14 November 2022
In February 2020, Huawei released a position paper that measured the aggregate total computing power of various nations. Most rankings of national computing focus on government-owned supercomputers, but this report instead analyzed distributed computing capability, combining cloud, device, and edge compute, adjusted for network limitations and dissipation effects. The paper – notable for its forward penned by Huawei’s chairman, unlike its other English language industry reports – has two key takeaways, both for investors and for all Americans concerned with China’s commercialization of geopolitical conflict. First, the current environment is a remarkable emerging opportunity for investors who track the enabling technologies of distributed computing. Second, the paper is a telling view of how established Chinese institutions view computing competition at the national level, particularly as they leverage economic power for political and military ends.
The concept of distributed computing dates back to the very first local-area computer networks and the invention of ethernet and ARPANET in the 1960s and ‘70s. The growth of the internet in the 1990s and the introduction of the smartphone (and other smart devices) in the early 2010s could be described as subsequent step changes for the capacity of distributed computing. As measured by the Huawei paper, the United States was home to a total of 2,522 gigaFLOPS (GFLOPS) of computing power. China trailed in early 2020 with 770 GFLOPS, but is closing the gap rapidly.
Our venture firm, First In, is investing in the technologists who are driving the world to the next frontier of distributed computing. With our investing focus on early-stage security technology, we are tracking a step change in distributed computing capability that is being driven by the convergence of three security-related trends in particular: zero-trust adoption, consensus mechanisms, and artificial intelligence. Together, these trends are remarkable opportunities for security technology investing – but also make “total computing power” a national security consideration.
Zero trust adoption. The design, implementation, and overall performance of zero-trust networks has been advancing steadily since Google’s internal Beyond Corp rollout in 2014, but we see an acceleration in the adoption of zero trust architecture across the commercial enterprise. Processing power, bandwidth, and – perhaps most importantly – cultural adoption of zero trust requirements have reached a breakthrough such that it’s not a best practice only reserved for the most-heavily resourced organizations. First In portfolio companies like ZeroTier and Appaegis are in the forefront of enterprise ZT adoption. In addition, methodologies such as zero knowledge proofs (ZKP) and fully homomorphic encryption (FHE) that until recently were purely academic pursuits are finally starting to bridge into the real world, thanks to new algorithmic techniques. Enterprise-scale FHE is no longer a distant cryptographic goal; startups are being built now that will be making it a reality.
Consensus mechanisms. Step outside the hype cycle for crypto and web3 and consider a fundamental security challenge to any distributed computing system: fault tolerance. The only way for a system of independent machines in a real-world environment to identify and overcome a component failure like a byzantine fault is through some consensus mechanism. Blockchain technologies represent the first practical means of fault-tolerant inter-machine consensus, and as such, have unlocked a new category of secure enterprise-grade application companies that First In will be exploring. Indeed, one of our portfolio companies, Antithesis, is central to automating testing of consensus mechanisms. Additionally, the decentralized nature of blockchain-based applications is immensely relevant to U.S. national security, not only for the defensive/law enforcement use cases that make the news, but also because of entirely new power projection capabilities coming online as well.
Artificial intelligence. Steady advances in neural network architecture and performance over the last several years are compounding in a non-linear manner, with startling new generative capabilities being introduced in only the last few months. We are focused on how AI will drive distributed computing acceleration by supercharging the efficiency and effectiveness of computing on the edge and by generating machine-speed offensive and defensive security strategies, postures, and proactive decision-making across an enterprise. “Security” in this context touches the traditional enterprise software considerations of authentication and monitoring. It also will include unstructured data analysis, automation, serverless computing, physical security, military hardware, and even social and economic applications like social engineering and business intelligence/analysis. First In portfolio companies like Grist Mill Exchange and Shift5 are at the forefront of this trend.
Unsurprisingly, Huawei’s position paper concludes that investment in distributed computing power is directly tied to increases in national geopolitical power. It is clear from the Chinese scientific, military, and political sources cited in the paper that deep thinking has gone into a strategy to advance China to more developed stages of distributed computing at every level of society. For the United States to maintain and grow its lead in aggregate national compute, policymakers as well as investors should also recognize and embrace enabling technologies for distributed computing as strategic investment opportunities.