In just five years, venture capital's focus has shifted rapidly from marketplaces, fintech, and software toward capital- and time-intensive hard science. The article attributes this to AI reducing discovery costs and timelines, bold founders making ambitious problem-solving prestigious, and geopolitics forcing technological independence and investment. Bubbles and wasted capital may accompany the shift, so optimism must retain independent thinking and discipline.

1. From Software to Hard Science

When the author began investing about five years ago, his inbox was full of marketplaces, "Uber for X," vertical software, and neobanks. These models remain viable, but represent a noticeably smaller share of proposals.

Instead, physics, biology, chemistry, and other hard sciences, once relatively neglected for their high upfront capital needs, have moved to the forefront. Technology often changes slowly and then all at once; this shift follows that pattern.

Over a decade ago, Peter Thiel criticized the industry:

"We wanted flying cars, instead we got 140 characters."

Initially underappreciated, that diagnosis now commands broader agreement. The article asks why and how the change happened.

2. AI Changed the Economics of Discovery

The author recalls showing ChatGPT to his grandfather in 2022 and hearing it called "magic." What seemed sudden rested on decades of largely unseen development. Technology expands the possible and can therefore feel magical. ✨

Turning Points in the AI Revolution

AlexNet's ImageNet win was pivotal for deep learning and neural networks. GPUs reduced its error rate to about 15%, compared with a previous best near 26%. This suggested insufficient compute and data, rather than inherent neural-network limits, had held progress back.

The next leap was the transformer architecture introduced in Google's "Attention Is All You Need." Earlier models processed words sequentially and could lose early context in long sentences, like the telephone game. Transformers examine relationships across a sentence together. Frontier models such as GPT, Claude, and Gemini still fundamentally use this architecture.

GPT-3, introduced in 2020, strongly supported scaling laws: more compute, data, and parameters could improve performance. AI moved from unpredictable experimentation toward a question partly of capital allocation. GPT-3 demonstrated capability; ChatGPT turned it into accessible usability.

Bringing Software Speed to Experimental Science

VC traditionally favored software for its lower starting capital, rapid iteration, and quick business-model testing.

Science involved expensive trial and error: choosing a few promising hypotheses from an immense space and spending time and money testing them. AI changes those economics. Trained models can perform or accelerate search, simulation, and design, bringing software-like scalability to atoms and molecules.

Borrowing Charlie Munger's phrase, the author calls converging forces a "Lollapalooza effect." By attacking hard tech's slow, costly search and experimentation bottleneck, AI is repricing its value and investment appeal.

Examples include:

• AlphaFold reduced protein-structure prediction from months to seconds and freely released predicted structures for around 200 million proteins.
• Insilico used AI to identify a disease target and design molecules, advancing Rentosertib to human trials in about 30 months, versus a process that might ordinarily take six to ten years.
• Proxima Fusion used AI simulations to design an entire stellarator fusion plant before cutting metal or building hardware, suggesting shorter design cycles for a problem that challenged physicists for decades.

But a platform shift alone is insufficient. Founders' determination and vision must carry technology into reality.

3. One Founder's Example Can Trigger a Wave of Imitation

The privilege of early-stage investing is finding one person who changes what a generation believes possible. This may sound exaggerated, but history offers examples.

Elon Musk reset launch-cost expectations with SpaceX, opening the space economy. Demis Hassabis expanded protein science through DeepMind and AlphaFold. Jennifer Doudna redefined gene editing through CRISPR.

As such examples accumulate, people recognize that reality is changeable. Hard problems can produce large commercial outcomes, and running a company toward five-year goals, rather than proving growth every week, becomes legitimate.

"It is no longer enough for technology's brightest minds to focus only on ad clicks. Think bigger. Think progress."

Work Becomes Status and Identity

Culture is changing. Bold missions and worldviews attract exceptional talent, who see spending their time on difficult, important goals as high-status.

Even success probabilities can become secondary: attempting a meaningful problem provides purpose. Some founders who chose their businesses before AI now worry their work is not ambitious or important enough. Employees leave for hard-science companies, while founders experience FOMO about larger opportunities.

This reflects human mimetic tendencies. Once someone accomplishes the seemingly impossible, others look and move in the same direction.

4. Geopolitics Drives Technological Independence and Capital

Alongside AI and founder examples, geopolitics is another major force.

