Software won’t – A core tech prophecy—that software would dominate every industry—has collided with the reality that most of life and most of the economy still run on physical matter. From retail and logistics to materials science, protein modeling, and drug discovery, the pie
Fifteen years ago. tech investor Marc Andreessen published an essay with a promise: software would “eat the world.” The prediction was built on two pillars—tech companies were seemingly undervalued. and low startup costs paired with near-infinite scalability would let software-based firms take over industry after industry.
The idea didn’t stay theoretical.. Today, so-called “Mag 7” stocks dominate the S&P 500, with market capitalizations in the trillions.. Even startups like Anthropic and OpenAI are valued at hundreds of billions of dollars.. Data centers are also swelling, reshaping industries from construction to energy.
But the central question remains unanswered: how much real value is being created, and where exactly does it land?. The argument here is blunt—our daily lives still largely take place in the realm of atoms. and that reality doesn’t bend just because software gets smarter.. If anything, many of the most consequential future technologies will be rooted in physical space.
The economy is not digital
Andreessen’s essay pointed to Amazon as an example of software’s power, writing that “the world’s largest bookseller” was a software company with a core “software engine” for selling virtually everything online—no retail stores required.
That line, the piece says, doesn’t match what Amazon actually looks like today. While software remains central to Amazon’s business, the company is “firmly ensconced in the physical world,” with retail stores, hundreds of warehouses, and a massive fleet of trucks.
The broader claim follows: most economic life—housing. transportation. energy. food. and. depending on age. health care—doesn’t show up as screens.. A quick look at monthly bills. the piece argues. would suggest spending on phones. computers. and internet services is dwarfed by spending on things you can touch.
It cites a report by the International Data Center Authority stating that the digital economy accounts for a mere 15% of global gross domestic product. That’s a large number in nominal terms, but it still leaves 85% of global activity outside the digital category.
And if software can “eat” industries, the piece asks, why do companies still build in-person infrastructure? It points to the expense and effort Amazon has put into physical spaces, and it adds that Netflix—another company Andreessen touted—is opening “real-life entertainment centers.”
The tension is practical, not philosophical.. Even as people feel glued to phones. the argument insists the physical world is where they eat. work. meet. and have fun.. It says Zoom calls are “never quite as satisfying” as real-life encounters—and that software has a big appetite. but the real world is simply too big and complex to be eaten.
Still, the piece doesn’t dismiss software altogether. It frames opportunity as a matter of use: software can shape the physical world in ways that unlock enormous value.
Matter is not digital
The discussion then pivots to a more technical claim.. Materials are constantly encountered. often without notice—clothes that are soft and warm. tools with high tensile strength. substances that conduct electricity or resist shattering on impact.. Traditionally. materials science is described as a field closer to a “cottage industry”: scientists work from desired properties and then search through trial and error. testing thousands of candidates before finding something useful.
In the early 2000s, an MIT professor named Gerd Ceder began developing computational methods to predict new materials.. That work, the piece says, led to the Materials Project at Lawrence Berkeley National Laboratory.. Instead of testing thousands of candidates. scientists can use digital simulations to eliminate most options before focusing only on the most promising.
As more materials data became available. two Stanford graduate students started applying machine learning to materials databases and found they could “dramatically” improve development economics.. The company they founded. Citrine Informatics. is described as a pioneer in the space and as having attracted large players such as Dassault. Schrödinger. and Microsoft.
Yet the argument is firm: materials are not digital.. The piece says there will always be some loss in information when digital systems model physical reality.. It points to a shift on the horizon—non-digital architectures like quantum computers—as tools that can model the physical world with far greater fidelity.
Biology is not digital
The paper then moves from materials to biology, bringing in the moment the machine-learning story hit mainstream scientific credibility.. In 2024. Demis Hassabis and John Jumper won the Nobel Prize for their development of AlphaFold. an AI model that can predict—with “incredible accuracy”—the structure of proteins.
