‍

Maintain.

In preview, we maintain our theme of conditional optimism for the future amid much pessimism, while turning a spotlight onto the maintenance of the present, a set of undervalued opportunities ignored by too many funding a future that is (in our view) too far away. Sustaining effort is the necessary precondition for the survival and growth of software code, businesses, machines, human bodies, cities, and even civilization itself. Maintenance is not about preserving the status quo but thoughtfully fueling forward progress by improving on humanity’s past achievements. Maintenance can be conducted in place, or it can take place elsewhere. Tech giants who in the recent past invested significant capital into long-developed teams working on the edge of what’s possible presently want to maintain their momentum by spinning them out into ready-made ventures poised for future success. The largess of the future will come from leadership in the present maintained on the legacies of the past.

In our last quarterly letter, our focus was on realism (of real rates and real science) as an antidote to naive narratives of unrealistic, unfiltered, and unfettered futurism. We continue this reasoning with a contrarian take: you can’t look too far into the future. Of course Lux finds and funds cutting-edge science and technology ventures that turn sci-fi into sci-fact in AI, biotech, defense, and beyond. Over the past two years, the cost of capital has risen, rightly reining in far-flung, fanciful (and sometimes fraudulent) futuristic ideas. 

But recent optimism around the near-term fall in rates—with markets pricing in almost twice the number of rate cuts this year than the Fed has repeatedly signaled—has incited the mania we last saw in 2021. Has venture learned its lesson? We see more and more investors feverishly funding futures closer to science fiction than the facts of science (to wit: fusion, quantum, longevity, space elevators, and more). 

One of our current (not widely shared) views is that too much attention and capital is currently allocated to venture ideas that are 10+ years out—and too little to maintenance of all the systems that already exist. As a result, the mere prospect of lower rates on the horizon has caused many other generalist firms to venture further into the future: many are rebranding as “AI investors,” or otherwise trying their hand at deeptech investing. Perhaps this is out of necessity, especially as these generalists’ own previous domains—consumer, enterprise SaaS, fintech, e-commerce, and the like—have become increasingly crowded and competitive. The net result: this frothy cohort of “frontier” investors is now funding projects that we believe will have 15 or 20 year time horizons to commercialization, or even longer. That timeline works for national laboratories funded from government coffers, but venture capitalists must calibrate between sci-fi and sci-fact within the 10-year horizon of a fund.

Over the next decade, we don’t need fusion, we need more fission. We don’t need quantum computers, we need more chip production. We don’t need snake oil salesmen peddling longevity schemes, we need secure reliable production of medicines in scant supply. We don’t need to build space elevators, we need to rebuild our arsenal of democracy to better equip ourselves and our allies to deter our adversaries. The breakthroughs of the future cannot—and will not—lie on the breakdowns of the present and a break up with the past.

The Merit and Marvel of Maintenance

Before going more in-depth on our thesis of maintenance, we first return to our notion covered in past letters of the Entropic Apex (“a moment of maximum instability and imbalance where things are easily torn apart or fall to pieces”). Economically, we observe the widespread consensus of a Goldilocks soft-landing by Powell (we remain skeptical) that may prove to be illusory under the pressures of a presidential election, rising government and consumer debt, increasing concentration of capital in too few companies (The Magnificent Seven), and continuing overvaluation of assets like commercial real estate. Politically, we observe growing global populism in response to rising food prices, migration crises, border insecurity, and climate dislocation just as half the world votes this year concurrent with fresh risks of good AI being used by bad actors to undermine election integrity. Geopolitically, we observe multiple regional wars underway, with other regions sitting on powder kegs from the South China Sea and North Korea to the Sahel and the Stabroek Block of Guyana.

One key tenet of physics is the second law of thermodynamics, which holds that entropy doesn’t decrease. Things tend to fall apart––physical objects, relationships, companies and countries—unless energy is expended for maintenance. Thus the value of maintaining existing systems grows as entropy accelerates, and as we reach the Entropic Apex, that value becomes concomitantly unbounded. Recent headlines of collapsing bridges, derailed trains, power outages and blown-out airplane door plugs visually and viscerally underscore the need for greater vigilance and investment in maintenance. Maintenance is especially important when systems are under increasing pressure, with greater usage requiring more upkeep or greater disuse requiring more repair. In the context of a higher cost of capital, rehabilitating existing systems is often frugal compared to buying brand-new ones. 

