👋 Hi, it’s Rohit Malhotra and welcome to the FREE edition of Partner Growth Newsletter, my weekly newsletter doing deep dives into the fastest-growing startups and S1 briefs. Subscribe to join readers who get Partner Growth delivered to their inbox every Wednesday morning.

If you’re new, not yet a subscriber, or just plain missed it, here are some of our recent editions.

Wispr Flow turns your voice into clean, professional text you can send the moment you stop talking. Not rough transcription you have to clean up. Actual polished text — ready for email, Slack, or any app.

Speak the way you think. Go on tangents. Change your mind mid-sentence. Flow strips the filler, fixes the grammar, and gives you text that reads like you spent five minutes writing it.

89% of messages sent with zero edits. Millions of professionals use Flow daily, including teams at OpenAI, Vercel, and Clay. Works on Mac, Windows, and iPhone.

Start flowing free

_{Interested in sponsoring these emails? See our partnership options here.}

Subscribe to the Life Self Mastery podcast, which guides you on getting funding and allowing your business to grow rocketship. 

Previous guests include Guy Kawasaki, Brad Feld, James Clear, Nick Huber, Shu Nyatta and 350+ incredible guests.

SAP's twin acquisitions on May 5 are not an incumbent panic buying its way into the AI era. They are a structural bet from a CTO who watched the market price SAP as a SaaSpocalypse casualty and decided he could own the entire stack enterprise agentic AI requires before Salesforce, Snowflake or Databricks figured out they were racing.

On the surface: Germany's most valuable company committing €1 billion over four years to an 18 month old Freiburg lab, three founders walking away with well over half a billion in cash up front, paired with a second undisclosed deal for a US lakehouse most enterprise buyers cannot pronounce. A spectacle. European incumbent flailing. The kind of headline filed under "too late to matter" and forgotten by the next earnings call.

Look closer, and the architecture gets sharper.

In February 2025, OpenAI's own COO conceded that "we have not yet really seen AI penetrate enterprise business processes." Fifteen months later, SAP CTO Philipp Herzig sharpened the diagnosis. "Enterprise AI doesn't stall because the models aren't good enough; it stalls because the data isn't ready for AI agents." That is the actual bottleneck. Pilots that cannot scale. Months of engineering work to integrate every new source. Duplicated pipelines. Compliance risk when nobody can explain how the AI reached a decision. The models work. The data underneath them does not.

SAP runs that data. Accounting, HR, procurement, expense management for a meaningful slice of the global economy, all sitting in tables. Structured rows and columns with business context baked in. The exact shape of data large language models were never designed to reason about, and the exact shape of data enterprises actually run on. As CFO Dominik Asam framed it to CNBC in January, "it's all about how quickly we as SAP actually also embark on these technologies in our R&D portfolio to keep the relative economies of scale advantage." Incumbents who move slow on agentic AI become rounding errors. Incumbents who move fast keep the keys to the enterprise.

The bet on how to move is where it gets interesting. Salesforce went open with Headless 360, letting customers plug in any agent including OpenAI's. SAP went the other direction. Its API policy now prohibits AI agents from accessing SAP products unless they run on SAP endorsed architectures. Joule Agents. Nvidia's NemoClaw on the Joule stack. Walled garden by design. Two incumbents under identical pressure made opposite bets on what enterprise customers want.

A walled garden only wins if what is inside the walls is genuinely better. Which is what May 5 was actually about.

Dremio for the foundation. An Apache Iceberg native lakehouse that unifies SAP and non SAP data without movement or format conversion, serverless economics, open catalog built on Apache Polaris. The data layer agentic AI needs before it can do anything useful at enterprise scale.

Prior Labs for the reasoning. A Freiburg startup whose TabPFN tabular foundation models have crossed three million open source downloads, with founder Frank Hutter pitching it as "a globally leading frontier AI lab for structured data, in Europe, in the open." Balderton partner James Wise, who led Prior Labs' $9.3 million pre seed in February 2025, called it "one of Germany's biggest ever venture outcomes."

Two deals. One architecture. A €1 billion bet on an 18 month old lab is not an incumbent panic buy. It is a wager on architecture asymmetry.

OpenAI's COO conceded in February 2025 that AI had not really penetrated enterprise business processes. Fifteen months later, SAP CTO Philipp Herzig sharpened the diagnosis. Enterprise AI does not stall because the models are not good enough. It stalls because the data is not ready.

That is the actual bottleneck. Pilots that cannot scale. Months of engineering to integrate every new source. Duplicated pipelines. Compliance risk when nobody can explain how the AI reached a decision. The models work. The data underneath them does not.

SAP runs that data. Accounting, HR, procurement and expense management for a meaningful slice of the global economy, all sitting in tables. The exact shape of data large language models were never designed to reason about, and the exact shape of data enterprises actually run on. CFO Dominik Asam framed the urgency to CNBC in January. "It is all about how quickly we as SAP actually also embark on these technologies in our R&D portfolio to keep the relative economies of scale advantage." Incumbents who move slow on agentic AI become rounding errors.

