As fewer and fewer humans are involved, uptime increases – and new issues arise.
Web hosting is a zero-sum game where anything but perfect service is a challenge to your operations. Unplanned downtime already costs firms an estimated $400 billion a year, or nearly 10% of the profits of companies, according to a June 2024 report by Splunk and Oxford Economics that surveyed 2,000 executives worldwide.
When something goes wrong in the hosting world, it has long been the case that a human employee needs to identify the issue and try to fix it – a costly solution requiring on-call teams that command higher salaries and extra charges.
But a change could be coming, thanks to the artificial intelligence (AI) revolution.
Hosting providers are coming to the realization that AI can slash incident response times and prevent human error, reducing that eye-watering cost of unexpected downtime on a business’s bottom line.
Work is already underway to help with that. At Google Cloud Next ’24, the search giant unveiled Gemini Cloud Assist, a generative-AI co-pilot that troubleshoots workloads, rewrites infrastructure-as-code, and recommends cost tweaks from a chat prompt.
Google isn’t the only competitor in the space working on these issues. Amazon recently rewrote its own playbook to try and accommodate such changes. The AWS Well-Architected Framework now advises architects to “automate healing on all layers”, from replacing failed EC2 instances to triggering cross-region database failovers without paging a human.
From NoCode to NoOps
But automation of this type isn’t confined to hyperscale data centres. Cloudflare’s Developer Week last year saw 18 upgrades or new features, from automatic Git-based build previews, gradual global roll-outs, and one-command logging, all of which were designed to make full-stack deployment easier.
Rivals Netlify and Vercel offer similar so-called “NoOps” pipelines, but Cloudflare argues that a wider network absorbs spikes more cheaply.
Even traditional registrars have joined the hype cycle. In May 2025, NameSilo claimed victory with “Self-Healing Hosting: How AI Is Changing Web Uptime Forever”, promising 99.99% availability by fixing faults before users notice. The language mirrors a real trend: customers now view predictive scaling and automated restarts as baseline features, not premium add-ons.
Separating the hype from the reality can be challenging. The 2024 Accelerate State of DevOps report – produced by Google’s DORA research group – found that while roughly a quarter of surveyed teams now embed AI in their pipelines, 39% of engineers “have little or no trust” in AI-generated code.
In the meantime, some high-AI organisations actually saw a 7% dip in software-delivery stability after adopting AI.
The big picture
How much to benefit from AI’s potential, versus how much to protect against its risks, is on the minds of regulators. In April, the European Commission opened consultations on a draft Cloud and AI Development Act aimed at tripling EU data-centre capacity while safeguarding competition and trust. Although still only a proposal, it hints that autonomous infrastructure may soon face the same disclosure and audit rules already applied to personal-data processing.
It all means that full NoOps hosting operations are likely to be a long way off. Even AWS’s guidance stops short of removing people; it stresses designing for automated failover alongside, rather than instead of, human oversight.
Even then, the industry is undertaking an experiment to see whether language models and policy-driven scripts can keep millions of sites online without blowing up budgets or spawning silent failures. The early signals are encouraging: Splunk’s research shows “resilience-leader” firms adopt embedded AI four times more than laggards and endured far fewer outages, although that’s in a handful of examples and is very early days. While the hard data is difficult to come by, the industry is excited by the opportunities ahead.
Your email address will not be published. Required fields are markedmarked