Maintenance might conjure images of manual labor and analog ratchets and wrenches, but in reality, many of these maintenance technologies are very cutting-edge, often marrying past, present and future. Software and sensors help supervisors detect aberrations in machines that are at risk of breaking down, offering opportunities for efficient preventative maintenance that can improve uptime. Artificial intelligence is increasingly augmenting burnt-out workers in industries as diverse as healthcare and the creative arts, allowing them to properly maintain their well-being and quality of life. New therapeutics help patients maintain homeostasis, body weight and cognition through disease and advanced age. Recall that capex spending has two components: growth and maintenance, often in competition with each other. Yet, we see that in the technological age, growth doesn’t come at the expense of maintenance, but rather is predicated on maintenance.

Recently, a group of technologists and archaeologists offered the Vesuvius Challenge, a $1 million purse to use the latest AI language models to resurrect and interpret the Herculaneum Papyri, a series of 1,800 papyrus scrolls discovered in the debris of the volcanic explosion of Mount Vesuvius in 79 AD. Due to their fragility, the scrolls cannot be opened, so the team used a particle accelerator to create high-resolution CT scans that could then be interpreted by AI models. In late 2023, it was announced that a model had succeeded in reading the scrolls. The entropy of a disastrous volcano in the past, coupled with our present technologies, allows us not only to maintain our collective cultural and historical legacy, but also push out the scientific and technological limits of the future.

Maintaining Markets, Macro and Micro

The concentration of capital in the 10 largest stocks of the S&P is the highest it’s been since the 1970s Nifty Fifty. Investors are optimistic that AI products for the ‘Magnificent Seven’ will deliver top-line revenue growth, cost efficiencies, productivity gains, and contribution margin expansion. Microsoft’s $10 billion commitment in a highly structured deal with OpenAI last year coincided with a $1 trillion gain in market capitalization. Meanwhile for the rest of the market, results are grimmer. Across all public companies, earnings fell more than sales growth, with margins compressing and record bond issuances increasing corporate indebtedness. A rising focus on margins has led to layoffs at over 1,000 tech companies totaling more than 250,000 workers last year.

Such concentration within the S&P’s components means that a handful of companies doing well can mask the weakness in hundreds of other stocks. Investor expectations have soared around AI, and this year will show whether the fundamentals follow and the tech giants cement their dominant positions or lose hundreds of billions as investors flee. Should the market caps of the ‘Magnificent Seven’ fall, we expect it will deepen pressure for cutting-edge AI/ML research groups to spin out voluntarily to maintain their momentum as independent ventures or else depart involuntarily due to internal fiscal discipline or external activist pressure. Over the past year, Lux has created new ventures around two key AI/ML groups. From Alphabet, we spun out the team pioneering work on digitize olfaction into Osmo, and from Meta, we spun out the team behind the Evolutionary Scale Modeling AI foundation models for biology into Evolutionary Scale. The latter’s open-sourced ESMFold model is now being downloaded in record numbers from fellow Lux company Hugging Face’s model hub. We are continuously evaluating funding spin-outs of large companies, or incubating departing teams who have maintained a seamless web of trust and who have benefited from years of working together and often tens or hundreds of millions of dollars already invested from their parent companies.

The consensus as recently as Q2 2023 was an economic hard landing following a period of heightened inflation, rising rates, collapsing banks and high volatility. That pessimism has since transitioned to widespread belief in a soft landing alongside robust employment figures and expectations of the Fed cutting rates to maintain economic prosperity. Prediction markets have Trump leading Biden, and some investors believe the market has been rallying in anticipation of pro-business policies should Trump win or in response to Yellen offering an abundance of liquidity. Yet, a soft landing is hardly assured. Many surveys suggest people—despite having empirically higher incomes and net worths—feel worse emotionally due in part to sticker shock from the higher prices of everyday goods. Real estate, a key part of people’s net worths, saw sales drop to the lowest level in nearly 30 years, with 7% mortgages all but unaffordable for new homebuyers, increasing pressure on the rental market. A recent analysis by the New York Fed found credit delinquencies rising, particularly for auto loans. With elevated interest rates, consumers may struggle to maintain their mortgages, auto loans and credit card debts, leading to a cutback on essential purchases and triggering a recession. Nationally, an inverted yield curve (3-month and 10-year) continues to send a warning signal, and globally, debt is now at a record high of over $300 trillion. In just the past three years, there have been nearly 20 sovereign defaults across 10 countries, more than all defaults in the last two decades. The risk of distressed debt rippling chaos out of emerging countries is high. A hard landing hardly seems improbable.

