Chipped: Part 1
imagination, invention, innovation & the new stone age
It has been a century since physicist Julius Edgar Lilienfeld, an immigrant to the United States, patented the idea of using a semiconductor material to make a transistor. A hundred years later, silicon microchips, some with tens of billions of transistors, are in everything from computers to cars to coffeemakers. They make our modern world possible. Now, with Artificial Intelligence, they are poised to run the world.
CHIPPED is divided into five parts. Each can be read independently, but there is an arc to the order.
In 1947, a team of Bell Labs physicists built the first working prototype of a semiconductor transistor, a device that could be used as a switch to transmit or block electrons. It also worked as an amplifier, boosting electronic signals. The scientists were sure they were the first to come up with the idea, but another physicist, who could only imagine such a transistor, beat them to the patent punch 22 years earlier, in 1925.
The difference was material. Literally. The Bell Labs team had identified the essential ingredient: a semiconductor, germanium, that could be tweaked to conduct electricity or serve as an insulator. (Semiconductors are metalloids, elements on the periodic table that have both metallic and non-metallic properties.) The focus quickly shifted to silicon, a semiconductor that not only worked better, but also was much easier to source. Silicon (Si) comes from quartz, aka silica (SiO2), one of the most abundant materials on Earth.
The ability of semiconductor transistors to manage the traffic of electrons (on or off, start or stop, yes or no) made them ideal for interpreting the ones and zeros of the binary code at the foundation of every computer program.
••••••••••••••••••••••••
1947 was also the year ENIAC (Electronic Numerical Integrator and Computer), the first computer that could be reprogrammed to handle new problems, was switched on for an eight-year run at the US Army’s Aberdeen Proving Ground in Maryland. The computer weighed 30 tons, filled a 30’ x 50’ room, and had 18,000 vacuum tubes. Its unprecedented computing power was originally used for ballistics analysis, but over its tenure, ENIAC’s “brain” tackled a broad range of challenges, including weather prediction and atomic energy calculations.
By 1955, the year ENIAC was decommissioned, transistors were coming of age. The transistor radio was the first mass market commercial success, followed quickly by the transistor car radio, just in time for Rock’n’roll.
Integrated circuits—silicon “chips” with multiple transistors—came along in the late 1950s. By the mid-1960s NASA was using them to build the Apollo program’s guidance computers. Chips got us to the moon.
It was the dawn of the Space Age, the Computer Age, the Digital Age and, although rarely understood as such, also the dawn of a new Stone Age. There is no gadgetry magic, no silicon, without quartz. The modern world simply isn’t possible without rocks.
Today, eight of the world’s top 15 companies by market cap either design or manufacture silicon microchips. Collectively, they are valued at more than $17 trillion dollars, which is roughly two-thirds of US GDP, or 17% of global GDP.
LIVING INTELLIGENCE
From a handful of chips in the 1960s to chips by the trillions today, these tiny slivers of silicon run our world. And if Amy Webb, founder and CEO of the Future Today Institute, is right, everything is about to get a whole lot chip-ier. Webb predicts the convergence of three “general-purpose technologies”—AI, Advanced Sensors and Bioengineering—that are poised to spark a supercycle of growth transforming every aspect of our lives. She calls this Living Intelligence.
By Webb’s definition, a general-purpose technology is an innovation that opens the floodgates to cascades of innovation: electricity, the internal combustion engine, the Apple app store, all general purpose technologies. To have three emerge at the same time is extraordinary. To have them converge is unprecedented. Chips are the foundation, enabling and amplifying impact.
Advanced sensors—chips—are embedded in wearables, phones, cars, appliances, satellites and soil. They are in everything around us, constantly monitoring the state of things. These sensors can provide endless feasts of real-time data for ravenous AIs, transforming Large Language Models (LLMs) based on text into Large Action Models (LAMs) based on everything happening everywhere, all at once.
