The Death of An Era and A New Renaissance In Computing
What are we trying to do as investors? We are looking for things that are small, which will grow to be large. We look for things that will be needed in the future that are often overlooked in the present. In a way, we are trying to be seers, prophets, and oracles— each with our own set of philosophies and world views. Many of us try but fall by our own swords.
Some of the most crucial aspects of this process include evaluating where we have been, where we are, and where we are heading. We need to properly vet what kind of teams and projects have the fortitude to withstand the sands of time and not misstep. Any miscalculation can lead to failure. Constant adjustments and course corrections are necessary.
This process is something that takes time. We learn from our successes, and perhaps we learn even more from our failures. Understanding history and updating our roadmaps to compensate for the changes that are happening in the present is of utmost importance.
I have quoted Aldous Huxley many times in my writings. Huxley was a techno-prophet that foresaw the environment we are currently living in.
Huxley feared that there would be no reason to ban books, for there would be no one who wanted to read one. He feared those who would give us so much that we would be reduced to passivity and egoism. Huxley feared the truth would be drowned in a sea of irrelevance and that we would become a trivial culture, preoccupied with nonsense.
I have written extensively on how our current social media landscape is destroying our information streams. So to Huxley's point, how do we know what is relevant? True? Or trivial? Do we really know what’s going on?
We need an accurate view of what is actually happening now. Marshall McLuhan, another techno-prophet of the 1960’s, warned us decades ago when he said: “We look at the present through a rear view mirror. We march backwards into the future.”
Wayne Gretzky, the famous hockey player, is famous for a very philosophical quote he gave to a reporter after playing a stellar game: “I skate to where the puck is going, not where it has been.”
Gretzky’s skill is almost unmatched to this day. His ability to read and interpret plays that the other team was executing was on the level of prophetic. Quick slap shots across the ice, followed by fake outs, all while players are trying to run you down, only to find the player you thought had the puck, didn’t. Not Gretzky. He saw the moves, the fakes, the sight of hands— and he knew where to skate to be in the perfect position to strike.
Anticipating the actions of others, working with good information and forecasting future trends aren’t the only factors. You, the observer, are part of the equation— and what makes up “you” is constantly changing. We can sometimes be our own worst enemy. Our vision can become cloudy and distorted and our fluctuating mental states can paint different pictures.
Heraclitus, an Ancient Greek philosopher said: “Panta Rhei” which means “everything flows” and nothing ever stays the same. Given enough time, all things eventually turn to their opposite, meaning that everything is a constant becoming and changing. We are changing along with everything else.
Socrates commented on this statement from Heraclitus: “Heraclitus says, you know, that all things move and nothing remains still, and he likens the universe to the current of a river, saying that you cannot step twice into the same stream.”
It is not just that you cannot step twice into the same river because the river is constantly changing. Even if the river remained static, there would never be two moments where you would be the same. We are not impartial observers to the universal process.
So what is going on? How are we changing? How is the river changing as we wade through its currents? I have some ideas that I have been pondering that I believe will be helpful on this quest of ours. Take a breath and follow me down to the deep end.
———
In my last article, The Internet Before the Internet and a History of Sharing Central Supercomputers, I gave you a birds eye view of the history of computers and how the invention of time-sharing enabled us to do more with the limited hardware capabilities before the advent of the personal computer.
Here’s a brief synopsis to refresh your memory:
Time-sharing played a crucial role in the development of a precursor to the Internet called SAGE.
Time-sharing went on to be an integral feature of the modern Internet.
Even though Moore’s Law brought down the cost of hardware and improved performance to the point where everyone had a supercomputer in their pocket, the architecture of the Internet is virtually the same as it was in the early days.
We still connect to shared supercomputers via personal terminals like we did in the 50’s.
Today's supercomputers are called data centers.
Data centers are multi billion dollar concrete warehouses with massive energy, cooling and security needs.
Data center are just a bunch of servers that we now call “The Cloud”
The Cloud is just someone else’s computer.
There is a metaphorical arms race between companies and nation states on who can build the biggest data center (supercomputer).
So why does this all matter? Why am I telling you all this and why have I laid out all this groundwork in my previous post?
I wanted to give you a primer on Moore’s Law and how computers have developed over the last several decades. I wanted to demonstrate how their rapid advancement played a role in how the Internet was developed. I wanted to show how the architecture of these systems hasn’t fundamentally changed since the 50s, despite having faster and cheaper hardware as a result of Moore’s Law. We still use time-sharing concepts and we still connect to central “supercomputers” which we now call data centers.
