Two Companies Got 43 Cents of Every AI Dollar. Plan Your Budget Around That.

The AI money is not spread out. It is stacking up under a handful of names, and the way vendors price their tools is starting to bend to where the users actually are. If you own a design or product team, both of those trends hit your next budget review. Let me catch you up on what moved this week and what to do about it.

The money is crowding into a few names

The headline number is stark. AI companies took 86 cents of every venture dollar in the first half of the year, according to PitchBook-NVCA data. U.S. startups raised $412.7 billion by June, past the full-year record from 2021. That sounds like a rising tide. It is not.

Strip out OpenAI and Anthropic and the picture changes fast. Those two alone pulled in roughly 43% of all global venture capital in the first half. Seed funding fell 27% nationally, and first-time fund formation is on pace for its lowest year since 2016. The Seattle region, an AI hub thanks to Microsoft and Amazon, saw venture funding drop 40% and slid from fifth to seventh among the top ten metros.

What this means for you: the vendors with the deepest war chests will keep shipping and cutting prices. The smaller ones you like may run out of runway. Factor survival odds into any tool you build a workflow around.

Big rounds are not only a Bay Area story

The crowding is real, but new large checks are still landing outside the giants. PixVerse, a Singapore-based video generation startup, closed a $439 million round that pushed its valuation past $2 billion, with Alibaba and Mirae Asset joining in. It claims 150 million registered users and 15 million monthly actives, and charges $4.80 per minute for image-to-video.

Co-founder Jaden Xie made a point worth remembering: "OpenAI exited the business when they shut down Sora 2," he said, arguing only a few players clear the quality bar. His edge, he says, is data labeling, not the data itself, built on ByteDance visual tech.

So the field is not one company. If your team touches video or creative work, the specialist vendors are funded enough to be worth a real trial, not just a demo.

Open source is pulling in real checks too

Nous Research, the maker of the open-source Hermes agent, is finalizing a round at a $1.5 billion valuation, raising at least $75 million. That comes less than three months after its $50 million Series A. Hermes has roughly 214,000 stars and nearly 40,000 forks on GitHub, and ships with built-in skills like web search and coding.

Here is the useful part for a team lead. Nous offers both a free self-hosted version and cloud tiers from $20 to $200 a month. That split lets you test the tool for free on your own machine, then pay only when you want the hosted convenience.

Open-source tools with funding behind them give you leverage in vendor talks. You have a credible fallback, which is worth more than any discount a salesperson offers.

Pricing is going local, and that tells you where the users are

Anthropic started listing Claude prices in Indian rupees, its biggest market after the U.S. India makes up 5.8% of global Claude usage. Claude Pro shows at about $21 a month there, close to the $17 U.S. price, with local taxes baked in.

The rollout is not finished. Anthropic has not enabled UPI, India's instant payment network, so users still pay by card. OpenAI already added UPI support for ChatGPT last August. Anthropic also opened a Bengaluru office and hired former Microsoft India managing director Irina Ghose to run the business there.

If your team is global, watch these local pricing moves. They tell you which vendors are serious about your regions and which are still treating them as an afterthought.

The deep cut

The cheap lifetime AI bundles floating around, $60 for GPT, Claude, and Gemini in one login, exist because the market is crowded and undifferentiated. Multi-model resellers like ChatPlayground win when no single model is clearly best and switching is cheap. That is a signal, not just a deal. When your tools are this easy to swap, lock-in is weak and your negotiating power is strong. Do not sign a long enterprise contract at list price when a one-prompt-to-many-models workspace proves the models are close enough to trade. Use that leverage before it disappears.

Three questions for your team

1. Which of our current AI vendors could run out of money in a year, and what is our backup plan if one goes dark like Sora 2 did?
2. For the regions our users live in, which vendors price and pay locally, and which force awkward currency and payment friction?
3. Where are we paying list price for a model that a free open-source or multi-model option could match, and what would switching actually cost us?