Thesis 2.0 - Sovereign Computation

The world has changed, so has our thesis.

First published in Nov 23; updated in Nov 2024

Our thesis is a living framework of ideas allowing us to operate with focus and towards what matters most to us. It is built upon our perspectives on societal and technological inflection points that are shaping markets. Within it we evaluate startups as complex, adaptive systems on a case by case basis.

Over the past few decades computational networks started to infuse every aspect of our lives. The web was a powerful enabler of the mobile, social and cloud eras. It helped initiate nascent compute platforms like crypto and AI. Software took the centerstage and ate the world. But it also led to over-funded incrementalism leaving the west de-industrialised and vulnerable.

But the tides are changing. The pandemic was a catalyst for supply chain disintegration and de-globalisation which we expect to be amplified by more geo-political volatility in the years to come. The rise of AI (maybe AGI soon) will accelerate such trends because it deeply integrates energy and compute. Energy sovereignty enables compute sovereignty. If we manage to scale both we will get intelligence abundance.

Despite the mobilisation of enormous resources towards such a future, there will be critical bottlenecks we need to overcome. Many of which are rooted in the physical world: our thinking machines need to operate interactively in the physical world to accomplish things and learn. Yet, cells and matter operate at much slower speeds than AI simulations, creating a latency bottleneck. Similarly, human brains, social systems and their institutions have high inertia and will struggle to keep pace with progress in the world of bits. This is where opportunity lies.

At Inflection we work towards overcoming sovereign computation's key constraints at the crossroads of bits and atoms.

Scale

As process scaling might come to an end it is a societal imperative that we find alternative paths to extend Moore’s Law. Current computational and networking frameworks are voracious and exponentially growing consumers of energy, accounting for over 20% of global electricity demand. The quest for new computing concepts calls for innovation in energy generation, conducting materials, lithography and the adoption of novel approaches to manufacturing. Novel technical but also legal frameworks governing interactions between humans and machines need to be developed to drive progress. We are excited to back entrepreneurs contributing to those problem sets, e.g. hardware acceleration for cryptographic algorithms (Fabric), re-programmable, general purpose edge AI chips (Stealth) or orbital edge AI running on satellites (Aptos).

Resilience

As compute scales and intelligence becomes more abundant, bits and atoms become more interconnected. We're more efficient but also more exposed to critical, large scale failures as the attack surface of our systems grows with their complexity. As much as nuclear weapons defined the 20th century world order, we believe that sovereign compute infrastructure will define that of the 21st century. AI safety, trust and alignment (for lack of a better term) as well as privacy and data-sovereignty of the individual are foundational to the liberal societies we must protect. Doing so requires us to rethink zero trust architectures from the ground up - never trust, always verify. Therefore, we like to team up with founders pushing boundaries in encrypted computation (Flashbots), local-first (Anytype), distributed systems and blockchain (Iron, Modulus, Entropy) all the way to space security (Lodestar).

Flow

Without exponential growth in data generation and access, more intelligence won't help. Improving our thinking machines capabilities necessitates new approaches to sensing, data interoperability and interfaces. We believe that novel sensing tech beyond the electro-magnetic spectrum (think quantum, bio, chemical) will unlock entirely new behaviors, aka opportunities. Data markets powered by private compute tech (see above), synthetic data or market mechanisms (think prediction markets) can be catalysts for the free flow of data. Due to the rising capabilities of our thinking machines, mouse, keyboard and screen seem archaic. We are excited to see new designs emerge, be it in a machine-to machine (think agents) or human-machine (think XR, BCI) context. We enjoy working with teams unlocking insights by providing e.g. high resolution imagery and sensing from the stratosphere (Radical), comprehensive, interactive maps of the earth's subsurface (Stealth) or data markets run on encrypted data at rest (Tune Insight).

Inflection is committed to a future where sovereign compute platforms power human progress by backing outlier entrepreneurs with capital, networks and technology.

Get in touch.