Stacking transistors vertically like apartments in a building
In late June 2026, IBM announced a working chip built at 0.7 nanometers — the first sub-1 nanometer process node ever demonstrated at scale — marking a quiet but consequential turn in humanity's long effort to compress more intelligence into less matter. The breakthrough, built on a vertical transistor architecture IBM calls NanoStack, does not merely shrink the chip; it reimagines the geometry of computation itself, trading the flat plane of decades past for the layered logic of a building. At a moment when artificial intelligence is rewriting the demands placed on hardware, IBM has stepped to the edge of what physics permits — and then stepped further.
- IBM has crossed the sub-1 nanometer threshold, a boundary semiconductor engineers have pursued for years, with a chip small enough to rest on a fingernail yet dense enough to outperform entire previous generations.
- The NanoStack architecture abandons the flat transistor layouts that have defined chip design for decades, stacking components vertically like floors in a building — a structural shift that multiplies processing power without expanding physical footprint.
- Markets responded with turbulence rather than triumph, as investors weighed the distance between a laboratory demonstration and the volume manufacturing that would actually translate this milestone into revenue.
- The deeper disruption is philosophical: where Moore's Law demanded ever-smaller components, NanoStack proposes a different answer — not smaller, but taller — opening a new design frontier IBM is calling 'More Than Moore.'
- If IBM can move from announcement to production, it enters direct competition with Nvidia and AMD in the data center AI chip market, a space where density, efficiency, and cost-per-computation are the currencies of dominance.
IBM crossed a threshold this week that the semiconductor world has been approaching for years: a working chip at 0.7 nanometers, the first sub-1 nanometer process node demonstrated at scale. The announcement, made in late June, matters because artificial intelligence has made the demand for denser, faster, more efficient chips more urgent than at any previous moment in computing history.
The technology behind it is called NanoStack, and its key insight is architectural rather than merely incremental. Instead of arranging transistors in the flat planes that have defined chip design for generations, NanoStack stacks them vertically — the way an apartment building fits more residents into a city block than a row of houses ever could. The result is a chip roughly the size of a fingernail that carries vastly more computational capacity than its predecessors.
The stock market's reaction was unsteady, moving in both directions after the announcement. That ambivalence reflects a real tension: laboratory breakthroughs and manufacturable products are different things, and IBM will need to prove it can produce these chips in volume before the market fully rewards the achievement.
What gives NanoStack its broader significance is the design philosophy it signals. The traditional path of Moore's Law — making transistors ever smaller — is running into the hard limits of atomic physics. By going vertical instead of smaller, IBM is proposing a different route forward, one the company is calling 'More Than Moore.' If it can execute on that vision and bring 0.7nm chips to data centers within the next few years, it could meaningfully challenge Nvidia and AMD in the AI hardware market — a competition where the stakes are measured in the future of intelligence itself.
IBM crossed a threshold this week that semiconductor engineers have been chasing for years: a working chip built at 0.7 nanometers, smaller than anything in commercial production today. The announcement, made in late June, marks the first time anyone has demonstrated a sub-1 nanometer process node at scale, a technical milestone that matters because it means packing more computing power into less physical space—exactly what artificial intelligence systems demand.
The breakthrough hinges on a design approach IBM calls NanoStack, which reimagines how transistors fit together inside a chip. Rather than arranging them in a flat plane, the way chipmakers have done for decades, NanoStack stacks transistors vertically, like apartments in a building. This vertical arrangement lets engineers cram significantly more transistors into the same footprint, multiplying the chip's processing capacity without making it physically larger. A single chip can now be described as roughly the size of a fingernail, yet contain vastly more computational horsepower than previous generations.
The timing matters. Artificial intelligence has become the dominant force driving semiconductor demand. Data centers training large language models and running inference workloads consume enormous amounts of power and require chips that can handle massive parallel computations. Every nanometer of miniaturization translates to faster processing, lower power consumption, and lower costs per unit of computation. IBM's 0.7nm achievement puts the company at the leading edge of this race, ahead of competitors who are still working with larger process nodes.
The stock market took notice, though not in a straight line. IBM's share price moved both directions in the hours after the announcement, reflecting investor uncertainty about whether the company can actually manufacture these chips in volume and whether the technology will translate into revenue. That skepticism is reasonable—there is often a gap between a laboratory breakthrough and a product customers can actually buy and use. IBM will need to prove it can move from demonstration to production, and that customers will pay for the performance gains the new technology enables.
What makes NanoStack significant beyond the raw numbers is the design philosophy it represents. For years, the semiconductor industry has relied on making transistors smaller and smaller—the traditional path of Moore's Law. But as transistors approach atomic scales, that approach hits physical limits. NanoStack sidesteps some of those limits by going vertical instead of smaller, a shift that opens new possibilities for how chips can be designed and manufactured. IBM is calling this approach "More Than Moore," signaling that the future of chip design may depend less on shrinking individual components and more on rethinking how they connect.
The company has positioned this breakthrough as a cornerstone of its strategy in high-performance computing, particularly for AI workloads. If IBM can deliver 0.7nm chips to market within the next few years, it could reshape competition in the data center chip market, where companies like Nvidia and AMD currently dominate. IBM's own Power processors, used in enterprise servers, could benefit from the density and efficiency gains NanoStack enables. The question now is execution: can the company move from announcement to production, and will the market reward it for doing so.
Citações Notáveis
IBM positioned the breakthrough as 'More Than Moore,' signaling that future chip design may depend less on shrinking components and more on rethinking how they connect.— IBM's strategic positioning
A Conversa do Hearth Outra perspectiva sobre a história
Why does going from, say, 1 nanometer to 0.7 nanometers actually matter? It sounds like a small difference.
The difference is exponential, not linear. At these scales, you're fitting exponentially more transistors into the same space. A chip the size of a fingernail can now do what previously required something much larger. For AI, that means faster inference, lower latency, less power draw per calculation.
And the vertical stacking—why is that a breakthrough rather than just an engineering tweak?
Because it solves a problem that flat design can't anymore. When you're at atomic scales, making things smaller hits hard physical limits. Stacking lets you go deeper instead of smaller, which is a fundamentally different way to think about density. It's the difference between building wider and building taller.
Who actually benefits from this first? Is it consumers, or is this enterprise-only?
Enterprise first, almost certainly. Data centers training AI models, cloud providers running inference at scale—they're the ones who can afford cutting-edge chips and who see immediate ROI from the efficiency gains. Consumer devices will follow, but probably years behind.
The stock moved both ways. Why would investors be skeptical about a breakthrough?
Because breakthroughs in the lab don't always become breakthroughs in the factory. IBM has to prove it can manufacture these at volume, at cost, with acceptable yields. And then customers have to actually adopt them. That's a long road from announcement to revenue.
What does this mean for the companies IBM is competing against?
It means the race just got faster. Nvidia, AMD, others—they're all working on their own advanced nodes. IBM just moved the goalpost. If they can sustain this lead and get to market first, they reshape the competitive landscape. If they stumble in manufacturing, it's just a good press release.