Global trade long relied on David Ricardo's comparative advantage: countries specialize where they are most efficient, lowering costs and increasing convenience. Industrial strategy is now shifting toward control and self-reliance over efficiency.

The question is becoming "Can we make this ourselves during a crisis?" rather than "Who makes it best and cheapest?" That directs capital into defense, biology, critical minerals, computing, and space.

Defense: From Imports to Domestic Production

Some defense capabilities always involved sovereignty, such as nuclear weapons, missiles, and submarines. France's independent military and aerospace capabilities, represented by the force de frappe, and independent defense primes in the US, UK, and Russia illustrate this.

Such independence was largely reserved for major powers. Most countries imported equipment, and post-Cold War production became deeply internationalized through specialization.

The Russia–Ukraine war changed those calculations. Previously outsourced or imported items became things that must be made domestically. Defense revived as a field for ambitious young founders, including Mach in the US and Harmattan in France.

Biology: COVID-19 Revealed Both Fragility and Capability

Pharmaceuticals exemplified globalization and comparative advantage. Cheap, large-scale production seemed best for patients and society. India became the world's generic-drug production base, while China's WuXi Biologics and similar companies underpinned Western chemical and biologic manufacturing.

The US led R&D and considered offshore later-stage production safe. SARS and Ebola offered warnings, but COVID-19 was the turning point.

The pandemic revealed:

1. Fragility: Western countries lacked masks, swabs, and even vaccine vials, and discovered supply chains passed through geopolitical competitors.
2. Capability: mRNA vaccines were developed within 11 months of the viral genome's publication, versus a prior record of about four years.

The US subsequently passed the BIOSECURE Act, restricting federal dealings with certain "biotechnology companies of concern." The author sees it as cementing biological sovereignty, though capital was already shifting after the pandemic, before the legislation.

Chai Discovery and Enveda in the US, Argo Bio in China, and PopVax in India show progress. They are only a starting point for founders seeking major biotech companies.

Competition Restores Urgency and Innovation

Technological independence extends into:

• Critical minerals: Phoenix Tailings, Vulcan Elements
• Computing and AI: Moonshot, Manus, Etched
• Space: Apex, Agnikul

Cross-border competition awakens domestic potential and restores urgency. Excessive nearshoring, bringing every capability home or nearby indiscriminately, remains a risk. Still, the author expects a more balanced long-term outcome and considers the overall change positive.

5. Capital for Important Problems, With Bubble Risk

He summarizes the shift:

"AI compressed discovery time, founders made it high-status, and geopolitics triggered the funding."

Is this the right direction for technology and VC? There are clear benefits: building in slow, important, neglected industries, and directing talent toward more significant problems than advertising or short-term consumer services.

He also calls it a bubble, but not necessarily a wholly bad one. A collapse could destroy considerable capital while leaving lasting, socially useful infrastructure.

Historical examples include:

• Britain's Railway Mania in the 1840s, which involved speculation and collapse but built much of its rail network.
• The dot-com bubble, whose failures nevertheless helped establish today's broadband infrastructure and cheap cloud computing.

The future is uncertain, but talent and capital moving toward important problems offers some reassurance.

6. Discipline Amid the Boom

Risks cannot be ignored. Apart from exceptions, fund management is overly focused on raising large funds and SPVs, special-purpose vehicles, in the name of capital-intensive futures.

Hard tech genuinely needs more capital than software for laboratories, manufacturing, approvals, and lengthy R&D. Yet layered structures can extract excessive fees, and implausible futures can be used to raise money.

He invokes Munger's "Serpico effect":

"When enough people profit from wrongdoing, they turn against anyone exposing it and become dangerous enemies."

Investors and founders should judge from direct participation in solving real problems, rather than following fashion. Participation as investors, founders, or consumers in one of recent technology history's most exciting platform shifts is a privilege.

His request is clear:

"Navigate this exciting tightrope with discipline, independent thought, and unrelenting optimism about a bright future."

7. Closing: We Are Building Cars Again

The article ends by recasting Thiel's statement:

"We wanted flying cars and got 140 characters. For the first time in a long time, we are building cars."

This is not literally just automotive. It means seriously building again in AI, biotechnology, energy, defense, space, and advanced manufacturing: fields changing physical reality and social foundations. 🚗

The boom may bring excess, failure, and capital losses. Still, talent and money are gathering around important human problems, and lasting technology and infrastructure may benefit future generations. The author therefore supports the shift cautiously but strongly.