IBM says its new chip is one of the most efficient yet. Using seaweed to 3D-print buildings. AI companies are staring down the laws of economics. Why investors should pay attention to nature. All that and more in this week’s Prototype. To get it in your inbox, sign up here.
I’m sure it’s not news to anyone reading this newsletter that energy has become a major concern for the tech world, as data center buildouts and surging AI demand have massively increased electricity demand for servers.
IBM suggests it may be able to help reduce some of that energy demand. It announced a new advance enabling the company to pack nearly 100 billion transistors into a tiny chip, which it claims can provide up to 50% more performance or 70% greater efficiency than the 2nm node chips it announced in 2021. The nodes on this chip, it says, are the equivalent of 0.7nm nodes, taking chips past nanotech and into angstromtech (to possibly coin a phrase).
“It’s not just an incremental step, it’s a meaningful leap forward,” IBM’s Jay Gambetta said at a media briefing this week.
It’s worth noting that the chip itself does not have transistors that are physically that small. That’s (probably) impossible. Instead, it has a 3D architecture, called NanoStack, that produces the computing equivalent of a chip with nodes that size. The transistors are stacked on top of one another in a staggered layout in order to maximize the number that can be placed on a single chip.
Crucially, IBM says that its chips could provide a 40% improvement in scaling SRAM (Static Random Access Memory), a technology increasingly being used for different types of AI applications because it puts computer memory physically closer to where data is being computed, speeding up processing. “This is a step change the industry hasn’t seen in decades,” Gambetta said.
Don’t expect to see these chips in data centers anytime soon, though. IBM itself doesn’t manufacture chips, instead licensing its technology to other semiconductor companies, such as Japan’s Rapidus, which is bringing IBM’s 2nm tech to market next year. At the press briefing, IBM’s Huiming Bu told reporters they expect the new chip to be ready for production within about five years.
Discovery of the Week: Seaweed Could Help Turn Dirt Into Buildings
From apartment buildings in Europe to Texas neighborhoods, 3D-printed construction is increasingly common. Now that it’s proven successful, scientists are searching for lower-carbon building materials that are more practical for construction. One promising prospect has been dirt, because, after all, it’s everywhere. But it’s not always an effective printing material, because different soils can clog printers or simply not hold their shape.
Researchers at the University of Colorado Boulder may have found a solution: seaweed. Turns out it’s not just good for holding sushi rolls together–according to a new study, sodium alginate, a compound derived from the marine plant commonly used in foods like ice cream, also makes sand and clay easier to print. It changes the electrical charges in clay particles, causing them to slightly repel each other and prevent them from clumping together and clogging printers.
The result is a much stronger final product. The researchers found that using only 0.12% sodium alginate, the material could withstand about 25% more pressure and be printed about 33% faster. Down the line, this could allow a construction team to dig a building’s foundation, use that soil as raw material, and then 3D print parts of the building on site. In their testing, they only used the mixture to print thin walls and other structures that could be part of full-scale buildings. The next step will be proving the substance can print a house that will last.
– by Sydney Goitia-Doran
AI’s Inevitable Pricepocalypse
A few weeks ago, several AI services shifted to usage-based pricing rather than flat subscription fees. The result, as my colleague Rashi Shrivastava noted in her AI newsletter earlier this month, is massive sticker shock–“ballooning costs with little productivity gain to show for it.” This is causing many enterprises to pull back on their AI spending, a trend that appears to be accelerating, especially as open-source and in-house models are “good enough” (and cheaper) for certain tasks.
This has been a long-simmering problem. I noted earlier this year that many companies aren’t seeing ROI from AI, while workers report that adopting the technology has increased their workloads instead of making their lives easier. Meanwhile, tech companies are building out data infrastructure at what is, in many ways, the worst possible time: supply chains are strained and energy costs are rising. (It’s a great time to be in the memory chip business, though.) That puts increasing pressure on companies like Anthropic and OpenAI to grow revenue quickly to offset the costs–but as the sticker shock shows, their customers aren’t quite ready to pay more until they start seeing value.
The thing is, history is full of brilliant inventions–from the Concorde to nuclear container ships to the satellite phones of the 1990s–that did their jobs well but failed because they weren’t economically sustainable. What does it profit a company to build AGI if nobody can afford it?