The bet on how to move is where two incumbents split. Salesforce shipped Headless 360 and let customers plug in any agent they want, including OpenAI's. SAP went the other direction. Its current API policy prohibits AI agents from accessing SAP products unless they run on SAP endorsed architectures. Joule Agents. Nvidia's NemoClaw on the Joule stack. The Information first reported the lockout in practice. Two incumbents under identical SaaSpocalypse pressure made opposite bets on what enterprise customers actually want.

Two layers have to work for agentic AI to function at enterprise scale. A unified data foundation where SAP and non SAP data coexist without movement or format conversion. And a reasoning layer built natively for tables, not a chatbot retrofitted onto a database.

On May 5, SAP announced both halves on the same day.

Two deals, announced inside the same press cycle, structured to lock in both halves of the architecture before competitors registered what was happening.

Prior Labs first. €1 billion committed across four years into the 18 month old Freiburg lab founded by Frank Hutter, Noah Hollmann and Sauraj Gambhir. SAP declined to disclose the acquisition price. Sources told Pathfounders it was an almost all cash deal with well over half a billion up front to the founders. That detail matters more than it reads. Half a billion plus distributed across three technical co founders, 15 months after a $9.3 million pre seed led by Balderton, implies a cap table that stayed unusually clean. No bridge rounds. No down round dilution. Founders walked away with the kind of personal liquidity that normally requires a decade and a public listing.

The €1 billion is not the purchase price. It is a four year commitment to grow the lab, which means most of that capital is structured as forward investment against milestones SAP has not publicly disclosed. Research output, TabPFN benchmark improvements, integration into SAP AI Core and Business Data Cloud, or revenue contribution through the Joule agentic layer are the candidates. The structure preserves research velocity, which was the founders' non negotiable, and gives SAP a productization path it could not have built internally. TabPFN stays open source. Three million plus downloads of developer mindshare comes with the deal. Balderton partner James Wise called the exit one of Germany's biggest ever venture outcomes. Closing Q2 to Q3 2026, pending regulatory approval.

Dremio second. The US Apache Iceberg native lakehouse becomes the open data foundation of SAP Business Data Cloud. Apache Polaris as universal catalog. Apache Arrow underneath. Federated reach across SAP and non SAP data with no movement or format conversion required. Serverless economics. SAP committed in the announcement to keep investing in all three open source projects, which is the only way the Iceberg bet stays credible to customers. Closing Q3 2026.

Three numbers tell the story.

One. Prior Labs raised $9.3 million in February 2025. Fifteen months later, SAP committed €1 billion against it. That is roughly a 100x markup in 15 months on a company with no disclosed revenue. The price is not for the ARR. It is for category scarcity. Tabular foundation models are a field with maybe five credible labs globally. Fundamental raised a $255 million Series A in February. Neuralk AI raised less. SAP just removed the most distribution ready of them from the market and set the comp at a billion euros.

Two. The data integration math. Enterprise AI projects today spend 60 to 80% of total budget on data integration, ETL pipelines and format conversion before a model sees a single row. Iceberg native plus federated catalog collapses that toward zero for SAP Business Data Cloud customers. The savings do not show up in SAP's P&L. They show up in customer retention, in faster time to value on Joule Agents, and in deals that close instead of stalling in proof of concept.

Three. TabPFN at three million open source downloads. That is the developer mindshare SAP could not have bought with marketing dollars and could not have built in 18 months internally. Reasoning layer and distribution, acquired in a single transaction.

The unusual part is not the price. It is the architecture of the deal itself.

Prior Labs does not get absorbed. It operates as an independent unit with €1 billion committed across four years, research velocity preserved, TabPFN staying open source. SAP gets a direct path to productization through SAP AI Core, SAP Business Data Cloud and the Joule agentic layer, without the integration drag that kills most acqui hires of frontier labs. The founders keep building. SAP keeps shipping. The capital is structured as runway, not a one time outlay.

Dremio plugs in differently. It becomes the open data foundation of SAP Business Data Cloud directly. Apache Iceberg as the native table format. Apache Polaris as the universal catalog. Apache Arrow underneath. SAP committed in the announcement to continue investing in all three open source projects, which is the only way the Iceberg bet works. Lock customers into a proprietary lakehouse and you become Snowflake. Stay open and you become the default.

Stitched together: a walled garden at the agent layer through Joule, an open garden at the data and reasoning layers underneath. Customers get portability where it matters to them, SAP gets control where it matters to SAP. That is the platform play hiding inside two acquisitions.

The deal works because both sides solve a problem they could not solve alone.

Hutter, Hollmann and Gambhir get what every frontier lab founder wants and almost none secure. Half a billion plus in cash up front. Sovereignty as an independent unit. Compute and capital runway for four years. Distribution into every Fortune 500 SAP already sells to. The alternative was raising a Series A at a markup, burning two years on enterprise GTM they have never built, and hoping OpenAI or Google did not ship a tabular model first.