Moore’s Law Is Dead
Moore’s Law has allowed us to enjoy an amazing ride for nearly 80 years. But there’s a huge problem that we have recently started feeling the consequences of— Moore’s Law is dead. It died almost 13 years ago and we’ve been scrambling to compensate ever since.
I’m not alone in this observation. I have another one of these techno-profits on my side. Let me introduce you to John Hennessy. John Hennessy serves as Chairman of Alphabet Inc. (Parent company of Google), he was the 10th President of Stanford University and previously a professor of electrical engineering and computer science. He won the Turing Award, which is the equivalent of the Nobel Prize for pioneering a systematic and quantitative approach to the design and evaluation of computer architectures with enduring impact on the microprocessor industry. He is often referred to as The Godfather of Silicon Valley.
John has been trying to warn us for more than 5 years that Moore’s Law is dead, but many have ignored it.
Industries have set their watches to this predictable increase. In the past, when we peer into the future we have always used Moore’s Law as a compass to navigate us forward. We have assumed hardware would be delivered to us at this exponential pace so we have all used this knowledge to design our systems, write our code, and forecast future growth. No more.
As Gordon Moore, co-founder of Intel and author of Moore’s Law, once said: “no exponential is forever.” Today, the imminent end of Moore’s Law scaling represents a major turning point in history.
This doesn’t mean we will no longer see improvements. On the contrary,but we need to get used to the idea that this rate of doubling won’t be seen anytime soon. Improvements will come, but they will be incremental. Exponential improvement may come, but it will be sporadic and unpredictable. Many product advertisements will be shrouded in marketing claims that will make it seem more advanced than what is real.
Also, it is important to understand that Moore’s Law wasn’t an actual physical law of the universe that was guaranteed to continue. Moore’s Law, as formulated by Gordon Moore in his 1965 paper “Cramming More Components onto Integrated Circuits,” is nothing more than an empirical observation that the density transistors in an integrated circuit (silicon) doubles every 18-24 months.
It’s a simple concept: The engineering problem behind Moore’s law is reducing the barrier between transistors. The smaller you can make them, and the closer you can fit them together without touching— the more you can fit on a piece of silicon. Engineers have estimated by 2025 we will get to about 2-3 atoms in proximity to each transistor. Any closer and that’s when we would need to transition into the quantum mechanics realm— and this isn’t ready despite the headlines we see on the Internet. We are still likely decades away, and even then we aren’t going to get them in our phones.
Improvements will come, but not in the realm of transistor density. New and creative solutions will have to be invented for us to get past this bottleneck. The death of this law means we won’t be getting any more free rides. And as you can see over the past 80 years, we’ve been spoiled. We’ve gotten lazy with our architectural design of these systems. We kept doing the same thing over and over again and got away with it.
Do not fear. Problems are what we look for as investors. Problems require solutions, and solutions present opportunity. These problems will propel us into a new renaissance or golden age of computing. We are going to start seeing extremely creative ways of doing things.
But first, I need to inform you of another recently deceased law.
Dennard Scaling Is Dead
Just like with Moore’s Law, we see a similar phenomena with power consumption about every 2 years. If the transistor density doubles, power consumption (with twice the number of transistors) stays the same. As a result, we are able to gain a power efficiency increase as our processors get faster. It was now also cheaper from an energy perspective— a pretty sweet deal.
But just like with Moore’s Law, this law died right around the same time in 2007. The primary reason cited for the breakdown is that at small sizes, electrical current leakage poses greater challenges and also causes the chip to heat up, which creates a threat of thermal runaway and therefore further increases energy costs.
The breakdown of Dennard scaling and the resulting inability to increase processor speeds without overheating caused most CPU manufacturers to focus on multicore processors as an alternative way to improve performance. A multi-core processor is a microprocessor on a single integrated circuit with two or more separate processing units, called cores, each of which reads and executes program instructions.
So why don’t we just add more cores if we want to speed up the processor? The problem we run into is a multi-core approach requires greater energy consumption and these processors become hot while accomplishing more work.
