Why AI Is Making Infrastructure a Management Theme (AI column)
Went out a few days ago. RBC Why businesses choose hybrid infrastructure. In this interview, RBC from 19 In May, Yandex Cloud CEO Grigory Atrepiev said that the corporate software market in Russia 2025 year-round 808 The two main factors of market change are information security and artificial intelligence.
Today I teach all day in the framework of the HSE program.AI Leaders: A Business Lab for Executives” and tell managers about cloud, DataOps, MLOps, and AIOps.” And this article perfectly fits into the main thesis of my story: AI in a company does not begin with a beautiful demo model. It starts with infrastructure, data, security, operation and management readiness to bring the pilot to industrial adoption.
If you go back to the theses of Gregory from the article, they are something like this.
1Artificial intelligence and information security is a combo. On the one hand, companies are living in a more aggressive environment: according to Yandex Cloud, last year, they were processing. 103 There are billions of events every day in your own SIEM system, three times more year-on-year. On the other hand, AI has already ceased to be just an experiment: on the platform, according to the company, created more 18 thousands of agents. 2Next, the most interesting thing for managers begins. AI is rapidly becoming an infrastructure challenge. The demand for GPUs is growing, according to a study by Yandex Cloud and Apple Hills Digital, the average annual growth rate of this segment is up to 2030 yearly 23%. Change the requirements for the data center: if previously the rack could consume 5-6 KW, now for AI-loads we are talking about 100 KW and higher. 3️ Hybrid infrastructure does not look like a compromise between cloud and hardware, but rather a working model for a mature company. The public cloud provides speed of experimentation, fast pilots and flexible resource consumption. A private circuit is needed where there are sensitive data, regulatory restrictions, information security requirements and already established corporate systems. 4️ But hybridity doesn't come for free. The article lists the barriers well: different tools and skills, Kubernetes, API, monitoring, synchronization, backup, security. And with AI, a new layer is added: you need to control agents in enterprise systems, delineate access to data, protect protocols, and observe chains of action. Observability is becoming not only the topic of SRE, but also the topic of AI risk management.
In general, the managerial conclusion is this: AI projects often break down not on choosing or accessing your favorite model. They are beginning to stall on the quality of data, the availability of infrastructure, the lack of a strong business sponsor, an obscure model of responsibility and the inability to turn a pilot into a repeatable process. Therefore, it is important for managers to understand not only what neural network to buy. It is important to understand what kind of data, clouds, MLOps/AIOps, information security, monitoring and operation are needed so that AI does not remain a presentation for several employees, but becomes part of the corporate system.
#AI #Cloud #DataOps #MLOps #AIOps #Management #Engineering