The tech market finds itself in heavy weather these days due to adversities like deteriorating trust, market saturation, disappointing earnings and technology being increasingly politicized. In an attempt to recalibrate, we will have a closer look at some of the fundamental trajectories within the tech industry across a few layers of the tech stack, ranging from infrastructure to the interface. As we will see, many of the developments are not isolated but related to each other due to the interdependent nature of these technologies.
An important determinant for the overall development of digital tech are the limitations of its physical computational infrastructure, including computation, storage and connectivity. We find ourselves at the beginning of the installation phase of the 5G network, which will not only offer higher speeds, but also a lower latency and higher device density, thereby enabling all kinds of new applications that were not possible before. The enormous influx of data (44 zettabytes in 2020 according to IDC) created by the billions of connected devices (31.4 billion by 2023) will drive demand for storage and specialized computing hardware. Furthermore, to facilitate increased demands for low latency and processing these technologies will not only be deployed in the cloud or on the edge of the network but on intermediary computation infrastructure known as fog computing. However, an important intervening variable in the roll out of the nfrastructure is the semiconductor industry’s expectation that Moore’s Law will run out of steam in the coming years. This has caused the industry to throw off its self-imposed shackles of simply increasing transistor density. Instead, with the ‘More than Moore roadmap’ we see more diversification in ways to increase computing power and solve a broader variety of problems (e.g. energy efficiency, specialized computational tasks, miniaturization, calculations/ dollar). First of all, material providers could find new opportunities, as the need for other types of materials such as graphene, indium, arsenide, high-k metal gate, silicon carbide, gallium nitride on silicon, and gallium oxide increase. Furthermore, we can see the emergence of alternative chip construction techniques, such as chiplets, FinFETS, GAA structures and 3D chip stacking techniques (e.g. 3D NAND, foveros). Interestingly, the chiplet approach also seems to enable ope-source hardware development which could help save cost, shorten time-to-market and draw on more innovation power. On the level of chips designs, as expected, SoCs, FPGAs and ASICs have gained popularity in solving more specialized computational tasks in the domain of AI, cryptography, 3D rendering and data analysis. Many of these designs and techniques will also rely on new chip manufacturing techniques. This year will see the first chips that have been manufactured with TSMC’s extreme ultraviolet lithography (EUV) enabling 7nm in chip designs, resulting in faster, smaller and more energy efficient chips. Lastly, in order to combat increasing R&D costs, deflating prices and consumer market saturation, we also see a shift in business model going from “silicon to services”, such as end-to-end IoT security solutions and platform-as-a-service. All in all, the heterogeneity in the chip market has increased competitiveness among chip manufacturers, as exemplified by AMD’s surprising comeback this year.
At the protocol level of the stack, we have seen that the crypto and blockchain space has gone through rough weather with a drop of 80% in value due to a shady ICO space, stricter regulation, market manipulation and unhinged speculation. Consequently, the semiconductor industry has been affected due to a drop in demand in ASIC mining hardware. In turn, the drop in hashing power made some cryptos even vulnerable resulting in a 51% attack”. However, under the surface it seems progress has not halted within the crypto space as can been seen by the ongoing growth in development activity, big tech’s involvement, crypto adoption doubling and few long awaited projects going live. More importantly, most of the hurdles that currently seem to limit the adoption and application of these technologies (scalability, energy efficiency, decentralization, governance, security, privacy) have attracted the attention of a wide group of developers with plans to solve them. Substantial breakthroughs on any of these issues could lead to a renewed interest in cryptocurrencies, which is nothing new looking at the 8 crashes in the past 10 years. If this space overcomes its adversities and is able to mature, it will have considerable consequences for the way we organize the remaining layers of the tech stack and how value is allocated. At the level of cloud infrastructure, we see that the evolution is mainly driven by interoperability between multiple clouds (whether private or public) allowing businesses to combine features from different vendors and/or augmenting legacy systems, enabled by containers and microservices. Furthermore, driven by the earlier mentioned progression towards edge and fog computing, we could witness a redrawing of the competitive landscape for data centers, while further entrenching current dominant public cloud players, as they have already specialized in the large scale maintenance of public computing infrastructure. One way to differentiate among cloud providers is through AI-as-a-service, in which AI services can be easily implemented in cloud-based applications. This could lower the threshold for businesses to experiment with machine learning (ML), as it gets rid of large upfront investments, which means that businesses could finally see some benefits of their stored data, and in turn could make both their front- and back-office applications smarter.
