Databricks is now expected to achieve an impressive top line of $6.9 billion by the end of the first half of fiscal 2027, representing more than 80% year-over-year growth, with AI products alone achieving an annual run rate of $1.7 billion and net retention of over 140%. That gives co-founder Arsalan Tavakoli-Shiraji better insight into how companies are actually using AI budgets than almost anyone else. On stage around SaaStr AIhe laid out where the money is and isn’t going, and a claim that should change every B2B manager’s mindset about competition:
Table of Contents
Any company that has a monopoly today will no longer have a monopoly in 12 to 24 months.
It’s in some ways a stunning claim from a company that has spent 13 years building a deep moat in data and AI. But we are all seeing it now, and it has direct consequences for both pricing power and competitive risk.
The summary:
- Three forces have broken pricing power simultaneously: construction costs, the lower price segment and migrations
- Everyone maximizes their token and almost no one knows their true AI ROI
- The real bottleneck in enterprise AI is the context, not the model
- Traditional BI is basically dead
Our deep insight:
Everyone is doing token maxing. Almost no company today knows its true AI ROI.
If you live by X, you would think that many companies have caught on to AI. They have all set up their own LLM research agencies. They all automate everything. The reality at the front is not like that.
By now, every CEO of their organization has said the same thing: If we don’t use AI, we’re behind. Uses tokens. This is how we measure your performance. Employees are doing just that and token spending is increasing rapidly.
The problem is what comes immediately after. Expenses are rising and most executives have no idea what they are getting in return. Tavakoli’s framework: Everyone has reached their token maximum, spending is increasing, and there is no clear statement about the issuance. This is the actual state of enterprise AI in 2026. Not “we figured it out.” More like this: “We spend a lot and try to find the result.”
For founders, this is the beginning. The companies that tie AI spending to a clear business outcome are currently winning budget. Standing agents receive zero points. Driving one number is enough.
Data is no longer a storage decision. It became a top-line device.
A few years ago, the pitch to a CIO was straightforward: put your data in a lake, replace the warehouse, get better analytics. Databricks used to whisper the AI part because AI made buyers think of self-driving cars and robots.
This urgency profile has completely changed. AI is now a top priority and not a factor for back office efficiency. And as soon as companies commit to this, they hit the same wall: data silos, no semantic layer, no context. The hard part isn’t the model. It’s about making data clean, managed and accessible to agents rather than people.
That’s what most people miss. The bottleneck in enterprise AI is not model quality. It’s the context. And context is not the same as data.
Context is the real bottleneck, and it gets boring quickly
Consider hiring a new employee. How do you explain everything that happens in your organization so that it can actually function? This is what a broker needs, and almost no company has it written down.
Ask yourself a simple question: Show me my biggest spenders in major clouds at the end of the last fiscal quarter in EMEA. Sounds trivial. But what is considered a “cloud”? What is a “Top Donor”? When is the fiscal quarter? Which countries are in the EMEA region for this business? Each of these definitions is a definition that someone learned years ago by asking a colleague. Multiply that to an organization with 100,000 employees. These definitions are buried in emails, meeting minutes, and call notes and are constantly changing.
Most companies tried to solve this problem statically by writing a context document. The document is out of date the day after it was written. If you point someone to a context document from two years ago, it is already wrong. The hard part is not even writing the context. It constantly pulls in new information and discards the old.
That’s why Databricks developed Genie Ontology, a self-improving context layer that extracts and continuously updates business knowledge from files, tickets, chats and meetings. The bottom line for anyone building brokers for the company: The lasting value does not lie in the broker. It’s about maintaining the live context.
Traditional BI is basically dead
Standalone BI is a dashboard graveyard. A handful of dashboards with long queries that no one looks at, created by the 5% of an organization that can actually write a query, with a one-week turnaround time for each new question.
The genius at Databricks does it differently. The evidence from the session: A car manufacturer just loaded 70,000 users onto it. Not the 5% who can write SQL. The 95% who run the company and know which questions are important. They ask their own questions and get answers in 30 seconds instead of waiting a week for a data analyst.
That changes behavior. Someone has a question, asks it in the middle of the meeting, the answer comes back and the decision is postponed in real time. And no one ever has just one question. An answer leads to a follow-up, which leads to going deeper. Tavakoli found that people are actually more stressed in the AI age because usage has increased. If you can always get the next answer, keep pushing.
The old BI tools struggled because they didn’t have a semantic understanding of the data. Putting “talk to your data” above just meant converting text to SQL, which didn’t work. You need the layer that interprets the requirements and maps them to the meaning of the data. Today, every byte is expected to be visible in the organization and to every employee in real time. BI as a category disappears into dashboards and answers.
Why no monopoly survives: The mechanics
Three things are happening at once and together they are breaking down the pricing power of the incumbents.
- First, the cost of developing software has collapsed. When everything was a monolithic stack, it was extremely difficult to create one and convince a company to adopt it. Now a new entrant can join an organization that already has data collected and managed, build on that, and quickly deliver something credible. More builders, more competitors in every valuable category.
- Second, the bottom end got good. The old low end product was cheap and crappy. It may have made a workflow bad, but technically it worked. Now AI is making those same low-end products great, especially when they leverage third-party APIs. Layer Salesforce, Shopify, or Databricks data on top of a lightweight app and it’s no longer a one-workflow toy. “You get what you pay for” breaks down. A new entrant with nothing to lose sets the price at 30-40% of the incumbent and wins in greenfield deals.
- Third, and most importantly, migration costs have collapsed. Migratory birds used to die on the vine. If a provider promises you a 50% savings, then the migration itself costs five times the annual savings. Nobody moved. The person who understood the legacy system retired before three successors.
This math is now reversed. The code is self-describing, so LLMs can go into a legacy environment, understand what it does, convert it, migrate the data, and write the harnesses to verify that the new output matches the old. Databricks completes enterprise-class migrations in 30 days or less, depending on complexity. When the costs of switching fall, the willingness to test a new provider increases. And once buyers actually try to switch, no incumbent can rely on lock-in to defend its price.
The evidence was on the floor at SaaStr AI. A large proportion of the exhibiting companies did not exist a year ago and already have significant sales. Buyers are ready to migrate in ways they simply haven’t before.
Lock-in is no longer a viable long-term strategy in B2B
There is a Cambrian explosion of AI apps that works in two directions. It’s the biggest app creation moment in B2B history. It also means brutal competition.
If you attack an incumbent, the wedge is there. Lower construction costs, an actually good low-end product, and migration costs low enough for buyers to switch. Be aggressive, lead with a clear outcome, and target the categories where a monopoly has been collecting rent for a decade.
If you hold the strong position, lock-in is no longer a strategy. Either reinvent yourself with real AI and earn relevance for the next decade, or watch the power base erode from the ground up as modern, cheaper, agent-based alternatives disappear. There is no holding the line.
Vibe coding your own CRM poses no threat, and Tavakoli is clear that this is not a real path for most companies. Building it is one thing. Maintaining it, developing it and taking responsibility is another matter. The threat is a thousand modern startups built for a world where agents are the primary users of software, migrations take a month, and pricing discipline has disappeared. That’s the next 24 months.
What it means for the next 24 months:
- When you attack an incumbent, the wedge is finally there: price aggressively, lead with the result
- If you are the incumbent, lock-in is no longer a strategy. Reinvent yourself with true AI or be wiped out from scratch
- Winning an AI budget means being tied to a number, not “we deployed agents.”
- The enduring value of agents is the live context, not the agent itself
https://www.saastr.com/databricks-co-founder-arsalan-tavakoli-every-software-monopoly-falls-in-the-next-24-months/
