It’s not a secret that AI is gaining more and more popularity not only within the IT sector but also for education and commerce. According to the Artificial Intelligence Global Executive Study and Research Project conducted in 2019, nine out of ten companies are using AI. Taking into account that the technology can be applied to everything from suggesting products based on previous choices and behavior anticipation to some more advanced case usage such as  SQL server monitoring and load predictions,  within the upcoming years, it’s expected that to cover all the business spheres.

With increasing demand, enterprises require extensive resources to develop and implement AI into their businesses using ML training and big data massives to ensure correct interpreting and acting on the analyzed results. Thus, it poses a certain challenge to Datacenters’ operations as they need to grow their capacity and develop their server infrastructure to be able to cover the users’ need for accessing more resources. It also contributes to the growth of Cloud technology since the ability of quick scaling is crucial.

Data centers benefits from AI

But is it only the challenges that are coming from AI popularity growth? Taking into account the optimization that it brings to the spheres where it’s introduced. As Erich Sanchack, EVP of Operations at Digital Realty said in Cabling Installation & Maintenance roundup, “Implementation of AI in the data center will move us well beyond current DCIM systems and their limitations. Using AI, we are able to create an environment in which not only are all of our power and facilities decisions and processes completely optimized but that our resource planning and even advanced functions like dynamic bandwidth and server allocation are fully automated as well.”

Data centers receive great potential for development too. Introducing AI  mechanisms it’s crucial to make sure that the Machine Learning training was conducted on the sufficient number of use cases and supply human supervision on the first stages as faulty suggestions and actions based on them can not only have a negative financial outcome but are harmful for Data center reputation. But with proper implementation, the benefits over weigh:

Monitoring automation

In order to ensure smooth system performance, all the systems require constant monitoring. While there exist a lot of monitoring tools such as Nagios, SAM, Zabbix, and many others, they still require a lot of human intervention to investigate the possible load reasons only when the load occurs.  AI systems can be trained to execute the primary commands aimed at reducing the number of active requests, killing idle system processes, and predict the load increase before the server usage statistics show the minimal overuse symptoms.   In turn, it increases server stability and minimizes downtime.

Security enhancements

While malicious attacks are getting smarter, there should exist smart solutions to these threats. For sure the defense systems by data centers provide the highest level of protection, but the risk of falling a victim of new vulnerabilities cannot be underestimated. Some malware protection software vendors already started implementing AI in their firewalls and malware protection tools. With access to tons of data, such systems are learning fast how to single out potentially harmful requests and files, improving scanning algorithms to detect cyber threats faster.  In terms of network protection, it surfs through both incoming and outgoing data that helps to predict where a potential threat can occur.  With the help of AI, it is possible to detect any abnormal activity on the server to signal about the potential threat and act on the received information to mitigate the attack before it scales.

Preventing hardware overheating

AI systems use smart sensors installed in the equipment to analyze the normal noise levels and acceptable device temperature to turn on additional cooling if it exceeds certain points. It would allow acting on the smallest symptoms of overheat and prevent the outages. Leading suppliers already start developing hardware with machine learning in mind with a liquid cooling system that can be automatically set up. Under temperature fluctuation conditionals ensuring smooth performance contributes to repair cost savings and prolonging the server lifespan.

Energy saving

The nature of data center operations presupposes an uninterrupted power supply.  Implementing AI can significantly reduce electricity and ventilation consumption by optimizing resource usage and effective distribution. Taking, for instance, Google’s experience after they implemented Deep Mind technology. The results are indeed impressive. They were able to reduce the cooling bills by 40% using Machine learning control recommendations. Extrapolating the technology to other systems may even reduce the power costs even more.

Decreasing need for human resources

Routine tasks automation, smart monitoring tools, and machine control of cooling systems mentioned above as well as inviting AI technology in disaster recovery processes to reduce the workload on data center personnel meaning that they invest the saved time in other tasks. For sure, it also entails staff shortening. But it’s not about machines dominance over humans but rather restructuring and updating vacancies responsibilities. It’s also expected that the current data center workers will need to increase their competencies and acquire new skills adapting to these changes such as basic machine learning mechanisms understandings, supervising training processes, and introduce updates into learning algorithms. For employers, it may mean temporary additional investment in staff training and helping people to adapt to the new circumstances.

Final Thoughts

Along with the challenges for data centers, AI can also open opportunities for development and workflow optimization. Smart forecasting and suggested actions can significantly reduce the time required for finding and eliminating downtime threats, preventing data leakages, and mitigate overload situations before they occur. Industry representatives that realize it and include the AI integration into their roadmaps and global strategic focus will be more prosperous compared to their competitors as they will be able to spend the saved resources for modernization, network improvements and hardware stock. Moreover, they will be able to offer more competitive prices than companies that neglect the AI potential.  So, it’s only a matter of time when data centers that do not use this chance to their advantage fall out of the market competition.

It’s a perfect storm. Liquefied natural gas prices have plummeted on a combination of demand destruction caused by the COVID-19 pandemic, and fast-growing supply as a wave of new export facilities have added to global capacity.

The extent of the price plunge was brought home to the market last week when trader Vitol sold a cargo of LNG to Gunvor at $2.05/mmbtu for early-June delivery to northern Asia.

That may be the lowest price ever for a spot cargo in the region; in January the price was more than $5/mmbtu. Gas for delivery at the Henry Hub pipeline in the US, the benchmark for most global trade, last traded at around $1.75/mmbtu. Asia prices reflect global price trends, particularly the delivered cost for US cargoes.