When it comes to the advancements in AI itself, we can expect further gradual progress with the implementation of more advanced computing infrastructure (e.g. improved sensors, 5G, specialized AI chips) as it enables a greater amount of training data to be captured, stored and processed resulting in more advanced ML algorithms. Next to hardware, we also see progress in the application of different ML architectures. In 2018, AI got its eyes through advancements in computer vision and Generative Adversarial Networks (GANs), an architecture in which two neural networks compete with each other in a zero-sum game. Similar to how advancements in natural language processing (NLP) resulted in today’s conversational interfaces, we will increasingly see the computer’s visual capabilities appear in tomorrow’s applications. Think of autonomous vehicles, AR-apps and photo and video manipulation software. Moreover, GANs are also expected to have a significant impact on fraud detection. However, at the end of the day, an important precursor for the advancement of AI is the access to talent and the integration of machine learning expertise within other disciplines.
At the service and interface level, we can see two clear trends that have disrupted the business as usual. Firstly, there are the companies like Facebook and Google, which both have data-driven consumer business models. Both had to face considerable backlash due to the misuse of customer data, whether it be hacks or through third party malpractices. Consequently, we will see that these companies will make considerable investments to regain customer trust. With the expectation that future services will only become more data intensive and sensitive, these companies will have to reconsider the way their systems are built in order to guarantee security and prove to customers that their data is handled appropriately. Then there are the more product-based companies like Apple and Samsung, who have profited for a long time from the 2-yearly smartphone spec upgrades. However, both are now facing plateauing customer device sales. One way of dealing with this is by putting more emphasis on revenue through services, as can already be seen in how Apple reports its quarterly numbers. Another appropriate strategy could be that these companies will be driven towards more exotic interfaces, features and form factors (e.g. foldable screens) in order to achieve more differentiation. A similar trend could be seen in the emergence of the digital camera market, which for years was also locked into a monotonous megapixel race. After that reached diminishing returns, companies invested heavily in new features like compact zoom lenses, stereo-photography and mirrorless designs. Another way of innovating the user experience (UX) is through the power of software, as exemplified by the Google Pixel, which has the superior camera in part enabled by superior photo improvement algorithms. However, from the perspective of the disappearing computer design paradigm, the deemphasis of the smartphone is to be expected as the main user interface becomes increasingly elusive, gradually spreading across multiple devices (e.g. smart speaker, smartwatch, AR glasses, wireless earpods), thereby allowing for more adaptivity.
Lastly, an important intervening variable which runs across the entire technology stack is the governmental and geopolitical dimension. Most apparent here is the increasingly escalating tech war. Against the backdrop of the ongoing trade war, Chinese tech companies are increasingly prohibited in buying technologies and /or deploying their technologies, which further reinforces China’s ambition to strengthen local production power. On the other hand, China and other countries (e.g. Russia, Pakistan) are attempting to gain further control over internet traffic in order to gain sovereignty. Hence, we can see that decisions in the development of cloud, GPS, chip manufacturing, AI and the roll out of 5G are not purely economically driven, but increasingly determined by considerations regarding national security, nationalism and hegemony. Hence, the contours of protectionism and isolationism are thereby increasingly being reflected in the way we organize the tech stack, thereby moving closer to the splinternet or in the emergence of three internets. Lastly, the impending interference of governments can also be seen in the growing likelihood of anti-trust measures with the aim to break the monopoly of big tech, further adding to the volatility of the tech stack.
