Unstructured Data: Understanding The Keys To Palantir's Future

Summary

• Palantir is not a data analytics company in the traditional sense, because Palantir's software is designed to do something very different: provide an OS for the modern enterprise.
• Much the same way that the analytics/AI/ML revolution we have lived through was enabled by increasingly better hardware. New hardware breakthroughs will enable Palantir's future.
• It took decades for processing power to catch up to theories of how structured data sets could be used. Now, enterprises are able to handle more and more unstructured data.
• We hypothesize that Palantir is Facebook for the industrial internet, in the sequel to our previous Palantir article. Palantir will enable companies to move AI/data from processes to organizational structure.

Introduction

Imagine the factory of the future. Autonomous trucks are pulling up from a supplier. The supplier's cloud is transferring information about the shipment to the factory's on-site data center. Using that data, picking robots are unloading pallets and boxes. Further down the line, even more robots are scurrying around, sorting and organizing the never-ending stream of parts and supplies into various locations, as yet even more robots pull these materials into the production line.

Blanketing this facility are sensors. Sensors for heat, for foot traffic, for production and quality control. Cameras. Radar. Lidar. GPS. Motion. Pressure. Some sensors are used for local analytics, a robot using its eyes to make sure it doesn't crash into its fellow robots. Some of this data is pooled and organized, sorted for important insights or anomalies, then transferred up to the cloud where it can be analyzed, recorded, and studied.

Somewhere in this facility are humans, doing complex work by hand. A small team of factory workers have noticed a problem. Small plastic rivets they are using keep breaking, costing the factory time and materials. At their workstation, the team leader makes a note on a computer.

Plastic rivets keep breaking.. breaking ~1 out of 6 PN#13345

Elsewhere are more humans, managing the factory. Supervisors and mid-level management are using the vast amounts of data collected to simulate various scenarios. The data center is hard at work crunching the numbers, done remotely in the cloud.

A mid-level supervisor receives the note from the factory floor, flagged by the system to indicate there is a problem. This is a different type of data and complex problem that a computer would not be good at sorting out. With the power of the organization's data at their fingertips, the supervisor is able to immediately look into the issue. The supplier is running the same software, and the two companies collaboratively share some of their data.

The computer has recognized the part number in the note, and with a single click, the mid-level supervisor can see the supplier's data regarding the plastic rivets. He turns to the specifications and notes. Because the tools used by the team on the floor are connected to the network, he can also see that the pressure they are using may be too high for the brittle plastic. The team is alerted in real time that they should reduce the pressure that they are applying. A note is made accessible to other assembly points that might have the similar issues with the same rivets. The issue is resolved quickly and efficiently. These tiny advances in efficiency add up.

Structured & Unstructured Data

This may be a hypothetical example, a made-up bit of futuristic fantasy, but it is the world we are rapidly headed towards. Much of this technology already exists, at companies like Ocado (OTCPK:OCDGF) and Amazon (AMZN), robots are already scurrying around the warehouse, aided by subset of data science known as computer vision. Facebook (FB), Google (GOOGL) (GOOG), and Amazon (AMZN) are all using powerful analytics to understand their customers, and what they might like to see or buy.

Some of this data is structured. What websites you visited and when, prices, or the balances of various accounts are all data that is easily structured. They can be organized, measured, and analyzed with relatively little effort. Computers are unparalleled in their ability to perform these tasks, and today's analytics run at a massive scale.

The most abundant data is unstructured, like documents or images. Computers are getting better at collecting, analyzing, and using this type of data, but we humans still have a tremendous advantage. Our powerful brains are processing units built for unstructured data, giving us the ability to see context or draw connections that would be challenging even for today's most advanced AI models.

So, how do we leverage what computers are good at today, so that we can make the most of our unique abilities? How do we enable the transfer of responsibilities to computers incrementally, as they gain ground in their own ability?

The answer to these questions starts with a centralized platform for data, where data is at the heart of the organization, supporting interdisciplinary collaboration. The business of the future is one that is not solely organized by budgets and organizational charts, but organized around its data. Pervasive. Everywhere. Everything. Augmenting and aiding decisions across the organization.

