Databricks does not need the public markets right now, and that is exactly what makes its IPO delay worth watching.
Databricks has become the kind of company Wall Street would normally chase hard: fast-growing, AI-linked, deeply embedded in enterprise data systems and valued at a level that would make most private software companies rush toward the opening bell. Instead, CEO Ali Ghodsi is telling the market to wait.
According to Bloomberg, Ghodsi said on Thursday that Databricks still expects to go public at some point so employees have a market-based way to sell shares, but he does not see 2026 as the right year. That matters because the comment lands in a year when investors have been looking for exactly this kind of listing. AI infrastructure is still one of the few technology themes with enough heat to carry a large IPO, yet one of the best-known private companies in the category is choosing to remain private.
This is not a small, cash-starved startup trying to avoid scrutiny. Databricks said in February that it had crossed a $5.4 billion revenue run-rate, growing more than 65% year over year. More than $1.4 billion of that came from AI products. The company also said it had delivered positive free cash flow over the previous 12 months, sustained net retention above 140%, and had more than 800 customers spending at least $1 million annually.
Those numbers usually create pressure to list. In Databricks' case, they may have done the opposite. The company has already raised enough money to behave like a public software giant without accepting the quarterly discipline of being one.
The old reason for an IPO was simple. A company needed capital, credibility and liquidity. Databricks already has plenty of the first two, and it has been finding private ways to handle the third.
In February, Databricks completed more than $7 billion of new investment capacity, including about $5 billion of equity financing at a $134 billion valuation and about $2 billion of additional debt capacity. That followed a late 2024 financing at a $62 billion valuation and a further jump past $100 billion in 2025. In practical terms, the company has raised public-company scale money while staying outside the public market.
That gives Ghodsi room to keep spending where the company thinks the next enterprise AI market will form. Databricks is pushing Lakebase, its serverless Postgres database built for AI agents, and Genie, its conversational tool that lets employees ask questions of business data in natural language. It has also built Agent Bricks for enterprise AI agents and earlier bought MosaicML in a deal valued at about $1.3 billion to deepen its model training capabilities.
The point is not just that Databricks has AI products. Many companies can make that claim now. The stronger position is that Databricks sits close to the data enterprises already trust, govern and protect. If companies are going to build useful AI agents, they need those systems to work with proprietary data, not just public internet knowledge. That is where Databricks wants to become infrastructure, not a feature.
The IPO signal goes beyond one company
Ghodsi's hesitation also says something uncomfortable about the 2026 IPO market. Public investors may want AI listings, but private AI companies may not like the trade being offered. Once a company lists, the valuation has to survive daily trading, earnings calls, software-sector rotations and every new question about whether AI spending is producing real returns.
Databricks competes with public companies including Snowflake and Oracle, which means the market already has ways to price data infrastructure. That can be useful, but it can also be limiting. A private investor may value Databricks as one of the main operating systems for enterprise AI. A public investor may compare it more directly with software multiples, cloud consumption trends and margins. Those are different conversations.
This is why staying private can be a strategic choice rather than a defensive one. Databricks can keep funding research, acquisitions and international expansion without having to explain every expense as either margin pressure or growth investment. It can also wait for more proof that products like Lakebase and Genie are not just early AI demand, but durable revenue lines.
There is still a cost to waiting. Employees eventually need liquidity, and the longer a company stays private at a very high valuation, the more complicated that liquidity becomes. Secondary sales can help, but they are not the same as a broad public market.
For the wider AI market, Databricks is a useful test case. If a company with strong growth, positive free cash flow, a $134 billion valuation and clear enterprise demand still does not want to go public, then the issue is not simply whether the IPO window is open. The better question is whether public markets are offering enough patience for the next phase of AI infrastructure.
That is what to watch next. Databricks will almost certainly remain an IPO candidate, but the timing will depend less on market excitement and more on whether the company believes public investors understand what it is becoming. Until then, private capital is giving it the one thing every ambitious AI company wants most: time.
Also read: AI data centers are making electricity the next industry bottleneck • AI data centers are turning electricity into the new chip constraint • China has turned brain implants into a commercial medical race
