Ben Lamm
Variety via Getty Images
This week’s Prototype covers AI models predicting evolution, AI’s significant debt, and a step back in defense technology. To receive updates directly, subscribe here.
Artificial intelligence is already employed in predicting weather, stock market trends, and protein structures. Ben Lamm, co-founder of the de-extinction firm Colossal, aims to expand AI’s capabilities further by developing models that can predict evolutionary changes across entire biological systems.
This ambition is being pursued through Astromech, a spinoff from Colossal co-founded by Lamm and geneticist George Church. Astromech announced yesterday that it secured $20 million in venture funding led by Bob Nelsen, co-founder of Arch Ventures. This investment increases the startup’s total funding to $60 million and values the company at $3.8 billion.
The core of Astromech’s models is the extensive biological data collected by Colossal, which sheds light on species’ evolutionary changes. For instance, they analyze genetic similarities and differences between modern elephants and woolly mammoths to forecast future genetic changes by working backwards.
This research is not just of interest to paleontologists. Lamm explained that understanding genetic evolution helps in developing better crops, discovering superior pharmaceuticals, and identifying emerging disease threats earlier, thus benefiting farmers, pharmaceutical companies, and public health officials alike.
Currently, Astromech concentrates on longevity, mapping 46 genes related to extended lifespans, cancer resistance, and cellular maintenance. Lamm noted that the new funds will primarily be used to hire research teams. He added that their models are specifically designed for these challenges, so computational costs are lower compared to other AI companies.
Astromech is not alone in creating datasets for modeling biological changes. However, Lamm emphasized that Colossal’s data provides a unique advantage due to its comprehensive geographic and temporal range. “I don’t think anyone has the geographic distance and the time distance that we have,” he remarked.
AI’s Enormous Debt Load
The rapid expansion of AI infrastructure demands around $2 trillion, which the bond market cannot support, according to Torsten Slok, chief economist at Apollo.
Despite significant spending, tech companies still require more funds. A Wall Street Journal analysis reveals that $3 trillion in AI-related debt is not reflected on the books of major tech firms, in addition to the $600 billion in reported capital expenditures.
The hope is that this investment will eventually pay off as AI demand grows. However, questions remain about its viability. While AI revenue is rising, it is not yet sufficient to cover expenditures.
Furthermore, despite the potential of AI tools, companies struggle to convert this potential into lasting productivity gains or profits, even when AI adoption boosts revenue. If this persists, merely creating useful tools will not be enough. AI companies must find ways to turn their tools into a genuine competitive edge.
The Hot Take: We Should Be Talking About CPUs
Each week, I consult investors for insights on tech trends in their fields. This time, the perspective comes from Nan Li, founder of Dimension Capital, which invests in frontier science and healthcare startups. I profiled the firm a few years ago, and last month it announced it has raised $800 million for its third fund.
What tech is being overhyped right now?
Consumer Robotics. Foundation models for robotics are showing promising generalization and scaling laws, which could lead to the belief that household robots are imminent. However, consumer settings present challenges like unbounded tasks, diverse environments, and strict safety requirements. The path for robotics will likely continue through industrial applications, offering early commercial opportunities and data collection for further model training.
What should more people be talking about today?
CPUs. For years, the AI economy has relied on GPUs. Regardless of competition among labs and models, the foundation has been matrix multiplication on specialized hardware. Following Nvidia’s dominance, CPUs are gaining attention. As data centers become more intricate and workloads evolve, CPUs are increasingly tasked with orchestration, logic, and I/O functions. Intelligence growth continues through compute, with CPUs playing a larger role for hyperscalers and neoclouds.
What are we all going to be talking about in five years?
Agentic Science. The convergence of agentic reasoning, foundational models in biology and chemistry, lab automation, and data generation will dramatically speed up scientific research in medicine, materials science, and physics. We’re seeing significant advancements across scientific capabilities, and the compounded impact of these technologies will likely lead to a research boom soon.
On My Radar
mRNA for cancer: Recently, Moderna released clinical trial results for a cancer vaccine it is developing with Merck for melanoma. The vaccine extended patient lifespans and prevented the cancer from recurring and spreading. This news has more than doubled the company’s stock value.
Moving military tech backwards: The Trump administration is reversing some advanced defense technologies. Last week, it directed the Navy to replace electromagnetic catapults on aircraft carriers with older technology, a costly decision requiring more sailors. Furthermore, the Army announced the closure of a unit focused on drone warfare in Europe, despite ongoing conflicts involving drone fleets.
Orbital data center race: Starcloud, a Florida-based company developing orbital data centers, raised $250 million this week, doubling its valuation to $2.3 billion.
Big tech socialism: The Cato Institute, a libertarian think-tank, points out that the federal government has now acquired stakes in 30 companies involved in emerging technologies, such as rare earth minerals, rockets, nuclear power, and quantum computing. While the government has previously provided bailouts or subsidies, taking partial ownership is a new approach and raises concerns about fair market competition if the government has vested interests to protect.
What’s Entertaining Me This Week
Despite Sydney’s suggestion, I haven’t yet watched Masters of the Universe. However, I’ve been enjoying the film’s score by Daniel Pemberton, who collaborated with Queen guitarist Brian May on several tracks. The music, reminiscent of 1980s rock-infused scores like Flash Gordon by Queen, is a perfect backdrop as I dive into the Dungeon Crawler Carl series, currently on the seventh book, This Inevitable Ruin.
