Ask a VP of engineering how their team feels about AI-generated code: It’s fast and it’s everywhere, but nobody’s fully sure it’s safe yet. And behind that unease is some staggering math.
What $1 trillion buys the industry
Start with what companies already spend trying to catch bad code before it reaches anyone. Global IT spending is projected to hit $6.31 trillion this year, according to Gartner. Fortune 2000 companies are estimated to account for about 75% of total spending in this area, with average quality assurance and testing allocations typically representing around 22% of an organization’s IT budget.
Run those three multipliers together, and you land just above $1 trillion — the size of the quality assurance and testing industry today. Salaries for QA professionals, staging and shadow environments, outsourced testing partners, and licenses for the automation and GenAI tools meant to catch problems before customers do. All of it, adding up to more than a trillion dollars a year, spent purely on defense.
What’s getting through anyway
Despite all of that spending, a lot of bad code is still getting through. And AI is increasingly writing it.
CISQ’s most recent Cost of Poor Software Quality report puts the annual toll from operational software failures — the outages and defects that actually reach production — at roughly $1.56 trillion. In our own 2026 research with Wakefield Research, 80% of organizations deploying AI-generated code say they’ve already traced a real production incident back to it. Apply that rate to CISQ’s operational-failure figure, and you land somewhere between $1.2 and $1.3 trillion in production problems now connected to AI-generated code specifically.
Two trillion-dollar figures, one uncomfortable picture
So, organizations worldwide are spending just over $1 trillion trying to keep bad code out of production, yet something like like $1.2 to $1.3 trillion in AI-linked failures are getting through anyway. Almost unbelievably, the cost of what’s slipping past the safety net is starting to outpace the cost of the safety net itself.
The safety net wasn’t built for the speed of AI
In a nutshell, this problem underscores the code verification crisis: The review process was sized for code written at human speed, but AI writes at a pace no QA org was ever staffed to match.
Every sprint, someone on your team is making a judgment call about which AI-written code gets scrutinized and which gets a quick glance, simply because there isn’t time to give the barrage of code generated equal attention. Multiply that single decision across every engineering team currently shipping AI-assisted code, and the gap shows up everywhere: late-night emergencies, rollback scrambles, postmortems, customer escalations, lost revenue, and a slow, quiet erosion of what “tested” even means anymore.
The crisis is precisely what Sauce Labs built AURA to solve, giving the verification process AI-level speed to match AI-powered code generation. With AI-unified release assurance, the trillion-dollar safety net actually stretches to cover what’s landing in it now.
Is your team feeling the crisis today? Your problems are not unique. The entire industry feels this pain at some level.
Let’s talk about what closing the gap between speed and release confidence could look like for your next release cycle. Book a demo to see AURA in action.