Another downside with the multi-core approach is that switching between each of these cores worsens CPU power dissipation, which results in a fraction of an integrated circuit that can actually be active at any given point in time without violating power constraints. The remaining (inactive) area is referred to as dark silicon. Dark silicon essentially means that all the transistors we tried to cram into the chip are now unusable.
So now we are running into some very tricky conundrums. We can’t cram more transistors onto the chips due to the death of Moore’s Law. Energy efficiency is declining and heat is increasing due to the death of Dennard scaling. Multi-core can’t continue without overheating the chips and resulting in unusable transistors with dark silicon spots.
The Problem
The death of these laws is causing some pretty serious problems that will require very creative solutions to get us through these bottlenecks.
We can’t make them faster. They require more energy than ever before and they get too hot.
Remember in my last article when I talked about these massive data centers that house hundreds of thousands of computers? This is where the death of these laws are creating the most disruption.
In data centers, the second most costly thing is the electrical and cooling infrastructure, as well as the energy needed to cool the machines and run the computers. Data centers have turned into power plants, and the heat generated from the computers requires massive cooling infrastructure to ensure they won’t blow up. This infrastructure requires even more power to operate, resulting in an ever greater need for cooling. See how this death spiral works?
The techniques we have been using for decades in computer architecture have hit a wall with respect to their energy scaling and speed limits.
We have now entered into an environment of diminishing returns for data centers. The law of diminishing returns states that increasing a factor of production by one unit will at some point return a lower unit of output per incremental unit of input. The law of diminishing returns does not cause a decrease in overall production capabilities, rather it defines a point on a production curve where producing an additional unit of output results in a loss and is known as a negative return. Under diminishing returns, output remains positive, however productivity and efficiency decrease.
Data centers that make up our “cloud” or Internet are reaching their capacity. They can’t keep up with the data that us users generate. Computational costs are increasing and they are constantly battling these heat and energy problems, not to mention their increasing negative environmental impact.
The way these data centers operate are much like how utility companies operate. You use their cloud and they measure how many resources you use. Things like power, bandwidth, storage, etc. They add up everything you use and send you a bill.
Since data centers have entered an environment of diminishing returns, they must pass the cost onto you to compensate for these problems. And the costs are rising each year. On top of these issues, there is another factor contributing to an increase in prices.
I have explained the business model of how publicly traded companies operate. In short, the goal is to post higher and higher revenue each and every quarter, year after year to ensure the stock price increases. If revenue decreases over time, the company’s board of directors is very likely to find a replacement CEO and/or management team to fix this problem. Because of this business model, it is almost unthinkable to consider reducing the price of cloud computation and storage for customers and users.
The problem landscape continues to get trickier.
Side Effect: Software In Decline
We’ve been blessed with decades of seeming unlimited speed, power and memory all while costs kept decreasing. As a result, this has caused an enormous amount of “bloat” in software architecture. Programmers of the 50s, 60s, and 70s had very limited constraints in what the hardware could do. And as we now have seen, computer science and mathematical wizards like John McCarty had to invent extremely creative ways to do more with less— things like time-sharing and inventing extremely efficient programming languages.
MIT professor Charles Leiserson in the 2018 “Engineering of Software Systems” course opened with the following remarks: “It's interesting these days, most software courses don't bother to talk about these things. And the reason is because as much as possible people have been insulated in writing their software from performance considerations. But if you want to write fast code, you have to know what is going on underneath so you can exploit the strengths of the architecture. And the interface, the best interface, that we have is the assembly language.”
Over the last several decades, what has happened to the software world? Software was getting free improvements without making lots of efficiency improvements. You could afford to write bloated code and use slow programming languages because hardware improvements would mitigate that.
Every two years programmers were blessed with massive speed and storage upgrades. New programming languages emerged like the scripting languages. Developers had so much speed at their fingertips that they no longer had to write software that was optimized for speed and resource usage— this wasn’t the case in the early days of John McCarthy.
Even the modern web is extremely bloated. In the early days of the Internet, the only code that would run on a webpage was HTML and CSS. Now it’s virtually impossible to visit a site that isn’t bloated with dozens of scripts running concurrently while using cookies and collecting data on what you are doing. The web browsers themselves have also followed this path.
We can’t get away with bloated software anymore. Speed and efficiency increases are going to happen through other creative solutions, instead of by gaining exponential growth through the hardware itself. Moore’s law has given us a free ride within the existing paradigms for more than 7 decades— and that free ride is coming to an end. We are entering a wild and messy disruptive time.