With a watchful eye on the macro, in our microcosm of venture capital, we are conditionally optimistic—as future venture returns are premised on present malaise produced from past excess. We described in past letters the bifurcation of VC into the minnows (the small, subscale funds) and the megas (the giant, multi-billion-dollar funds). The proliferation and pace from the former funded a record number of startup launches, while the giant pools of capital put out by the latter pumped obscene amounts of cash into pre-scaling startups unready or unworthy of absorbing and allocating it all. There were 12,000 startups in 2010, but 50,000 last year. Yet, venture investments in U.S. startups hit the lowest level in four years—from a peak of $300 billion in 2021 to around $170 billion in 2023. As the death rate of startups and funds increases, past decisions will look foolish and future returns may well look wise.

Last year, we closed Lux 8, our largest fund ever, at a time when the fewest funds were raised since 2013 (481). With capital allocators directing both assets and attention elsewhere, we maintain our prior prediction of the powder, the pace and the pullback. First, we expect a lot of dry-powder to prove wet (capital that will be used to shore up weak companies in underwater portfolios rather than invested into new ventures). Second, we expect that pace of industry-wide investment will continue grinding slower (one large growth fund that did nearly 200 deals in 2021 executed just 20 investments last year; another once-active giant fund conducted just seven; many other funds have run into fundraising speedbumps). Third, the pullback: there will be many more dispiriting exits in venture, with soft landings for talent but hard landings for investors.

We remain conditionally optimistic even as venture broadly contracts for two reasons. First, the loosening of talent across the tech industry allows for the maintenance and cross-pollination of scientific progress. We spend a lot of time recruiting top talent to our companies––in recent months, recruiting dozens of employees from Apple, Spotify, Scale AI, Google, Stripe, Intel, Meta, and Adobe into our portfolio. Second, three key Lux investment areas we summarize as GPT-4, the G7, and GLP-1 (which is to say artificial intelligence, defense for global allies and biotech/health) are poised to continue taking advantage of the need for maintenance in creating momentum and magnetism. We will walk through each of these sectors in turn.

Open and the AI of Technology Maintenance

Software is how we maintain most of today’s complex systems, cobbling together code into the combinatorics of capabilities that power our applications. Software engineers constantly maintain their code to address security vulnerabilities, fix bugs, prevent data loss, improve performance, and ensure compatibility with new hardware and software releases. Future functions are merely the present code evolved from past engineering efforts. As much as 80% of financial transactions use FORTRAN or COBOL––60-year-old programming languages. Nearly half the banking system uses COBOL, and it runs 95% of the time you swipe your ATM card.

With so many systems profoundly reliant on software, proprietary engineering efforts aren’t sufficient to maintain our systems. To compensate, technology companies for the past two decades have increasingly moved toward open-source and community-driven models of code for key technologies and infrastructure. All of us today depend on a long tail of hundreds of thousands of contributors who commit code to improve the world’s collective software, while a small handful of valuable maintainers ensure the code commits and pull requests meet standards of quality, functionality and security even as they handle bug reports and feature requests, moderate the community and settle conflicts.

Lux has devoted exceptional attention to funding robust and open software platforms, particularly in artificial intelligence where Lux family companies Hugging Face, MosaicML (acquired by Databricks), Runway, Osmo, Together AI and Evolutionary Scale have all used open and maintainable approaches to build world-class technology businesses. Take Hugging Face. Its research team open-sourced the Transformer technology (the T in ChatGPT) that powers all large language models (LLMs) today by publishing open-access papers and providing free libraries to the AI community to build upon their work. Now through its platform, Hugging Face has become the center of the global AI community, where thousands of public and private maintainers work together to advance the state of the art in AI. In fact, on Hugging Face’s model hub, any developer can freely browse, download and instantly use over 500,000 open-sourced AI models of varying sizes and capabilities—many of which are fast approaching, or have exceeded, the performance of proprietary models in specific tasks. 