The piece calls this a breakthrough comparable to computational approaches in materials science because it allows researchers to identify potential drug candidates hundreds, if not thousands, of times faster than with conventional methods.
The promise has already surfaced in real-world biotech timelines.. It notes that in 2023, Insilico, a Hong Kong-based biotech startup, advanced the first AI-generated drug candidate into human clinical trials.. It also says there are currently dozens of potentially life-changing drugs in the pipeline that were discovered in a fraction of the time it would take using conventional methods.
But biology isn’t digital either, the piece argues.. No matter how powerful the model, proposed cures still have to be tested on humans.. Researchers must determine whether therapies are safe, non-toxic, and better than existing molecules and methods.. And that—rather than discovery alone—is said to make up the bulk of development costs.
The article cites a 2024 paper suggesting AI discovery could double the overall success rate from 5% to 10% to 9% to 18%. It then pushes back on the hype: it says claims that “AI will cure cancer” are more than overblown.
Anyone who has worked in a hospital, the piece argues, knows healthcare remains incredibly labor-intensive and depends on capable, caring professionals. It adds that the U.S. faces an extreme shortage of those professionals—and that software will do little to solve the staffing problem.
We need to focus more on atoms, less on bits
From there, the argument turns outward to a longer time horizon and a personal tone.. The piece looks back 50 years to 1976, when life expectancy in the U.S.. was 72 years versus 78 today.. It recalls that American families typically had one car and one television. that houses were smaller. nutrition was worse. pollution was worse. and there was no internet.
It contrasts that era with today.. The piece acknowledges how many things have improved, but it insists others have gotten worse.. While incomes have risen in the aggregate. it says much of that gain has gone to top earners. leaving many households feeling worse off.. It points to “amazingly cool gadgets,” while also saying costs for basic needs—housing, healthcare, and education—have soared.
The core critique is about innovation incentives. The piece says digital innovation is fast, cheap, and low risk, which helps explain why software keeps advancing. But it argues the result has been lots of incremental digital innovation and not enough transformational change in the real world.
It then frames a harsh conclusion: it’s hard to see how Americans have become meaningfully better off over the last 50 years.. Despite decades of Silicon Valley messaging, it says most American families are materially struggling and mental health is declining.. It attributes that not to an exogenous shock. but to choices that have been made—arguing that the technology exists to improve lives. but the benefits aren’t accessible to most.
The ending returns to the theme stated at the outset. The piece says the world is not digital. People live, eat, travel, and breathe in physical spaces, and it argues “no amount of algorithms and data centers” will change that.
It invokes philosopher Martin Heidegger, saying technology is “less a creation than it is an uncovering”—bringing possibilities that still require human responsibility to “enframe and direct them” so they benefit people.
In the end, the message is shaped like a principle: we live in a world of atoms, not bits, and technology matters only if it makes lives better.
software eating the world Marc Andreessen Mag 7 S&P 500 Anthropic OpenAI data centers economy not digital materials science Gerd Ceder Materials Project Citrine Informatics AlphaFold Demis Hassabis John Jumper Insilico AI drug discovery
Atoms still win? So like… robots can’t replace everything, shocking.
I feel like this article is just saying “AI isn’t magic” but with extra words. Still though, software is running my whole life, so idk how they’re concluding “atoms win.”
Wait so they’re saying drug discovery isn’t real because atoms? That makes no sense, my cousin works in a lab and they use computers all the time. Also those Mag 7 valuations got real money, so “where does it land” like it’s not landing in employee pay and investors? maybe they just don’t like the profits lol.
Every time I hear “software won’t eat the world” I think of that old internet post like 10 years ago. But then they mention data centers swelling and construction and energy, so… doesn’t that mean software is eating the world still? Like, atoms are involved, but software is still the thing driving the power bill. Kinda contradictory, unless atoms winning means we all just pay more for electricity.