Now, generative AI probably won’t go the way of the Concorde, but over the coming months and years, you’re going to see increasing attention paid to actual governance and management of AI. The companies seeing ROI on their AI spending are those that have paid attention to their workflows, are finding real bottlenecks, and using automation to get past them. Expect other enterprises to follow suit. It will probably be very different from some of the “AI will just replace a bunch of jobs” fantasy scenarios.
The AI companies that make it through will be the ones that are best able to serve those customer requirements and focus on them while keeping a sharper eye on their capital expenditures. That means not just more efficient models, but hardware advances like IBM’s discussed above, as well as the new LLM-focused chip that OpenAI and Broadcom announced this week.
The Hot Take: Nature Is An Overlooked Investment Category
Each week, I ask investors for their take on tech trends within their industries. Today the answers come from Tom Quigley, cofounder and managing director of Superorganism, a venture fund investing in startups developing technological solutions to problems impacting biodiversity.
What tech is being overhyped right now?
Everything that isn’t solving real-world problems. Never before in history have we had such unlimited access to the tools to build – we should be picking hard, humanity-critical challenges to tackle, not another email tool.
What should more people be talking about today?
Nature as an overlooked investment category. Trillions in global GDP directly rely on nature, and there are treasures hidden in both the genomes of and the ecological services provided by the microbes, fungi, plant, and animal species across the world. From lab-grown plant material like seeds and fiber, to low-cost long-duration ocean robotics, to new proteins from nature for new therapeutics, embracing the investable opportunities in nature can rewrite the future.
What are we all going to be talking about in five years?
Robots in the wild. Robotics are going to move from specialized use cases (factories, etc) to being common-place in public, home, and remote environments. We’re expecting an explosion of robots in the wild, being used for difficult, dangerous, or infeasible tasks that humans can’t, shouldn’t, or simply don’t want to do.
On My Radar
State of the Strait: Some traffic began flowing again through the Strait of Hormuz, mostly on its southern route near Oman. But on Thursday, Iranian forces attacked a container ship with drones, hours after warning ships to no longer take that route. The attack has virtually halted traffic as of this writing, adding still more uncertainty to a region. Prices of other commodities like helium and sulfur remain high. Friday afternoon, U.S. forces struck back on Iranian targets, leaving the state of the current “ceasefire” in question.
An Emerging Satellite Bottleneck: SpaceX isn’t currently accepting rideshares for small satellite launches beyond early 2029, and its manifest is mostly full until then, according to reporting from SpaceNews’ Debra Werner and Emma Gatti. That’s leaving the emerging space economy scrambling to secure launches that are in short supply. That’s likely going to lead to higher prices in the near term, though as more new rocket companies come online the long-term result might be tighter competition for SpaceX, which currently dominates the market.
Pro Science Tip: Diversify Your Produce
When it comes to fruits and vegetables, I’m guided by flavor, and I assume eating enough is enough. But a recent study found even people who ate the recommended portions every day (not a large group) often weren’t getting enough flavanols, plant compounds linked to good heart health. So next time you’re in the produce aisle, be sure to throw some apples, pears, berries, stone fruits, beans and legumes into your cart. If you have a sweet tooth, cocoa and chocolate are also rich in flavanols. (Unfortunately this doesn’t include KitKats–we’ll have to suffer through dark chocolate for our health.)
– Sydney
What’s Entertaining Us This Week
Last weekend I caught a theater screening of Raiders of the Lost Ark for its 45th anniversary. Do I need to sell you on seeing this movie? Probably not. It’s a classic for a reason. Still, most of my life I’ve only caught showings of it on TV, and after seeing it on the big screen I have to say that’s the best way to see it. Movies are just better in theaters. Especially ones about archaeologists fighting Nazis.
– Alex
I was not expecting to like Masters of The Universe, the He-Man reboot starring Nicholas Galitzine and Jared Leto. Can you blame me? It got mixed reviews, had a corny trailer, and Reddit posts calling it “Disaster of the Universe.” So I was surprised when I saw it and really enjoyed it. It captures the energy of the comics and cartoons in a fun way and never took itself too seriously. The jokes landed and the characters were updated for a modern setting without losing their original flavor. Its greatest win was impressing both my dad, who grew up with He-Man, and me, who grew up with Netflix. Go check it out and decide for yourself.
– Sydney