Our first investment in Japan was Sakana AI, which brings together the world’s greatest AI talent into the heart of Tokyo pursuing new approaches to large language models (LLMs) through nature-inspired intelligence and complex evolutionary systems. Sakana, a Japanese word meaning fish, evokes a school of fish coming together and forming a coherent entity from simple rules. It’s a great example of the open spontaneity of technological progress. Co-founder David Ha published his ML research and demos on a once-anonymous viral blog, while his partner Llion Jones co-wrote the foundational paper “Attention Is All You Need”, whose groundbreaking transformers technology Hugging Face open-sourced to the wider public.

Open-source projects advance through just such collaborative contributions, underscoring the beauty in the continual process of code improvement. In Japan, there is a concept called “kintsugi” where a broken porcelain object is repaired with glue sprinkled with gold dust powder to celebrate, accentuate and emphasize the beauty of maintenance in a fragile and entropic world. We see a reflection of this philosophy today in cutting-edge 3D printed prosthetics, which showcase form and function with striking design instead of hiding a missing limb. Engineers and users, weaned on open systems, are increasingly joining the “right to repair” movement, challenging throwaway culture and emphasizing the beauty and functionality of maintenance. So too with software patches and updates. In cybersecurity, the idea of “zero trust” architecture assumes imperfections inside of a company’s systems and demands strict identity verification (never trust, always verify). Lux companies like Nozomi and ZEDEDA strengthen cyberdefenses through progressively more secure networks, a reflection of kintsugi within software infrastructure.

Maintenance for American and allied defense
Defense and aerospace is another industry solving big problems and seeing rising investment demand—and where we’ve been investing for many years. Given a growing set of global kinetic conflicts, the world is realizing that a tense war has taken form between two tenses—the future and the past. On one side: America and our allies—amid all our imperfections—advancing toward the future: progress, prosperity, pluralistic secular societies, tolerance, open exchange, scientific collaboration, and individual freedoms alongside a law-respecting world order working together to increase wealth, human dignity and open competition for all. In opposition: an increasingly aggressive and autocratic axis of revanchists and revisionists pining for the past; attacking ships, sea lanes, civility and the world order all at once; instigating instability by invading sovereign countries; and pugilistically promulgating conflict, chaos and wars while using their oppressive, censorial, and intolerant politics to suppress any form of dissent. America and its allies stand for maintenance and flourishing, while its opponents emphasize entropy to end the evolution of progress.

Peace requires maintenance of materiel––si vis pacem, para bellum (Latin for “if you want peace, prepare for war”). Leading Lux companies like Anduril, Saildrone, Planet, Impulse Space, Epsilon3 and Nominal are collectively winning publicly announced deals of unprecedented size and are also meaningfully involved in valuable private transactions. The tailwinds for them are strong. In the aerospace and defense industries, it will take some years for a bubble to form, but the setup is beginning: starting with standout companies (like Anduril) rapidly attracting talent and capital, turning product launches into programs of record that are then reflected in rising revenue and valuations. That public success provokes ambitious entrepreneurs to create adjacent ventures and attract additional capital––which creates a growing investable universe. What we expect to follow are dedicated ‘defensetech’ fund-of-funds giving LPs access to underlying companies, albeit two layers removed to give moral distance from the messy morass of actual warfighting. Eventually, unconstrained avarice will supply too much capital and lower future returns for these industries, a trend we are closely watching. But that time is not today.

The global balance of power is determined by a country's maintenance of its tactical and strategic edges. New programs in predictive maintenance using AI and a range of sensors (vibrational, thermal, positional, pressure, fiber optic, air data) have improved tactical readiness and helped to avoid costly or fatal crashes (like the Army’s implementation on Apache AH-64 and Black Hawk UH-60 choppers). Strategically, the Pentagon and Space Force are investing to maintain and upgrade our missile defense and detection systems. Satellites scan the globe for the heat signatures and trajectories of rockets, which can be quickly relayed to counter missiles. These space systems defend ground-based systems, but they in turn need defenses themselves, especially as adversaries like Iran, Russia and China are trying to interfere and disable them. To maintain defense and deterrence capability, these systems will be expanded across Earth orbits like LEO, GEO, MEO, and HEO in a complex constellation requiring precision placement and long-term maintenance. Lux company Impulse Space, conceived and run by longtime SpaceX founding rocket engineer and CTO Tom Mueller, is developing the in-space propulsion systems necessary for these kinds of high-precision delivery and positioning missions. America’s destiny is in the stars, a future predicated on winning the present and building upon our past legacy of scientific and technological excellence.

