Gmail" feature, promising to automatically generate email responses that mimic your voice. Useful? Possibly. Or according to the media, creepy? To me I'm fine with it, whatever helps me generate text without sounding like a 22 year old marketing major from some Ivy league. Here’s a quick demo of me using the tool.

Google introduced new pricing tiers: Google AI Pro ($20/month) and Google AI Ultra ($250/month). The Ultra plan offers enhanced features, including the latest versions of Gemini and advanced filmmaking tools. However, the steep pricing raises questions about genuine value. Is Ultra’s 10x premium justified? While some power users might justify the cost, like if you make commercials, most users will probably settle for the Pro version or look elsewhere, especially if rival services bundle similar features more affordably.

In other news, OpenAI has officially merged with Jony Ive’s company in a $6.5 billion equity deal for 55 people which works out to $116M per head, wow. While the merger itself is significant, the announcement was notably short on details. OpenAI’s increasingly Apple-like marketing—a lot of style, little substance which provides me with opportunities to troll.

On the healthcare front, AI startup Abridge is gaining serious traction. Abridge, which transcribes medical conversations, reached a $5 billion valuation backed by solid annual recurring revenue. It's a clear example of AI solving a genuinely pressing need—reducing doctor burnout by automating note-taking and documentation. While other AI startups flounder, those tackling real-world, regulated challenges like healthcare seem best poised to succeed.

Google’s I/O 2025 had highs and lows. Real-time language translation and agents (assuming they work) are cool. Tech giants should consider less theatrics and more tangible innovation—demonstrations that clearly showcase immediate user benefits rather than vague promises.

You can check out our full conversation about this on the latest episode of the SVIC podcast. We’re giving away 1 month of access to our bonus content (Deep Dives into IPOs, M&A Activity, private interviews, reactions and more!) and Discord—no credit card, no strings. Grab it here: svicpodcast.com/gift

Jordan