We are going to have to get creative. New ways of wiring code that allows our hardware to do more. We need to experiment with completely different computer architectures on both the hardware and software domains.
Software Solutions
Programmers have performed tests using different programming languages that are running the same code to calculate test problems and determine the time it takes to get an output.
In one of the tests programmers used a modern language like Python to see how long it would take to calculate a problem. Then, they rewrote it in C (developed in 1978) and it ran 47 times faster than it did in Python.
Then they introduced something called parallel loops which is another software solution that can increase speed and efficiency of the programs. This added another 30x improvement. Then they introduced memory optimization techniques and got even more factors of improvement. And finally, they got down to the architecture level— the closest you can get to the processor— and implemented vector instructions in the Intel chip architecture. The final result— 62,000 times faster. The lesson to be learned is that designing effectively from the architecture layers of the computer to the programming and software layers of the system can bring massive improvement.
We will need more solutions like this going forward. The folks at Apple recognized this need.
Apple’s Switch to M1
Applying this new lens of understanding can give us tremendous insight on a strategic move Apple made a few years ago.
Apple recently switched to using ARM processors. Previous to this switch they used Intel processors which have been the dominant player in the CPU space for decades.
At first glance, you may be thinking that this was just switching one physical piece of hardware for another, right? It’s true that they did switch hardware, but what this switch meant was that Apple could now change the architecture, or software, that the processor used to run its processes. They tackled the problem at the software layer, just like the example I gave above where programmers increased the performance exponentially through implementing elegant code architecture.
The difference between Intel’s x86 architecture and Apple’s ARM architecture is in the instruction set the software uses to talk to the processor X86 uses complex instruction set computing (CISC)— which means it executes complex instructions over several clock cycles all at once. ARM uses reduced instruction set computing (RISC)— which means it executes a single instruction per clock cycle.
CISC puts the burden on hardware to perform and RISC puts the burden on the software to perform. Switching to an ARM processor allowed Apple to take the software optimization approach to achieve the performance they wanted.
Apple saw the winds change and realized sticking with traditional chip technology wasn’t going to get them the continued improvement that they were seeking. They knew Moore’s law and Dennard scaling were coming to an end, and they knew that their current partnership with Intel wouldn’t take them much further technically— and would increase their costs at the same time.
Intel has been significantly behind its projections for getting more performance out of its processors due to the limits of Moore’s law and it was failing to catch up.
Apple started making their own silicon chips using this completely different architecture. Making the chips themselves gave them complete control to optimize and change any aspect of the chips they wanted— both on the hardware and software levels. This allowed them to have full control over the software architecture of the processor. They could now optimize every piece of code in the stack. Because of this more unified integration they have been able to achieve the higher performance they sought.
ARM processors using this RISC architecture have a lower cost than x86 processors. They have minimal power consumption and lower heat generation— all because of a more elegant approach to how the code interacts with the hardware. ARM processors are desirable for light, portable, battery-powered devices, including smartphones, laptops and tablet computers, and other embedded systems. They are also used for desktops and servers, including the world's fastest supercomputer.
Even though Apple was pushing up against the same level of transistor density of the chips that other companies were, they used their creativity and innovation to better control everything else within their limits.
Apples M1 represents the largest performance boost in computers in the last 13 years— since the death of Moore’s law.
Apple is now on track to be the next dominant silicon chip maker because of their innovation— and It will spark even more competition.
The Data Center Problem
Now that we know the world is pushing up against the limits of Moore’s Law and Dennard Scaling, we know that solutions are required if we are to overcome these bottlenecks.
We’ve now seen a brilliant strategy that Apple implemented to solve this problem. Let’s now go back to our data center friends who run the cloud. Data centers are experiencing these challenges at an ever greater scale because it’s not just a single computer— data centers are enormous buildings full of hundreds of thousands of computers.
Energy demand, heat problems and computation limits are all factors in the diminishing returns they face.
Data centers have become emblems of grandiose human achievement in scale and size. They pride themselves in the complexity and size of the operation, how many square feet in a building, gallons of water consumed, energy consumption, output, etc. They have essentially become massive power plants where the entire output is directed into running its internal process and cooling. The industry is expected to consume 20% of the world’s energy supply by 2025.