(Image Source: Palantir)

Though it may seem esoteric, a platform for unstructured data is the first thing required to quantify and manage such complex pools of information. That's exactly the type of software that Palantir (NYSE:PLTR) has built and continues to develop. A core "data OS" where information is shared and analyzed, visible to many different levels of the organization so that they can make quick and timely decisions. This is a radical departure from how most businesses are run today, where things are often siloed or opaque.

Even after 17 years of Palantir, we are still only in the early stages of this transformation. Things are just starting to scale, and this transformation will be enabled and accelerated by breakthroughs in advanced computer hardware.

Understanding the Hardware: Why it matters

Taking unstructured data such as documents, memos, notes, images, videos and so on and putting it onto an open and shared platform can be thought of as pseudo-quantification, in the same way that our brains are pseudo-processors for unstructured data. Evolutions in hardware are expanding software's ability to quantify this type of data, matching and exceeding our own capabilities.

Today, computers are mastering images (computer vision) and tomorrow's computers will become ever more sophisticated in their ability to read and write documents, memos, and so on. Palantir's software will enable businesses to leverage this technology.

So, let's expand on the hardware side, because it is critical to understanding this profound macro trend. Long before computers were doing analysis on structured data, scientists and researchers were developing theories and models on machine learning and artificial intelligence. It took around 20 years for the hardware required to catch up to the theories that had been developed. That convergence of theory and capability literally created Google, Facebook, and Amazon, companies whose business models revolve in large part around the ability to train algorithms on large amounts of data, and training them to do things humans could never do on their own.

The central tenet of this is hardware. The processors, the storage, the datacenter hardware, created the need for software the ecosystems that are layered on top. So, what is so special about today? Why is this moment so unique?

The hardware capability is always the bottleneck to progress. Once you have hardware that can do it, you can program it. There have been massive breakthroughs in hardware in recent years, and this doesn't seem to be slowing down. Technologies such as EUV are enabling us to push the boundaries of Moore's Law. More importantly, there is the emerging field of acceleration hardware. These are new architectures and designs being developed that are incredibly more efficient at doing artificial intelligence and machine learning (AI/ML).

AI/ML is software written by computers that is coincidentally pretty good at handling unstructured data. It will only get better as the hardware capability expands. To some readers, this is no secret. They've seen the epic rise of companies such as Nvidia (NVDA), AMD (AMD), and Xilinx (XLNX). Perhaps, you may have even read about companies such as Cerebras, which has created the world's largest chip, or Graphcore, a company that has created an architecture inspired by the design of the human brain.

(Image source: Nvidia)

Better hardware enables more powerful AI/ML software, more powerful AI/ML software drives the need for Palantir's data OS, because most data is unstructured. It drives the capability for companies to operate and manage their businesses based on data, not just on the analytics side gathering business intelligence, or in some siloed process, but in a holistic and centralized way.

If every factory operated like the factory of the future previously described, a sophisticated and coordinated cyber offensive might grind an entire economy to a halt. With so many connected devices and multiple data centers, imagine all of the entry points an attacker could use to target a network. Alibaba (BABA) is secretly working on such a factory concept, for example, but China's internet features a robust firewall where traffic is tightly monitored and controlled. Security in the open source and collaborative western world will require a wholistic platform. More on that later.

Palantir is more like Facebook than Tableau

So, at this point, it is probably easier to see why comparisons to Google and Facebook are far more accurate than comparisons to Tableau. Google was the halcyon to emerge from the convergence of 1980's AI/ML research and 2000's computing capability, but Facebook was truly the pioneer in unstructured data.

Users on Facebook experience a convenient platform that enables them to quantify and organize their social lives. They can measure how many friends they have and then calculate how popular they are (structured datasets of likes and friends lists). But they spend most of their time sharing and managing unstructured data such as comments, messages, photos, videos, and audio.

This platform was powerful enough to transform society (we will reserve our opinions on whether it was for the better or not). It put data at the center of our lives because it gave us an intuitive platform to handle both unstructured and structured data in a meaningful and organized way.

Steadily, the power of AI and ML (enabled by steady advances in hardware) has been embedded into Facebook's platform. Today, Facebook can recognize faces and draw conclusions about friend groups and personalities. It can curate content to hook users, driving engagement. It powers chatbots where companies direct their customer service. It records audio, scanning conversations for clues as to what people might be interested in buying, then curates advertising accordingly.

