SAP Drops $1.16B on an 18-Month-Old AI Lab and the Week Everyone Raised Money

SAP bets a billion on German AI — before the lab turns two

SAP announced it's acquiring Prior Labs, an 18-month-old German AI startup, and will pour €1 billion (roughly $1.16 billion) into the company over the next four years. The plan is to turn it into an AI lab focused on structured data — the tables, schemas, and databases where enterprise information actually lives.

SAP didn't disclose the acquisition price, which tells you they either paid a lot or want you to think they did. Either way, committing over a billion to a company that's barely walking yet is a loud signal: the enterprise AI race is now about owning the full stack, not just plugging in a third-party model and hoping it understands your ERP system.

Prior Labs will also get to work with Nvidia's NemoClaw, a next-gen agent framework SAP is now allowing select customers to use. The move feels like SAP hedging: buy the lab, fund the research, but keep one foot in Nvidia's ecosystem in case your own bet doesn't pay off.

Structured data is where the real money sits in enterprise AI. Unstructured text is flashy, but if you can't make sense of your own inventory system or supply chain database, you're still running on guesswork. SAP is betting that owning the AI layer for that kind of data is worth more than licensing someone else's foundation model.

Seed rounds everywhere — law, e-commerce, biotech, agents

This was a funding week. Not the kind where one megadeal dominates the headlines, but the kind where a dozen companies quietly closed rounds and started hiring.

Moritz, an AI-native law firm, raised $9 million in four days — one week before graduating from Y Combinator. The round came from 20VC, Urban Innovation Fund, and a "party" of individual investors including Reddit, Instacart, Dropbox, and Runway founders. The pitch: legal work is still too manual, and agents can draft contracts, review filings, and handle discovery faster than junior associates. Moritz is building the firm around that assumption from day one instead of bolting AI onto an existing practice.

ReFiBuy, a Raleigh-based e-commerce startup, raised $13.6 million to help online retailers adapt to AI-driven shopping trends. The founder is Scot Wingo, a known name in the Triangle startup scene, so the round probably closed on reputation and a slide deck. The product thesis is that e-commerce sites need to rethink discovery, checkout, and fulfillment now that shoppers expect ChatGPT-style interfaces and agents placing orders on their behalf.

Pumpkinseed closed a $20 million Series A for protein sequencing. Executive chairwoman Jen Dionne told Axios the platform is built to handle the scale and complexity drug developers need when they're trying to understand how proteins fold, bind, and fail. Sequencing is one of those unsexy infrastructure plays that quietly underpins the entire biotech stack, and Pumpkinseed is betting there's room for a new player if the tech is fast and accurate enough.

Nace.AI raised $21.5 million from Walden Catalyst to build custom AI models for enterprises. The pitch is straightforward: off-the-shelf agents don't understand your business, so you need a platform that lets you fine-tune models on your own data, workflows, and compliance requirements. It's the same problem every enterprise AI vendor is trying to solve, but the fact that Walden wrote a $21.5 million seed check suggests Nace has either a killer demo or a roster of early customers who actually deployed it.

Altara raised $7 million from Greylock to solve the data fragmentation problem in physical sciences. If you're a materials scientist or a hardware engineer, your technical documentation is spread across PDFs, internal wikis, vendor spec sheets, and someone's Google Drive. Altara is building an AI layer to pull it all into one place so you can actually search and reference it. The round included Jeff Dean, which means Google's senior AI leadership thinks the problem is real.

Crypto made one big move this week

MoonPay acquired DFlow for $100 million in stock. DFlow is a Solana trading infrastructure platform — basically middleware that makes it easier to build trading apps on Solana's blockchain. MoonPay paid in equity, which tells you they either don't have the cash or they want DFlow's team locked in for the long haul.

DFlow raised $7.5 million total before the acquisition, so the exit multiple is decent but not wild. The real story is that crypto infrastructure M&A is back. After two years of contraction, the companies that survived are now consolidating to own more of the stack.

a16z also raised a $2.2 billion crypto fund, which sounds big until you remember they raised $4.5 billion in 2022. Crypto is cooling, but the firms that stayed in are still deploying capital — just slower and pickier than before.