At 1:23:04 a.m. the experiment began. The steam to the turbines was shut off, and a run down of the turbine generator began, together with four (of eight total) Main Circulating Pumps (MCP). The diesel generator started and sequentially picked up loads, which was complete by 01:23:43; during this period the power for these four MCPs was supplied by the coasting down turbine generator. As the momentum of the turbine generator that powered the water pumps decreased, the water flow rate decreased, leading to increased formation of steam voids (bubbles) in the core. Because of the positive void coefficient of the RBMK reactor at low reactor power levels, it was now primed to embark on a positive feedback loop, in which the formation of steam voids reduced the ability of the liquid water coolant to absorb neutrons, which in turn increased the reactor's power output. This caused yet more water to flash into steam, giving yet a further power increase. However, during almost the entire period of the experiment the automatic control system successfully counteracted this positive feedback, continuously inserting control rods into the reactor core to limit the power rise.

At 1:23:40, as recorded by the SKALA centralized control system, an emergency shutdown or scram of the reactor was initiated. The scram was started when the EPS-5 button (also known as the AZ-5 button) of the reactor emergency protection system was pressed thus fully inserting all control rods, including the manual control rods that had been incautiously withdrawn earlier. The reason the EPS-5 button was pressed is not known, whether it was done as an emergency measure or simply as a routine method of shutting down the reactor upon completion of the experiment. There is a view that the scram may have been ordered as a response to the unexpected rapid power increase, although there is no recorded data convincingly testifying to this. Some have suggested that the button was not pressed but rather that the signal was automatically produced by the emergency protection system; however, the SKALA clearly registered a manual scram signal. In spite of this, the question as to when or even whether the EPS-5 button was pressed was the subject of debate. There are assertions that the pressure was caused by the rapid power acceleration at the start, and allegations that the button was not pressed until the reactor began to self-destruct but others assert that it happened earlier and in calm conditions. For whatever reason the EPS-5 button was pressed, insertion of control rods into the reactor core began. The control rod insertion mechanism operated at a relatively slow speed (0.4 m/s) taking 18–20 seconds for the rods to travel the full approximately 7-meter core length (height). A bigger problem was a flawed graphite-tip control rod design, which initially displaced coolant before neutron-absorbing material was inserted and the reaction slowed. As a result, the scram actually increased the reaction rate in the lower half of the core.

A few seconds after the start of the scram, a massive power spike occurred, the core overheated, and seconds later resulted in the initial explosion. Some of the fuel rods fractured, blocking the control rod columns and causing the control rods to become stuck after being inserted only one-third of the way. Within three seconds the reactor output rose above 530 MW. The subsequent course of events was not registered by instruments: it is known only as a result of mathematical simulation. First a great rise in power caused an increase in fuel temperature and massive steam buildup with rapid increase in steam pressure. This destroyed fuel elements and ruptured the channels in which these elements were located. Then according to some estimations, the reactor jumped to around 30 GW thermal, ten times the normal operational output. It was not possible to reconstruct the precise sequence of the processes that led to the destruction of the reactor and the power unit building. There is a general understanding that it was steam from the wrecked channels entering the reactor inner structure that caused the destruction of the reactor casing, tearing off and lifting by force the 2,000 ton upper plate (to which the entire reactor assembly is fastened). Apparently this was the first explosion that many heard. This was a steam explosion like the explosion of a steam boiler from the excess pressure of vapor. This ruptured further fuel channels—as a result the remaining coolant flashed to steam and escaped the reactor core. The total water loss combined with a high positive void coefficient to increase the reactor power.

A second, more powerful explosion occurred about two or three seconds after the first; evidence indicates that the second explosion resulted from a nuclear excursion. The nuclear excursion dispersed the core and effectively terminated that phase of the event. However, the graphite fire continued, greatly contributing to the spread of radioactive material and the contamination of outlying areas. There were initially several hypotheses about the nature of the second explosion. One view was that "the second explosion was caused by the hydrogen which had been produced either by the overheated steam-zirconium reaction or by the reaction of red-hot graphite with steam that produce hydrogen and carbon monoxide." Another hypothesis posits that the second explosion was a thermal explosion of the reactor as a result of the uncontrollable escape of fast neutrons caused by the complete water loss in the reactor core. A third hypothesis was that the explosion was caused, exceptionally, by steam. According to this version, the flow of steam and the steam pressure caused all the destruction following the ejection from the shaft of a substantial part of the graphite and fuel.

According to observers outside Unit 4, burning lumps of material and sparks shot into the air above the reactor. Some of them fell on to the roof of the machine hall and started a fire. About 25 per cent of the red-hot graphite blocks and overheated material from the fuel channels was ejected. ... Parts of the graphite blocks and fuel channels were out of the reactor building. ... As a result of the damage to the building an airflow through the core was established by the high temperature of the core. The air ignited the hot graphite and started a graphite fire.

However, the ratio of xenon radioisotopes released during the event provides compelling evidence that the second explosion was a nuclear power transient. This nuclear transient released ~0.01 kiloton of TNT equivalent (40 GJ) of energy; the analysis indicates that the nuclear excursion was limited to a small portion of the core.

Contrary to safety regulations, a combustible material (bitumen) had been used in the construction of the roof of the reactor building and the turbine hall. Ejected material ignited at least five fires on the roof of the (still operating) adjacent reactor 3. It was imperative to put those fires out and protect the cooling systems of reactor 3. Inside reactor 3, the chief of the night shift, Yuri Bagdasarov, wanted to shut down the reactor immediately, but chief engineer Nikolai Fomin would not allow this. The operators were given respirators and potassium iodide tablets and told to continue working. At 05:00, however, Bagdasarov made his own decision to shut down the reactor, leaving only those operators there who had to work the emergency cooling systems.