In addition to these limits they are facing, there is another problem that they have created for themselves. As the data center market grows, these operations are becoming owned by fewer and fewer organizations. Smaller and more efficient data centers are shutting down because they cannot compete with the larger players.
Reducing the number of data centers that are distributed across the globe make the few large centers targets for security concerns. Natural disasters causing disruptions at these locations can put this cloud computing at risk, including the data we have stored there.
In Greek mythology, there is a story of the Hydra— a great and terrible water dragon with many heads. If you cut off one of its heads, two more would grow in its place, making it an almost impossible creature to defeat. If we relate this story to the way data centers operate, then we can see that this strategy is vulnerable to failure. If one of these centers goes down, it’s going to take billions of dollars and massive resource allocation to get it back online. They can’t easily regenerate two heads to replace the one that was cut off.
One of the biggest data centers in the world was created by the NSA in Utah. The finished structure is characterized as a Tier III Data Center, with over a million square feet of space that cost over $1.5 billion to build. Toward the end of the project's construction it was plagued by electrical problems in the form of massive power surges that damaged the equipment. This delayed its opening by a year and took several years to complete— and this isn’t even factoring in the time it took for the planning and approval process before construction even began.
The Solution
Here is where our next project comes into the picture. Let me introduce you to Storj - Fast and secure cloud storage at a fraction of the cost. Just like our innovators above, Storj has taken a software approach to the hardware limitations that plague centralized data centers.
The idea is simple and it takes John McCarthy’s idea of time sharing to the next level. Storj allows anyone to share their computers’ resources and use them to build the biggest, most powerful, and most decentralized data center the world has ever seen. It unites the computing and storage of any computer and brings it into a global network that anyone can participate in.
Here’s how it works. Storj first encrypts your data ensuring that no other participant in the network has access to it— keeping it safe and secure. Then, the encrypted data is split into pieces that can't be distinguished from any other object’s pieces.
Each piece of an object is distributed over a massive global network of nodes so data is never in just one place—it's all over the world with better security and privacy.
When you retrieve an object, only 29 of its 80 pieces are needed to reconstitute that object. With no central point of failure, your data is always quickly available, all over the world.
One node going offline won't impact any files. Their network's automatic repair process reacts when too many pieces for any files are lost, and repairs them within a very healthy margin of safety.
We are now looking at a system that looks like our mythical Hydra.
At the core of the Storj lies the STORJ token. This token incentivizes all participation in the network. Anyone that wants to offer their storage and bandwidth the global cloud gets rewarded with STORJ.
By allowing anyone to participate in the network, it allows actual data centers to monetize their operation in ways they previously could not. Earlier I wrote about how the centralized top tier data centers were causing some of the small and medium operations to go out of business. Storj enables these operations to now monetize their unused storage and bandwidth. Having this second stream of income sweetens the economics for these businesses.
Even individuals with computers in their basements or attics can participate and gain passive income from their unused resources. Imagine how many untapped extra resources that are just waiting to be connected to the Storj network.
By decentralizing the data center, it reduces the problems that the large players incur such as heating and cooling.
Because these data centers operate like a utility company, customers can spend hundreds of thousands of dollars a month on these cloud services. Cost is a huge factor for businesses and individuals that are hosting extremely popular apps and websites from a service like Amazon AWS. This is where Storj really shines.
Because Storj uses the existing infrastructure and unused machine resources, it decreases the cost of customers who are looking to purchase cloud servers. As a result, it’s a fraction of the cost compared to a centralized service like Amazon’s AWS. Storj’s cost is 10% of what a centralized operator charges. In other words, you get a 90% discount by switching to a decentralized model. This alone is a major strategic advantage.
Storj is a major breakthrough in distributed computer systems that will unlock new opportunities to build upon.
At the time of writing this, Storj currently has a market cap of $107 million. It can be purchased on Coinbase and most decentralized exchanges under the token symbol $STORJ
For more information you can visit the Storj website at:
https://www.storj.io/
A Renaissance In Computing
As we get further from the death of Moore’s law and Dennard scaling, we will see the effects manifest themselves in an even greater way.
We can now use this new lens to look at the investment landscape and see opportunities we previously could not. We can also use it to warn us of potential problems a business will run into.
Creative waves of disruption are coming. We need brilliant architects, programmers and visionaries to find solutions like Apple and Storj have provided us.
We need to cut out the bloat and get back to basics if we want to continue on this path of exponential growth.