For a number of months global gas prices have been in decline as more and more liquefaction capacity has come on line, particularly in the US. In 2017 the United States became a net gas exporter for the first time, and while it’s a long way from challenging Qatar and Australia as the leading shippers of the fuel, American gas looks set to remain a feature of the global market.

Meanwhile, the onset of the pandemic has hit demand. US natural gas inventories have grown by vast increments in recent weeks: in the week ending April 10, the US Energy Information Administration reported a huge 73 Bcf injection into storages.

At the end of March, European gas storages were over 60% full, a record for the time of year due to a combination of a mild winter that dampened consumption, and high deliveries in 2019.

Buyers in Europe and Asia have been delaying deliveries and in many cases trying to cancel cargoes. Spain’s Endesa has cancelled two deliveries from the US in June and July, while buyers in Asia are reportedly in talks to push back cargoes that are scheduled to arrive between now and October.

Purchasers are also exercising options to reduce by as much as 10% the size of cargoes delivered under contract, as import storages reach near maximum capacity.

On the supply side the picture continues to look gloomy, as even more gas deliveries, and export capacity are on the way.

After the US and China signed a trade deal earlier this year, China has lifted its 25% tariff on natural gas, and the US has been quick to take advantage. The first shipment of post-deal LNG arrived in China on April 20, and reports indicate up to six more cargoes may be heading east.

The US federal government has awarded a conditional permit to Pembina Pipeline Corporation’s Jordan Cove LNG terminal in the state of Oregon, though local permission is still pending.

Australia, which last year became the world’s to exporter, is also progressing plans for world-scale import terminals to supply other parts of the country.

Producer Santos has signed a letter of intent to sell a 12.5% stake in its Barossa field to Japan’s JERA, output from which will feed the Darwin LNG terminal project. At the same time, PetroChina and Shell’s Arrow project has taken the final decision to proceed with the Surat Gas Project that will start supplying Australia’s east coast market from 2021.

A number of deals are also under way to develop LNG marketing capacity throughout Asia, which points to a continued healthy supply of gas into the region.

Even as demand is confidently predicted to continue to grow for many years to come, particularly in Asia, the current oversupply is threatening many producers, particularly marginal shale producers in the US.

The low Henry Hub prices have already led to the bankruptcy of one producer – Whiting Petroleum Corporation – and analysts warn that a prolonged period of low prices may drive others into insolvency as well.

With WTI oil prices now down to less than $20/barrel, shale oil and gas producers aren’t covering their costs, and company analysts are confidently predicting a surge in bankruptcies in the coming months. It may take this kind of pain in order to rein in enough supply to rebalance the market.

European carbon allowance (EUA) prices reached in mid-July a 14-year high of €30.80, just 20 cents shy of an all-time record, as buyers flooded into the market and speculators hedged options positions.

Prices have fluctuated between €26 and €29 since then, as some short-term traders took profit, but there seems to be a strong undercurrent of demand supporting EUAs.

The current market is a far cry from the middle of March, when EUA prices plunged 35% amid a general sell-off in energy. The initial bearish reaction to the spread of the virus and to lockdowns around the world reflected concerns that demand for oil, gas and electricity would be negatively affected.

But from mid-May, as the markets absorbed the real scale of the reduction in demand, prices began to recover. Some of this reflected a surge in liquidity as governments rolled out stimulus programmes to bolster their economies. Equity markets in particular rallied strongly, carrying some commodities with them.

Today, both German power and API2 coal prices have recovered to pre-Covid levels, while crude oil and TTF natural gas are around 15% and 10% respectively below their levels in the middle of March.

Carbon, however, kept on climbing and reached an intraday peak of €30.80 on July 13, its highest since April 2006 and just 20 cents short of the all-time record of €31.00.

Why has carbon recovered so strongly?

The fundamentals of supply and demand in 2020 look very weak: the lockdowns, which are only now being relaxed across Europe, are estimated to have cut around 200 million tonnes of demand, equivalent to more than 10% of a year’s total emissions from installations covered by the market.

The answer lies in a combination of speculative trading and long-term optimism. Next year sees the start of the fourth trading phase of the market (ending in 2030), in which supply will be tightened and fewer installations will receive free EUAs to protect them against international competition.

2021 will also bring a review of the functioning of the Market Stability Reserve (MSR), the mechanism that works to maintain a balance between supply and demand. Until now, the MSR has made small inroads into the glut of allowances; while it has removed as much as 400 million surplus EUAs from the market in both 2019 and 2020, demand has fallen sharply as well, as coal plants shut and are replaced by gas-fired power plants that need half as many allowances.

Regulators will begin next year to consider how to beef up the MSR so that it can absorb larger chunks of oversupply and return the market to balance sooner.

On top of this, the EU’s new Green Deal proposal sets ambitious targets to reach net zero carbon emissions by 2050, and this includes a steeper emissions cut by 2030. Lawmakers will be debating whether to raise the current 40% reduction compared to 1990 levels to a 50% or even 55% cut.

Whatever the outcome, there will be an impact on the EU ETS: it’s likely that from 2026 (half way through the fourth trading phase) the cap will be altered to target the 2030 goal and installations will face much tighter EUA supply.

Trading sources suggest that it’s this longer term view that has begun to drive prices higher, looking past the low demand expected in 2020 and focusing instead on the impact of the Green Deal. If this is true, then we may not have seen the end of carbon’s surge.