“If LLMs predict what to say next, LAMs predict what should be done next, breaking down complex tasks into smaller pieces. Unlike LLMs that primarily generate content, LAMs are optimized for task execution, enabling them to make real-time decisions based on specific commands. LAMs use a constellation of sensors everywhere, all around us, collecting multiple streams of data at once from wearables, extended reality devices, the internet of things (IoT), the home of things, smart cars, smart offices, and smart apartments. As LAMs become more embedded in our environments, they operate seamlessly, often without users’ direct engagement.”
Within a few years, the number of IoT (Internet of Things) connected devices is expected to double to nearly 38 billion, each laced with multiple sensors. By 2030, global demand for chips is predicted to reach $1 trillion.
That’s a lot of transistors. A lot of integrated circuits. And a whole lot of silicon.
FROM ROCK TO CRYSTAL
There is no AI, no internet, no cell phones, no TikTok, no Netflix, no drones, no video games, no streaming, no social media, and no “mining” crypto without the actual mining, smelting and refining of silicon.
The process begins with a bang in a quarry where a type of sandstone rich in quartz called quartzite is blasted out of rock, then chopped up into fist-size chunks and hauled away for smelting. Smelting is how silicon (Si) is separated from quartzite (SiO2). First, the rocks are melted together with coke, a pure form of carbon made from coal, at temperatures hovering around 3,000°F. The high heat splits the quartzite’s component silicon and oxygen atoms, while coke supplies carbon atoms. These carbon atoms readily bind with the newly liberated oxygen atoms to form CO2, a gas. What’s left in the furnace is silicon.
Smelting generates huge amounts of greenhouse gases both chemically (the binding of carbon and oxygen atoms to make CO2) and through the burning of fossil fuels to generate electricity to run an electric arc furnace.
There is a better way. A team of researchers at the Norwegian University of Science and Technology has come up with new, greener smelting process that uses aluminum in place of coke to bind oxygen atoms. This eliminates the chemical reaction that makes CO2. The process also uses one third less energy than conventional silicon smelting, so there are fewer emissions from generating the energy to run the furnace.
The new process is also less materially wasteful, able to accommodate “fines,” a powdered form of quartzite whose small particles must be filtered out for conventional smelting because they gum up the works. The aluminum can be sourced from slag (the leftovers from aluminum smelting) or from aluminum waste. And as an added bonus, slag from this new process can be used in the steel industry.
••••••••••••••••••••
Smelting gets the silicon to 98% purity, which is good but not good enough. So it is chopped up and sent along for a second process where the silicon bits are melted and chemicals added to tear apart and reconstruct the material’s crystalline structure, turning it into polysilicon.
China, which has 70% of the world’s quartzite, dominates the polysilicon trade with a market share estimated between 80% and 90%. It is easy and cheap for other countries to send their smelted silicon to China to be turned into polysilicon.
Polysilicon that is “six 9s” pure (99.999999%) can used to make solar panels, another market for which China holds a near global monopoly. Disturbingly, there have been reports of forced labor used to produce both polysilicon and solar panels in China.
Silicon that is at least “ten 9s” pure (99.9999999999%) pure is good enough to make into microchip. For that there is a third process, one with a more globally distributed manufacturing network that includes Japan, Singapore, Germany and the US, as well as China.
Once again the silicon—now polysilicon—is melted, but this time in a special crucible made from the purest silicon sand in the world in order to keep contamination to a minimum. (This special sand comes from only two mines, both located in the small town of Spruce Pine, North Carolina. When Hurricane Helene swept through the state in the fall of 2024, both mines were briefly shut down, sending shockwaves through the industry.)
Next, a seed crystal is dipped into the crucible filled with melted silicon. As it is slowly lifted out, a large sausage-shaped boule forms. From the multi-crystal polysilicon, a single giant crystal of pure silicon six to ten feet long emerges, which is then sliced into thin “wafers” and polished.
You have to squint to see the geologic origins of what is now an elegant, flawless, shiny black disk. The rock that was blasted out of a quarry, then atomically torn apart and reassembled three times, is finally ready to become a chip.
CHIPPED: Part 2 | chip fabs, superintelligence & money