The private credit party just got very selective about who gets in.
The Summary
- Thoma Bravo's $2.5 billion refinancing for cybersecurity firm Sophos is struggling to find private credit lenders, a sharp reversal from 2025 when lenders competed for software deals
- AI anxiety is driving the pullback — lenders now questioning whether traditional software firms can compete as AI rewrites entire product categories
- This is the credit market's early warning system blinking red on legacy enterprise software
The Signal
Private credit just drew a line between software that survives the AI transition and software that doesn't. Sophos, a 40-year-old cybersecurity company owned by Thoma Bravo, is learning this the hard way. Twelve months ago, this $2.5 billion refinancing would have been oversubscribed in days. Now lenders are walking away.
The shift isn't about Sophos specifically. It's about what Sophos represents: enterprise software built before agents became the default interface. Lenders are asking a new question before they write checks: Can this company's product be rebuilt by an AI agent in six months?
"AI anxiety has snarled up the market so much that Thoma Bravo's $2.5 billion refinancing is facing hesitant lenders."
Cybersecurity should be agent-proof, right? Threats evolve, humans need protection, the product writes itself. But lenders see something different. They see legacy architectures, manual configuration, and user interfaces designed for IT teams who might not exist in five years. They see software that requires humans to operate it, which means software at risk.
The private credit market moves fast because it has to. These firms lend billions based on cash flow projections, and those projections just got a lot harder to trust. If your revenue comes from seats, subscriptions, or support contracts tied to human workflows, you're in the danger zone.
Key factors driving lender hesitation:
- Traditional software pricing models (per-seat, per-user) vulnerable to agent consolidation
- Long sales cycles and complex implementations that AI-native competitors can skip
- Technical debt in core products that makes rapid AI integration expensive
This is the same pattern we saw in 2023 when SaaS multiples collapsed, but the stakes are higher now. Back then, investors worried about growth rates and burn. Now they're worried about obsolescence. The difference: growth rates can recover. Product categories don't come back once agents replace them.
Thoma Bravo has $130 billion in assets under management and a two-decade track record of buying, optimizing, and flipping enterprise software companies. If they're struggling to find $2.5 billion for a refinancing, not an acquisition, that's information. The playbook that worked for 20 years just hit a wall.
What makes this especially telling is the timing. We're not in a credit crisis. Private credit firms are sitting on record amounts of dry powder, desperate to deploy capital. They're saying no not because they lack funds, but because they lack confidence in the business model.
The Implication
Watch where private credit flows next. If lenders are pulling back from established cybersecurity firms, they're going to be even more skeptical of HR software, project management tools, and anything else that assumes humans are the primary users. The capital is moving toward infrastructure that serves agents, platforms that agents build on, and tools that get better as automation increases.
For founders: if your pitch deck still shows human seats as your unit economics, start rewriting. For investors in traditional enterprise software: the repricing has started, and it's coming from the debt side first. When lenders lose confidence before equity holders do, the correction tends to be swift.