Maintaining biological and mechanical machines

The human body is a marvelous machine of maintenance, but requires therapeutic help to maintain its peak performance. Biotech’s miracle molecules modifying metabolism added massive market cap value to Eli Lilly and Novo Nordisk, but they will need maintenance to keep weight off––closely watch Kallyope, created and co-founded by Lux, on their clinical milestones along the gut-brain axis. Biotech, and particularly the intersection of biotech with AI, remains an exceptional area for future investment returns, particularly as high interest rates have reset valuations from recent excesses. Lux companies like Eikon, Aera, Cajal, Evolutionary Scale and many others are maintaining humanity’s defenses against disease and disability, building out platforms for rich pipelines of discoveries with rigorous promise for patients.

Maintenance doesn’t stop at the human body, but radiates out to the systems, vehicles and buildings that protect them. Hospital and healthcare equipment from MRI to ventilators demand routine checks, repairs, part replacements and calibration to ensure accuracy and reliability. Lux company Lumafield does rapid multi-spectrum computed tomography (CT) scans of complex technical and electronic products to maintain working parts and spot defects in high-margin products. SkySelect uses AI and logistics software to help procure airplane parts and service MROs (maintenance, repair and overhaul) to improve and assure the safety of flight. On the ground, Kinetic uses state-of-the-art robotics and software for maintenance and calibration of all the sensors, hardware and software in next-gen electric and autonomous vehicles, while Applied Intuition helps automakers build robust software platforms for their increasingly autonomous cars. Finally, 3D scan and data platforms like Matterport and OpenSpace help in the maintenance and preservation of physical spaces, ranging from construction zones to cultural, architectural and historical monuments and buildings, ensuring that these treasures and shelters remain robust in the face of an ever-entropic world.

‍
Maintaining the future in the present through the past

In all societies and systems, there are invaluable maintainers, what some call “keepers of the fabric” (which comes from the actual title of a maintainer at Salisbury Cathedral where one of the oldest clocks in the world still runs). Examples of these leaders range from cultural leaders preserving heritage through oral, culinary and costumed traditions, to experienced board members at startups coaching founders through tough times, to senior cabinet advisors counseling presidents dealing with diplomatic conflicts, to high-tech product managers navigating supply-chain components, to drug-hunters working with medicinal chemists. Each of these keepers maintains past historical knowledge, narratives as well as stories of successes and failures to help present-day decision-makers morph mayhem into a science of meticulous maintenance in pursuit of the future.

Without maintenance and maintainers––systems fail. In nature, maintainers are known as keystone species, playing critical roles in keeping up the structure of ecosystems that impact other organisms. Sea otters keep urchin populations in check and prevent them from overgrazing on kelp forests. Beavers cut down dying trees, making room for new ones while creating wetlands that serve as habitats for other species. Wolves act as apex predators upon deer and elk thus maintaining the ecological balance of grasslands. Bee pollination helps maintain flowering plants, fruits, nuts and the extended food chain.

Many investors are driving toward the impossible, gripping a wobbling steering wheel on a rattling chassis of axles and wheels while careening off toward some future fiction. They’re convinced by their own disruptive narrowness that the speed limits don’t matter, the brake pedal and pads are irrelevant, the turn signals and headlights be damned—who needs vision anyway when the future is so close? After all, objects in mirror are closer than they appear. But recklessness leads to wrecks, not the future delivered. Tomorrow’s progress and growth depends on today’s maintenance and investment in the past’s legacy of achievements and discoveries. Science is nothing more than stochastic steps toward transcendence. If the underlying systems don’t work, nothing else does either. Amid breakdown in the venture markets, we find ourselves very focused on maintaining both strong discipline and a very high bar for very high future returns.

A final story. Legendary technology philosopher Stewart Brand recently released “The Maintenance Race,” his first chapter of a forthcoming book that details the 1968 first-ever, solo, non-stop, no-help-allowed, 30,000-mile, around-the-world yacht race told through the lens of maintenance. It profiles three sailors, one a meticulous maintainer prepared for worst-case scenarios and whatever was thrown his way; another who embraces simplicity and joy and designs his boat for minimum breakage and could win but opts out of the race to enjoy the ride; and the last who relies on cutting-edge technology that fails soon after starting, leading to tragic choices and tragic fate. We remain conditionally optimistic that the parable will be keenly understood for its central message: those who build to last are those built to win.