Click to learn more about author Samuel Bocetta.

Days after the highly-anticipated Wall Street debut of Uber Technologies, technology and finance analysts were still trying to determine what went wrong with the initial public offering.

On IPO day, Uber shares failed to capture the attention of institutional investors; instead of reaching the $49 price target, Uber stock closed below $42, and the subsequent trading days were equally terrible.

Granted, extenuating circumstances prevented Uber from reaching full valuation, case in point: the global trade war between the United States and China, but there is more to consider when evaluating the company’s future profits, particularly in terms of artificial intelligence and autonomous driving.

Profit Center or Cash Hemorrhage?

There is an uncomfortable truth about Uber, Lyft and other personal transportation companies that combine mobile technology with massive data centers and independent contractors working for pennies on the dollar. Finance analysts and economists have applied various models to determine the profit potential of these companies, and their findings have been disheartening.

When investment banking firm Morgan Stanley took over stock underwriting duties for Uber last year, the $120 billion valuation of the company raised some eyebrows because it was mostly based on the excitement the IPO would bring; beyond this hype, there are ride-sharing economics at play, and drivers take up 80 percent of the cost.

Under this business model, Uber may never become a profitable company, but what if there was a way to bring ride-sharing costs down 50 percent? This is where self-driving cars, AI and massive data centers come into play.

The prospect of robotaxis makes investors dizzy with excitement, and this is what Uber has pledged to continue working on. It should be noted that billionaire entrepreneur Elon Musk, founder of Tesla Motors and SpaceX, believes that he can beat Uber, Lyft and Google to the robotaxi future by forming an autonomous driving firm that would be valued at $500 billion.

Although the timeline for full autonomous driving is uncertain, the major players are locked into a race against time to achieve this goal. In a strange twist, the Uber driver strikes that took place in Los Angeles, New York, San Francisco, and Tokyo just before IPO day is exactly what the company hopes to eliminate with the advent of self-driving taxis.

Autonomous Driving and the Constraints on Data Centers

The technology impacts created by the race to deliver a functional robotaxi service will be largely felt at the data center level. The sheer complexity of self-driving car systems poses a great data management challenge: it is estimated that a single autonomous driving vehicle will generate about 300 terabytes of data per year, equivalent to the data encoded within 300,000 Blu-ray movies. While each robotaxi will be equipped with its own supercomputer, a significant portion of data will be managed by remote data centers.

The technology impacts created by the race to deliver a functional robotaxi service will be largely felt at the data center level. The sheer complexity of self-driving car systems poses a great data management challenge: it is estimated that a single autonomous driving vehicle will generate about 300 terabytes of data per year, equivalent to the data encoded within 300,000 Blu-ray movies. While each robotaxi will be equipped with its own supercomputer, a significant portion of data will be managed by remote data centers.

The future of telematics involves reading hundreds of sensors per vehicle on the road and letting AI constructs handle development, management, maintenance, and improvement.

As this evolves, imagine the current complexity of commercial and military air traffic control, not only in keeping the parts moving but deploying adequate countermeasures that go beyond what a commercial security software suite and a virtual private network (VPN) can provide to deter determined hackers bent on nationwide mischief.

Multiply this current complexity by several factors of magnitude, and you can get an idea of the challenges that will be placed on operating and protecting data centers once Uber, Lyft, Tesla, or Google launch the first robotaxi service.

The Future of Global Data Centers Will be Defined by AI Demands

The current landscape of data centers is largely regional; the rise of AI will force a pivot towards global data centers built with edge computing architecture. CPUs and GPUs will no longer be sufficient to handle robotaxi fleets, let alone all the other self-driving vehicles on the road, which need to be handled by supercomputers.

The transition to edge computing data centers is already underway: Microsoft’s Project Natick, which places underwater data centers off the Scottish coast, is an example of the data center future.

In addition to computational power, data storage and transmission, we also have electricity and cooling factors to consider, which explains why Microsoft is deploying data center at the bottom of frigid ocean basins.

We have already heard numerous concerns about the economic inefficiency of cryptocurrencies as they relate to electricity consumption and cooling; naturally, this is not something we can allow to happen with autonomous driving. Thankfully, the current demands placed on data centers by AI and machine learning will eventually become the key to efficiency.

Smarter and More Efficient Data Centers

We have to admit that AI has not been very good at solving human problems, and we should not expect it to do so because we are not very good at it either. What we have learned, however, is that AI constructs are very good at improving themselves. Google has already trained AI to improve energy efficiency, and the most promising result has been a 40 percent reduction of electricity consumption.

You know those mechanical tasks we perform to save electricity at home? Turning off lights when they are not use, installing skylights on the roof, promoting air circulation with fans to ease demand on A/C systems; these are all functions that AI can do a lot better through machine learning.

As for optimizing servers, protecting data storage, safeguarding access, and maintaining data integrity, AI routines are also better than humans; we just need to keep figuring out ways to code them so that they learn to improve themselves. We have already done this through analytics for the benefit of small business owners, and the next step will be to pass on monetary savings to consumers.

One area where consumers should see financial benefit is in web hosting, an industry where competition among low-cost web hosts already applies downward pressure on price. As AI-tuned data centers start to realize more cost savings in the way of improved efficiency related to lower electricity consumption, better storage, more bandwidth, and greater uptime, expect that the holdout hosting companies at higher price points will be able to charge less and maintain the same profit percentage.

The Bottom Line

From the current perspective, many of the technological complexities that we envision for our future will likely be solved by AI and machine learning. The ability of data centers to hold and manipulate the tsunami of data flowing through them is just the first step.

And while other problems will be created in the process, humanity has shown a remarkable ability to lurch forward into progress no matter the obstacle placed before us. Despite movies like Terminator 2, where the AI-powered machines became self-aware and attempted to annihilate the human race – something definitely to be avoided – here’s to hoping the reality of an AI future is more rosy.