The HubSpot-Salesforce integration has a reputation. And it's not a kind one.
It's the integration people warn each other about. The one that comes up in every migration conversation and every source-of-truth debate. It's the integration that has launched a thousand Slack messages to me that start with some version of "Why isn’t this syncing..."
It didn't get that reputation because HubSpot didn't care. It got it because everyone has been afraid to update (and potentially break) an integration that nearly 11,000 customers use every single day.
As a result, the backend that powers all of it was last refactored in 2014.
That's not a typo. The architecture that sits between the two most-used CRMs in B2B is twelve years old. And finally, FINALLY…HubSpot is rebuilding it.
What HubSpot (Quietly) Announced
Last week, HubSpot published a post in their community forum titled Rebuilding the Engine: What's Next for the HubSpot-Salesforce Integration. It got a fraction of the attention the Spring Spotlight features got, which is criminal because this affects far more customers than most of the AI features being rolled out.
Here's the TL;DR:
HubSpot has been internally developing v2 of the integration: a near-complete refactoring of the backend architecture that processes the data flowing between HubSpot and Salesforce, including workers, sync pipelines, logic definitions. They've been quietly moving objects onto the new platform for over a year, starting with custom objects, tickets, and activities.
As of last month, company sync is live on v2. Deal sync ships in the coming weeks. Contact sync is targeted for a Q4 2026 beta. And the goal is to fully deprecate the legacy sync engine by mid-2027.
The Problems with the Old Salesforce Integration
Before we get to what's new, let's talk about what's been broken. And a lot has been broken for a long time. So much so that my good friend Lauren Ryan basically built her career on helping companies fix the integration before HubSpot eventually hired her as a Solutions Engineer.
These are just a few of the major problems faced by users:
Duplicate Data - The legacy sync engine only deduplicated contacts on email address, which falls apart the moment you encounter job changes, role-based emails, or Salesforce's separate Lead and Contact objects. Most RevOps teams either bought a third-party dedupe tool, wrote custom scripts, or just accepted that a meaningful chunk of their CRM was duplicates.
The Owner Sync - There was no native sync for the owner field, which means every territory change, reorg, or rep transition creates a silent drift between the two systems. Every RevOps team that has scaled past 50 reps has built some custom middleware to handle this, but that often breaks the moment leadership reshuffles the org chart.
API Limits - HubSpot shares Salesforce's daily API call limit with every other tool integrated to your instance, and bulk operations can blow through the allocation and suspend the entire integration. When that happens, syncing stops and you often don't notice until a rep complains a deal isn't showing up.
Picklist Mismatches - HubSpot stores most dropdowns as open text on the backend, which means you can have values outside of the options in settings. Every time you evolve your data model (refined lead sources, updated lifecycle stages) you create a fresh batch of silent sync errors.
The Salesforce Leads Object - HubSpot has Contacts. Salesforce has Leads and Contacts. And reconciling those at sync time is genuinely hard. The hardcoded logic that handles Lead-to-Contact conversion is the same logic HubSpot's own engineering team admits triggers many errors.
The Debugging Experience - When sync errors stack up, the typical workflow is: open a support ticket, wait, get asked for screenshots, wait, eventually get told "this is a known limitation." HubSpot has now publicly confirmed the legacy platform was hard to debug even for the people who built it — which is the right kind of transparency from a vendor, and which means every RevOps team that has fought this integration for the last decade had completely legitimate grievances.
What v2 of the Salesforce Integration Unlocks
The new architecture brings a set of capabilities that customers have been begging for for years. Every object that runs on v2 gets these out of the box: