The hosting market can no longer be reduced to disk space, CPU cores, and server uptime. “Machines now make up most of the internet,” Ivan Čačija, a Management Board member at Info-Impero, which operates the Fastserver.hr brand, tells ICTbusiness Media - ICTbusiness.info. According to the figures he cites, automated systems now generate more than 53 percent of web traffic, creating both a security and an economic challenge for hosting companies.
Every request consumes computing, network, and security resources, regardless of whether it comes from a real user, a legitimate AI agent, or a malicious bot. Čačija therefore believes that understanding traffic and deciding what should be allowed to reach an application will become a key part of modern hosting infrastructure. What business customers are buying is also changing: gigabytes and hardware matter less than the availability and performance of their own operations. AI is accelerating that shift because it can help predict anomalies, scale resources, and manage infrastructure, while at the same time creating new demand for GPU capacity, power, cooling, and networking.
In that environment, Čačija does not see the future as a choice between local providers and global cloud platforms, but as a combination of private, local, and public infrastructure tailored to each workload. He sees the advantage of regional providers in proximity to customers, a more individual approach, local infrastructure, and knowledge of the regulatory environment. Fastserver’s partnership with Blackwall is intended to identify and stop unwanted traffic earlier, before it starts consuming resources in customer applications. Čačija sees Fastserver moving towards an integrated platform that brings together compute, storage, networking, security, traffic protection, monitoring and automation.
More than half of internet traffic is now generated by automated systems and bots. How does this change in the structure of internet traffic affect hosting companies, data centres and their customers?
The internet has changed significantly over the past few years. We can no longer assume that there is a person behind every request. According to the latest data, more than 53 percent of web traffic is generated by automated systems. In other words, machines have already become the majority of the internet.
Of course, not every bot is bad. We have search engines, monitoring systems, API clients, AI crawlers, and a whole range of legitimate automated services. However, a large share of traffic consists of scraping, credential stuffing, brute-force attempts, automated exploitation of vulnerabilities and various forms of DDoS attacks. A year earlier, when automated traffic accounted for 51 percent of total web traffic, as much as 37 percent of all traffic came from so-called “bad bots”.
For a hosting company, and ultimately for the customer, the problem is very concrete: every request costs something. Someone has to accept the connection, terminate TLS, process the HTTP request, perhaps run PHP, make several SQL queries, and return the result.
If a significant share of those requests does not come from real users, the customer may effectively be paying for CPU, memory, bandwidth, and security resources consumed by machines that bring no value to the business.
That is why bot traffic is no longer only a security problem, but an economic one as well.
An additional challenge is that it is becoming increasingly difficult to distinguish between a legitimate user, a legitimate AI agent, and a malicious bot. Modern bots use real browsers, execute JavaScript, change IP addresses, and try to imitate human behaviour. With the emergence of AI agents, that boundary will become even less clear.
That is why I believe the ability to understand traffic — who is accessing an application, why they are accessing it, and whether that traffic should be allowed to reach it at all — will become one of the key elements of modern hosting infrastructure.
Hosting used to be viewed primarily in terms of storage capacity, processing resources, and server availability. Today, speed, latency, security and infrastructure availability directly affect business results. How much has what customers are actually buying from a hosting company changed?
Quite radically. Hosting used to be sold in terms of gigabytes of storage, traffic allowances and the number of email addresses. Today, a serious business customer is not really buying gigabytes and CPU cores.
They are buying the availability and performance of their own business.
If you run an online store generating several million euros in annual revenue, the key question is no longer whether hosting costs 100 or 300 euros a month. A much more important question is how much it costs you if the store is slow or unavailable during your biggest sales campaign.
There is also a very concrete link between performance and business results. A Deloitte study found that an improvement of just 0.1 seconds in mobile site speed was associated with an 8.4 percent increase in retail conversions and a 9.2 percent increase in average order value. Of course, that does not mean every webshop will achieve the same result, but it clearly shows that milliseconds are no longer just a technical metric.
I like to explain it very simply: the end customer does not care how many processors a server has. What matters is how quickly the product page opens and whether the purchase can be completed.
I think Croatian companies increasingly understand this, especially those for which the internet is a direct sales channel. But we still see the paradox of tens or hundreds of thousands of euros being invested in digital marketing while, at the same time, companies try to save a few dozen euros a month on infrastructure.
You pay Google or Meta to bring you a customer, and then greet that customer with a slow online store.
That is why a quality hosting service today is not defined by hardware alone. The difference comes from architecture, network, storage, redundancy, backup, security, monitoring, response time, and the people who know what to do when something goes wrong.
Artificial intelligence and automation are changing almost every segment of the IT industry. How is AI changing the hosting industry itself and the way data-centre infrastructure will be managed?
AI will change the hosting industry in two directions: we will use it to manage infrastructure, while at the same time we will have to build infrastructure capable of running an increasing number of AI workloads.
Data centres already generate enormous amounts of data today: CPU and memory metrics, storage status, network traffic, latency, application logs and security events. AI can connect information that a person simply cannot track in real time.
A system will be able to recognise an anomaly before it becomes an incident, predict an increase in load, automatically add resources or move a workload.
In security, that may be even more important. The classic approach of “this IP address is bad, so we will block it” is no longer enough. We have to analyse traffic behaviour and make decisions practically in real time.
I think we will gradually move towards infrastructure where we increasingly specify what we want to achieve, and less and less how it should be done technically. The platform itself will decide where a workload should run, how many resources it needs, and when it should scale.
The administrator will not disappear, but the role will change — less repetitive work and more architecture, security, and decision-making.
On the other hand, AI is creating entirely new demand for infrastructure. It is no longer only about CPU, RAM and storage. GPU infrastructure brings completely different requirements in terms of power consumption, cooling and network infrastructure.
AI therefore will not change only the software in a data centre. It will also change how the data centre itself is designed.
Companies can now choose between the global public cloud, local cloud, dedicated infrastructure, colocation and different hybrid models. In that environment, where do you see room for domestic and regional hosting and data-centre service providers?
I do not think regional providers should try to become a small AWS or Azure. Hyperscalers have fantastic platforms and thousands of services. Our advantage lies elsewhere.
It is proximity to the customer, an individual approach, local infrastructure, knowledge of the regulatory and business environment, and the ability for customers to speak directly with the people managing their infrastructure.
Cost predictability is also important. Public cloud is excellent for certain scenarios, but not every workload is ideal for a hyperscaler. A system that continuously uses large amounts of CPU, RAM, storage, and traffic often has very different economics on dedicated or private-cloud infrastructure.
Questions of data location and digital sovereignty are also becoming increasingly important. For some customers, it is no longer enough to know that their data is “somewhere in Europe”. They want to know where it is physically located, who has access to the infrastructure, where the backups are, and under whose jurisdiction it falls.
This is where regional providers have a very clear role.
But I do not see the future as a choice between a local provider and AWS, Azure, or other global platforms. I see it as a combination.
A core business application and database can run on private infrastructure in Croatia; the backup can be at another location within the EU; CDN and protection can be globally distributed, while a particular AI or analytics workload can use the public cloud.
The point is not to put everything on one platform. The point is to choose the infrastructure that makes the most sense for each workload.
You have entered into a strategic partnership with Blackwall. What does it bring you in concrete terms, and which customer problem are you trying to solve through this cooperation?
Blackwall fits very naturally with the issue we discussed at the beginning of the conversation: an internet where an increasing share of traffic is no longer generated by people.
Our goal is to identify and stop unwanted traffic as early as possible, before it starts consuming the resources of the customer’s application and infrastructure.
We are implementing Blackwall GateKeeper, a protective reverse proxy positioned in front of the web infrastructure that analyses traffic before it reaches the application itself. The system combines behavioural analysis, signature-based detection and heuristics to distinguish legitimate traffic from bots and other threats.
In addition to bot management, it provides protection against L7 DDoS and brute-force attacks, along with other security and performance features such as a WAF, rate limiting, caching, HTTP/3 and TLS 1.3.
What is particularly interesting to us is that we can implement GateKeeper directly within our own infrastructure. Traffic analysis can therefore take place locally, right at the entry point to the hosting infrastructure.
Security is moving towards the edge of the infrastructure. It makes no sense to let a malicious request reach WordPress, a webshop or an API, allow it to execute application code, make ten SQL queries and consume CPU, only to reject it at the end.
The cheapest malicious request is the one that never reaches the application.
We therefore see the partnership with Blackwall as part of a broader transformation of Fastserver. The hosting company of the future will not sell customers only a virtual server or a certain number of CPU cores.
It will have to offer an integrated platform that includes compute, storage, networking, backup, performance, security, traffic protection, monitoring and automation.
In the end, the customer does not care which of our servers is running the application. What matters is that it is fast, available, and secure. And that is precisely the direction in which we want to develop Fastserver.
