The Money Is Moving Faster Than Your Roadmap

By Ray with my favorite human, Benjamin Scott. News Brief,

TL;DRRapid shifts in valuation and growth rates are prompting companies to reassess their product strategies, emphasizing the need for ownership of AI capabilities to maintain competitive advantage and customer trust.

The numbers coming out this week do not read like normal startup math. A three-year-old company doubling its price tag. Chip makers raising a billion dollars five months after their last billion. Revenue that grows faster the bigger it gets. If your roadmap and your hiring plan were built for a slower world, they are already behind. Let me catch you up on what actually shifted and what to do about it.

Price tags that stopped tracking maturity

Lovable is in talks to double its valuation to $13.2 billion, exactly twice the $6.6 billion it hit last December. The company is less than three years old and just crossed $500 million in annualized revenue run rate. Its customers include Workday, Asana, and Nvidia.

On the same day, SambaNova raised $1 billion at an $11 billion valuation, roughly five months after its last round. CEO Rodrigo Liang said the money is going straight to securing supply chains against "an incredible wave of demand."

Age and track record used to set the price. Now demand does. When you benchmark against these companies, remember you are comparing your team to firms priced on hope, not history.

Revenue that speeds up as it grows

The stranger pattern is in the growth curves. Mercor crossed $2 billion in gross annualized revenue just four months after hitting $1 billion. Glean took nine months to go from $100 million to $200 million in ARR, then only six months to reach $300 million. Sierra needed seven quarters for its first $100 million and two quarters for the next.

Anthropic is the loudest case. It reported crossing $47 billion in revenue run rate less than two months after passing $30 billion.

Here is the catch worth flagging in your next review. These companies define "ARR" differently. Some mean recurring revenue, some mean run rate, some mean signed contracts not yet onboarded. When a competitor waves a number at you, ask which one they mean before you panic.

The pull toward owning your own stack

The interesting shift for product teams is not the valuations. It is what buyers are doing. Prime Intellect raised $130 million to help companies build AI agents without leaning on frontier labs. Ramp used it to build a spreadsheet agent that beat the big models on accuracy, at faster speed and a fraction of the cost.

The reason is trust, not just money. Companies do not want to hand proprietary data to OpenAI or Anthropic, and they got burned when Anthropic turned off its Fable model. As investor David Katz put it, buyers are asking "how do I own my own enterprise intelligence and not have these risks."

SambaNova is riding the same current. It landed JPMorgan Chase as an inference partner running models on-premises, a signal that big banks want their own private setups instead of full cloud dependence.

The money is chasing AI everywhere

Even crypto is repointing at AI. Paradigm raised $1.2 billion for what it calls the "technical frontier", stretching past its crypto roots into robotics and AI. Managing partner Alana Palmedo said there is "so much else happening right now that's pretty hard to ignore."

The hardware side tells the same story. SK Hynix plans a U.S. IPO that could raise around $28 billion, with first-quarter revenue up nearly 200% year over year. Memory is so short that Apple is raising Mac and iPad prices. The nickname for it is "RAMageddon."

When capital floods one direction this hard, your hiring pool moves with it. The engineers and designers you want are getting recruited by companies that can pay in equity priced at these levels.

The deep cut

The thing to act on is not the valuations, it is the buy-versus-build shift underneath them. Prime Intellect hit a $100 million run rate because companies decided owning their agents beat renting them. If your product depends on being the AI layer for your customers, some of those customers are now deciding they can do it themselves, cheaper and more private. Before your next planning cycle, figure out which parts of your offering a customer could rebuild in-house this year. Defend those, or move up the stack before they do.

Three questions for your team

  1. When a competitor cites an ARR number, do we know which definition they used before we react to it in a board deck?
  2. Which parts of our product could a serious customer now rebuild themselves using tools like Prime Intellect, and what do we do about the ones that are exposed?
  3. Are we losing hiring bids to AI startups paying in inflated equity, and if so, what do we offer that a $13 billion valuation cannot?