Herzig gets what no internal R&D budget could buy in 18 months. A credible frontier AI story before the next earnings call. Three million developer downloads of mindshare. A reasoning layer purpose built for the data SAP already owns. And the Iceberg foundation underneath it that turns SAP Business Data Cloud from a product into a platform.

Mutual asymmetry. Both sides walked away with the scarce resource the other had.

SAP is wagering on four things being true at once.

One. Structured enterprise data is the next decade's defensible moat. Language models commoditize. Tabular reasoning over governed enterprise data does not, because the data itself is the moat and SAP already owns the customers who own it.

Two. Open formats win the lakehouse war. Iceberg becomes the default table format the way Linux became the default operating system. Snowflake's proprietary moat erodes. Customers consolidate on whoever offers the best Iceberg native experience, and SAP makes itself that default for the enterprises it already runs.

Three. Walled garden agents beat open agents inside the regulated enterprise. Salesforce is betting the opposite with Headless 360. SAP is betting that compliance, audit and data sovereignty pull customers toward curated Joule plus NemoClaw architectures, not bring your own agent.

Four. European AI sovereignty becomes a procurement filter. Buyers in Germany, France, the Nordics and the GCC start to weight EU built and EU governed AI in tenders. Prior Labs in Freiburg plus SAP plus Aleph Alpha plus Cohere plus Mistral starts to look like a stack, not a slogan.

Snowflake loses first. Their proprietary format thesis got materially harder the moment SAP went Iceberg native and committed to keep investing in Apache Polaris and Arrow. Every SAP customer evaluating a lakehouse migration now has a credible open alternative sitting inside the ERP they already run. The Snowflake sales motion of "we are the data cloud" meets a counter motion of "we are the data your business actually runs on, on open formats, with the reasoning layer built in."

Databricks loses second. They have the lakehouse, they have the AI ambition, they do not have a tabular foundation model lab or an ERP customer base. SAP just stapled both together.

Salesforce loses third. The open agent bet through Headless 360 looks weaker if SAP's curated Joule architecture ships better outcomes for regulated enterprise buyers. Two opposite bets, one quarter apart, and procurement teams will pick a winner.

Aleph Alpha and Cohere, mid merger, now have a much larger European AI competitor sitting directly upstream of their enterprise distribution.

And every tabular AI startup behind Prior Labs just had their exit comp set at a billion euros for 18 months of work. Great for them. Brutal for the next round of VCs marking entries.

Three threads worth pulling.

One. Europe is finally building an enterprise AI stack rather than renting one from US hyperscalers. Prior Labs in Freiburg. Aleph Alpha and Cohere merging into a single European AI powerhouse. Mistral in Paris. SAP in Walldorf committing real capital instead of polite press releases. For the first time, a sovereign procurement filter for EU and GCC buyers has a credible answer behind it. That changes what gets bought in Frankfurt, Paris, Riyadh and Abu Dhabi over the next 24 months.

Two. Tabular foundation models went from invisible category to billion euro exit in 18 months. Fundamental, Neuralk AI and the next wave of structured data labs just had their futures repriced. Some get acquired at premium. Some get crushed by Prior Labs running on SAP distribution. The window between obscure research and incumbent acquisition is now measurable in quarters, not years.

Three. The SaaSpocalypse response is no longer "build internally and hope." It is buy the reasoning layer and the data layer in the same quarter, before competitors register the move. SAP just wrote the playbook other incumbents will copy. Watch Oracle, Workday and ServiceNow next.

The bear case is real. Frontier labs run independently inside incumbents have a long history of being suffocated by the parent within three years. Iceberg is open, which means Snowflake and Databricks can match the format and compete on execution. OpenAI or Anthropic could ship a tabular model that erases the Prior Labs moat in a single release. Walled garden agent strategies look smart until customers decide they want choice more than they want curation.

The bull case is sharper. SAP is the only incumbent with the structured data, the customer base, the reasoning layer and the open foundation in one stack. The €1 billion is not the bet. The bet is that Herzig and Hutter ship faster than Salesforce, Snowflake and Databricks can respond. Speed of execution from here decides the decade.

What to watch by year end. Joule Agents general availability. TabPFN benchmarks against the OpenAI and Anthropic equivalents that have not shipped yet. SAP Business Data Cloud customer logos converting off Snowflake.

Here is my interview with Aly Madhavji, Managing Partner at Blockchain Founders Fund, a Singapore-based VC investing in early-stage Web3 startups globally. A Schwarzman Scholar and INSEAD MBA, he consults for the United Nations on blockchain for financial inclusion and is a judge on CryptoKnights, the upcoming Shark Tank-style show for crypto founders.

If you enjoyed our analysis, we’d very much appreciate you sharing with a friend.

Here are the options I have for us to work together. If any of them are interesting to you - hit me up!

And that’s it from me. See you next week.

What do you think about my bi-weekly Newsletter? Love it | Okay-ish | Stop it