On June 15, 2026, Ververica, a company founded by the original creators of the open source Apache Flink project, published a blog post called “Sovereign By Design. No Exceptions.” As of June 30, 2026, Ververica is shutting down its fully managed, multi-tenant cloud service. Every customer is being migrated to either BYOC (Bring Your Own Cloud) or a self-managed, on-premises deployment.

“BYOC is the cloud, deployed where customers have the governance to run it.”

The blog posits that every serious customer of a fully managed cloud service eventually migrates off of it, once governance requirements intervene. So, rather than maintaining a deployment model that regulated buyers are abandoning, Ververica is preemptively killing it, and pivoting to BYOC.

Ververica’s sudden pivot is unusual, but it springs from a trend that has been growing for several years: the move from SaaS to BYOC.

What is BYOC?

BYOC is a software deployment strategy that bifurcates software deployment into two “planes.” 

• The data plane consists of customer data, as well as the application software that interacts with the data. It runs locally inside the customer’s own cloud account. 
• The control plane is software that allows the vendor to manage the data plane by pushing updates, observing server health, and gatekeeping which customers get which features. This layer is run by the vendor. 

In this two-part approach, the customer has broad latitude to keep house on its own. It can apply its own IAM (Identity and Access Management) policies, VPC (Virtual Private Cloud), and encryption keys, all to suit its own security needs. The vendor only manages the software, and does not have custody or control of data. This approach tries to be the best of both worlds. The customer gets the benefits of SaaS, like low costs of onboarding and continuous software update management, without the biggest risk of SaaS: a vendor data breach that is outside the customer’s control.

Why Now?

Three forces are converging:

Growing Regulation. The wild west days of software are over. Europe is advancing new regulations on the storage and movement of data every year, and the US is following suit (albeit in a slower and less coordinated way). The EU’s GDPR (the General Data Protection Regulation) fines exceeded €3 billion in just the first half of 2025. China’s Network Data Security Management rules, in effect since January 2025, impose strict in-country storage requirements on certain industries. Regulations in the US for sensitive industries (like HIPAA in healthcare) require that companies track exactly where their data resides and who can access it. 

Eroding Supply-Chain Trust. After the SolarWinds debacle, and a string of SaaS-vendor breaches including the widely reported Snowflake customer-credential incidents, customers began to realize that they had inherited their vendors’ security issues. While some customers might trust the security of their vendors–particularly larger vendors–some customers might reasonably think they can do better. Or, at least, a customer wants to know whether a specific incident has affected them, as opposed to the entire clientele of the vendor. Moreover, large vendors may be bigger targets for security attacks, and segregating data by user disperses the targets for attack vectors. BYOC means enterprises are not at the mercy of their vendors’ security lapses.

Cloud Economics. Enterprise software buyers today often have multi-million dollar, multi-year commitments with AWS, Google, or Azure. In these agreements, customers commit to purchase significant compute capacity in order to reduce per-unit computing costs. This can make paying a SaaS vendor’s bill for infrastructure a bitter pill to swallow. Taking advantage of economies of scale in computing costs has become a key business metric in our age of heavy AI workloads. BYOC aggregates the enterprise’s buying power–kind of like arranging a wedding reception at your supper club to meet your minimum food and drink purchase commitment. It consumes compute capacity against the customer’s AWS commitments, and pushes into lower marginal costs. 

None of this means SaaS is a bad idea. SaaS is still the right model for a huge swath of software, particularly software that manages data with modest sensitivity. But for the categories where the data is highly sensitive—financial transactions, medical records, telecommunications, or human resources data—BYOC is winning the day. Ververica’s own customer base (banking, healthcare, energy, telecom) exactly fits this profile, so perhaps it is no surprise that Ververica is leading this charge.

Ververica made a bold move, but it isn’t alone. For example, Grafana now offers a BYOC deployment of its full managed stack, with the same code and release cadence as its SaaS Grafana Cloud. 

COSS Is the Final Piece of the Puzzle

BYOC solves an infrastructure and trust problem, but BYOC alone doesn’t solve vendor lock-in problems. If a proprietary vendor’s software runs inside your cloud account, you’ve gained control, but you’re still locked into a single vendor’s roadmap, pricing decisions, and continued existence. If that vendor gets acquired, pivots, or simply raises prices, you have to become a price taker, or engage in an expensive and disruptive migration.

Enterprise users demand autonomy and transparency, mainly to ensure business continuity. Commercial open source software (COSS) is the only model that provides all the elements for vendor independence. 

• Sovereignty and auditability. The customer can inspect the software itself, not just the infrastructure. A customer doesn’t have to take a vendor’s word for what the software does with data; it can read the source.
• No lock-in. If the vendor disappears, changes its license, or simply loses its edge, the customer can keep the software running. It may be forked, self-supported, or handed to another COSS provider, but it’s still available to use. 
• Ability to customize. In BYOC, the vendor sets the software image, and customers can’t easily change that. On-premises COSS, in contrast, gives customers complete freedom to customize software. But even in a BYOC arrangement, COSS makes it easier to create custom integrations, because APIs and software logic are always visible.
• Community-hardened reliability. Widely adopted open source projects have been stress-tested “in the wild” against real-world edge cases. Ververica’s own product is a good example: Apache Flink has been tested over a decade of use across thousands of organizations in high sensitivity environments.

COSS is a win for the vendor as well. Research from the Linux Foundation’s 2025 State of Commercial Open Source report, drawing on 25 years of data across 800 VC-backed startups, found that commercial open source companies have consistently outperformed closed-source peers on valuations, funding speed, and liquidity outcomes–particularly in infrastructure software. Ververica itself is a good example; its sovereignty pitch has apparently resonated with the banks and healthcare companies that form its customer base.

Proprietary vendors can build BYOC, too, of course. But a proprietary vendor offering BYOC is still asking its customer to trust a software black box, just one that happens to run on the customer’s own hardware. A COSS vendor offering BYOC can provide transparency, plus sovereignty, plus continuity.

SaaS is Still Alive, but it May Become BYOC 1.0

This trend does not mean that BYOC will overtake SaaS everywhere, or that COSS automatically wins. BYOC requires engineering investment. Building a control plane is not a trivial project; it’s not the kind of project that can be easily vibe-coded. For low-sensitivity software, or for smaller companies that can avoid regulatory constraints, multi-tenant SaaS will remain simpler and cheaper. 

And plenty of COSS branding is a veneer. Some vendors call themselves COSS when their open source offerings are dumbed-down versions, not truly useful for enterprise–they keep key features under proprietary or SaaS terms. Buyers still need to do their diligence on vendors, but the question has migrated from “Can I trust your data center?” to “Can I trust your license, your governance, and your roadmap?”

Still, this is a serious trend. Ververica didn’t just quietly change its product offerings, as do many COSS vendors when they adjust deployment strategies. It published a manifesto and set a hard shutdown date. For years, SaaS has been the go-to COSS business model pioneered by darlings like Confluent, Redis, and Elastic. It remains to be seen whether the rest of the COSS sector follows this migration.