Artificial Intelligence Data Centers and Mining Data Centers

Hodlx Guest Post Send your message

In recent months, work burdens have moved from artificial intelligence (artificial intelligence) from theoretical standards to economic pressure in the actual time of global infrastructure.

One of the linguistic models that serves millions of queries per hour to spread models that require wide groups of GPU to infer, the pressure on energy networks and resource calculation accelerates.

Surprisingly, the infrastructure that is better placed to absorb this load is not found in the silicon valley or the superficial farmer – But in mining data centers.

From the prisoner of war (proof of work) to obstetric artificial intelligence

Current currency mining centers were built on the hypothesis of high -density and energy density – Implementation of efficiency, employment and thermal control.

These are the same foundations required for the Natural Intelligence.

But there is a decisive difference – Although mining operations are relatively enjoyable and can be boycotted without losing work, the burdens of artificial intelligence work are sustainable, precisely driven by delay.

This contrast is displayed an opportunity.

By upgrading cooling systems – Especially through indulgence and liquid -based techniques – Improving the infrastructure for the distribution of energy and mining Data centers It can become hybrid environments.

They can run encryption mining when energy costs are low and switch to intelligence inference functions when ordering GPU mutations.

Emerging synchronization platforms, along with AI’s scheduling tools, allow the dynamic switch between tasks.

These tools have shown up to 27 to 33 % in times of completing the job and 1.53X Discounts In the last waiting list.

The economic layer is equally convincing – If the demand for artificial intelligence is invested through the inference markets, mining operations may find that it is more profitable to rent an arithmetic energy instead of extracting certain assets.

Some mining centers Truly with FPGA on the basis Settings, which are ASIC resistance and original suitable Artificial intelligence training.

This opens the door to full inter -employment – Where the same infrastructure treats Pow blocks and Transforme modelsDepending on market conditions.

When the scale becomes responsible

Despite its early progress in artificial intelligence investments, the United States is facing a wall of infrastructure on the horizon. in VirginiaConsuming data centers More than 25 % From state electricity.

In Santa Clara, more 50 data centers Now draw 60 % of the total energy use in the city, which is forced Silicon Valley Authority Largely Expand Transport systems – Raising rates for both industrial and residential users.

numerous research This shows Global demand for electricity It can be more than three times by 2030, to a large extent Because of the artificial intelligence.

If these expectations are designed, the United States will not need Achieving a balance between strategies – What is the excessive traditional AI facilities, associated with solid SLAS solid, well.

To meet this high demand, the United States must quickly diversify energy sources.

Renewable energy sources limit – Including solar energy, wind, wind and hydroelectric energy – You will play an important role.

However, these sources are intermittently intermittently, creating volatility on the network. This is where mining data centers provide an amazing stability feature.

It is designed with a flexible structure of the order, and it can stop operating or suffocating on the basis of network pregnancy, absorbing excess generation during renewable peak and decreasing during low production periods.

In Texas, this flexibility has already led to cooperative cooperation agreements between mining operations and network operators, as these facilities were placed as a great value in energy management of the next generation.

Alternative strategies are also emerging. electricity Imports from CanadaEspecially through HVDC (the current voltage) lines that exploit the electrical energy, are below Active Explore.

On the local front, SMRS (small units reactor) represents a promising path.

SMRS has already been developed by many companies – Ideal for conjugation With regional artificial intelligence centers Heavy facilities account.

The following artificial intelligence limits

Bitcoin mining was the early engine in this direction. However, the real story is not only related to mining – It is related to what comes next.

Mining infrastructure paves the way for artificial intelligence to calculate the size.

These facilities test the reasons – When local talents are trained, operational operations are refined and organizational tracks are explored.

Through modest and improved upright upgrades, many mining centers can burn the burdens of artificial intelligence, which provides low -cost spinal and costly effective to infer the global model.

The door for full interface

What is required is to reformulate the infrastructure of the data center you should look at in the era of artificial intelligence.

Instead of failing to pay for hyper Easter, the future may be normative, flexible and geographically distributed, led by hybrid centers that know how to manage thermal loads, improve the cost per watt and change the operational models in the actual time.

Patter is the founder and CEO of uminersMining infrastructure provider in a full cycle. It has a deep background in the development of the data center, the coded currency mining and the techniques driven by artificial intelligence.

Distinguished Image: Shutterstock/Iurii/Vladimir Sazonov