Cloudflare Says: Making Money from AI with Legacy Applications Is Nearly Impossible

I was recently chatting with a technology director. They were lamenting the hundreds of thousands of dollars they spent on AI and the - let's be honest - mediocre results they got. "We feel like we're doing everything right," they said, "but we're not seeing transformation." The source of the problem is often hidden in the most unexpected place: the foundations. And the new research data shared by Cloudflare today points exactly to this, and in a rather harsh way.

Money Pits and Empty Back Doors

Companies, especially large organizations, exhibit a kind of schizophrenic attitude towards AI. On one hand, they are racing to pour money into the latest, shiniest LLM models. On the other hand, the applications that host, analyze, and present the data these models will feed on often stubbornly persist with an architecture from the early 2010s. This is no different from trying to fit a Ferrari engine onto a horse-drawn carriage chassis.

Cloudflare's analysis shows that this fiasco is not just a theory; it's backed by concrete data. Companies that modernize their applications are three times more likely to achieve measurable financial return from their AI projects compared to those that don't. Three times! This is not a "slightly better" difference; it's a "either you change the game or you get wiped off the field" kind of gap.

Don't Let 'Modernization' Just Mean 'Moving to the Cloud'

Let's correct a critical misconception here. For many executives, 'modernization' simply means taking the application and moving it to a cloud provider's virtual machine. No. The real issue is the application's flexibility, speed, and security. Microservices architecture, containers, APIs, advanced security layers, and global-scale performance. AI models demand instant responses. An old, monolithic application gasps for breath with every query. Data flow gets clogged, latency increases, and your expensive AI model ends up watching the old system's "loading" spinner in the background when it should be responding to the user in seconds.

This situation coincides with a period where AI's role is being questioned. As we discussed in our article titled AI Will No Longer Just Answer, It Will Manage People, systems are moving towards the center of decision-making mechanisms. In such an environment, your infrastructure not being reliable and agile becomes not just an inefficiency, but an existential risk.

The CIO's Nightmare: Investment Priority Wars

So why do so many companies still try to skip this step? The answer is simple: Unglamorous work struggles to secure budget from glamorous projects. Telling a board of directors "Let's modernize the currently used customer portal" doesn't have the same impact as saying "Let's develop a custom large language model for our company." One is marketed as boring infrastructure cost, the other as an investment in the future. But Cloudflare's data reveals that the real future investment lies precisely in that 'boring' infrastructure.

This is also the micro-scale reflection of the global technology leadership debates we touched on in the article Davos 2026: How the World Economic Forum Transformed into a Technology Summit?. While leaders focus on flashy projects, they can neglect the technological foundations that form the basis of competition.

So, What Should You Do Now?

First, take a breath and look ruthlessly at your current application portfolio. Which applications are critical data sources? Which ones are at the heart of processes expected to integrate with AI? Your modernization roadmap should start exactly with these applications. The "lift-and-shift" method is not enough. A real architectural transformation is required. It's an expensive and arduous process, yes. But Cloudflare's data shows that every investor or executive expecting returns from AI must first accept this bill. Otherwise, the AI dream will end with old-system insomnia.

Frequently Asked Questions