(Image source: Density, a ceiling mounted sensor for detecting foot traffic in a room. One of many new innovations creating IoT data.)

So, how big can Palantir get? Think Google or Facebook, because it is essentially Facebook for enterprise. Everyone who has a job has information that they need to find, analyze, or share. It's not hard to imagine a world where Palantir achieves similar ubiquity. Even a local restaurant may find it useful to have radar sensors in the ceiling to collect floor traffic data. A server is alerted that customers at a particular table have been waiting too long or perhaps that data is shared with the local neighborhood commission to improve tourism in the area. It's certainly how today's businesses will be run tomorrow if they want to be efficient and competitive.

That technology exists today, and it will only getter better, cheaper, and more connected. The future of the internet is enterprise and industrial innovation, rather than more abundant consumer-oriented content. Palantir is Facebook for the industrial internet. Furthermore, technologies such as 5G and IoT will have a similar impact on Palantir as smartphones and highspeed cellular data had on Facebook's proliferation.

Geopolitics: The Catalyst for Rapid Adoption

(Satya Nadella, Eric Schmidt, Alex Karp arrive at the 2019 Bilderberg conference in Montreux, Switzerland. Presumably, the intersection of the economy, technology, and geopolitics were top of mind. Image Source: Newsweek)

As discussed in our previous article, geopolitical concerns are a catalyst for the rapid adoption of Palantir. Today, we think of the internet as a largely benevolent and peaceful place, with only the "dark web" lurking in some shadowy corner. But, if you've received the iOS-14 update limiting the access that applications have to your phone/iCloud data, you have seen firsthand that defensive fortifications are rising.

Stuxnet seems like a distant memory now, but to us, it is an omen of the dangers we may face in a new era. The Stuxnet virus, created by the US government and used to covertly destroy centrifuges in the Iranian nuclear program, escaped containment and infected industrial equipment in various companies, all over the world. (Great documentary on it here.)

It was the need to supply a complete stack of software to an enterprise. [It] was built on the bias that we believed that crisis in the western world would be so extreme that the Frankenstein monster way of servicing an enterprise with a 50 different products that have to be sewn together by a myriad of engineers, none of whom particularly are coordinated, but on the back of power points that take years to develop… and a cost structure that can go into the hundreds of millions would not work in a time of macro either economic or political crisis.

Of course, we didn't know that Covid would come, but we were betting that there would be economic and political conditions that meant the old way of building the Frankenstein monster internally, or purchasing 50 products that are supposed to be sewn together according to a PowerPoint generated by a non-technical consultant, would not be efficacious when your business would either live or die based on whether you can you can rebuild your enterprise stack overnight and in our case within hours. - CEO Alex Karp, Palantir Investor Day

Should geopolitical tensions continue to rise, the internet of the future may look very different than the one we have today. Every now and then, we see stories of some cyber-thieves infiltrating a company or government agency, but these rarely take on the sophistication that a nation - state is capable of. The warfare of the future may be one that is fought in large part over the internet, and companies/industries may be primary targets.

Palantir is one of the answers to this challenge, given that its software was originally designed and built for intelligence agencies and military purposes, which are the target of more sophisticated attacks than the average corporation. This may seem like fearmongering, but a proactive approach to adopting a highly secure platform would discourage such a scenario from ever occurring in the first place.

Conclusion

So, in conclusion, we turn to George Soros. Soros said recently that he will sell Palantir when his lockup expires, "because does not approve of Palantir's business practices" and that he "made this investment at a time when the negative social consequences of big data were less understood". Soros also owns Google, which he could sell tomorrow. Furthermore, he thinks that China, which uses a vast network of data collection to assign its citizens "social credit" scores, should lead play a bigger role in a new financial "world order". Quite ironic.

(Palantir Revenue, Image Source: Palantir Investor Day)

The thinly veiled, highly political, smear campaign against Palantir seems like a joke at this point. We hope we've explained why. And it's been fun and exciting to watch short interest that bought into the media narrative get absolutely torched, whimpering about it along the way (alternate link). Especially as Palantir has continued to pick up new government contracts, developing technology that it can leverage as it steadily continues to pivot towards the enterprise.

The long-term prospects of Palantir are quite solid. We are long Palantir.

Disclosure: I am/we are long PLTR. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.