Kurzweil Scorecard: The Energy Bet He Quietly Walked Away From

In The Singularity Is Near (2005), Ray Kurzweil placed a confident wager that the 2020s and 2030s would be defined by a single technology: the nanoengineered fuel cell. Tiny ones inside the body, running on glucose or hydrogen, would power our implants. Bigger ones, networked across rooftops and infrastructure, would replace central power plants. Solar would help, but the fuel cell was the headline act.

Nineteen years later, Kurzweil published The Singularity Is Nearer (2024). The fuel cell prediction is gone. The word “fuel cell” does not appear in the new book’s discussion of energy. The architecture he forecast did not arrive — but a different decentralized energy story did. He swapped the cast.

That swap is the most interesting thing about this batch.

The predictions

This run scores four 2005 Kurzweil claims about how civilization would extract and distribute energy. Three are testable now; the fourth is generational.

• “Energy for the body will be provided by microscopic fuel cells using either hydrogen or the body’s own ATP fuel” (ch. “on the Human Body”). Timeframe: by 2020s.
• “Nanoengineered fuel cells and solar power will decentralize energy production, embedding massively distributed energy resources into infrastructure and reducing reliance on centralized plants and fuel transport” (ch. “on Work”). Timeframe: by 2030s.
• “Substantial progress had been made in MEMS-scale and nanoscale fuel cells, including some using the body’s own glucose and ATP energy sources” (ch. “on the Human Body”). Timeframe: circa 2005 — a status claim, not a forecast.
• “Human civilization will reach Kardashev type II status by the twenty-second century, harnessing the power of its star for communication” (ch. “on the Intelligent Destiny of the Cosmos”). Timeframe: long.

Where we actually are

Body fuel cells: behind, and the lab knows it.

If you wanted Kurzweil’s body-fuel-cell vision to be on schedule, US 10,797,336 (granted October 2020) is roughly what you would have expected to see: an apparatus that converts glucose in cerebrospinal fluid into a hydrogen-rich fuel, then runs it through a proton-exchange membrane to make electricity. It is the right architecture. The patent has one inventor and no corporate assignee. Nothing has shipped.

The closer-to-real work sits at UC Berkeley. Patents 12,161,874 (December 2024) and 12,329,979 (June 2025) describe “self-charging” implantable biofuel cells with twin-tube anode and cathode assemblies, openings up to 900 μm, and functionalized electrodes designed for biological fluids. The 2025 grant is a divisional of a 2019 application that traces back to a 2018 provisional. Seven years from filing to a follow-on grant; still no implanted product.

The peak academic demonstration is on the same order of magnitude as it was a decade ago. The leading 2022 paper in Nano Letters (Liu et al., “Transient, Implantable, Ultrathin Biofuel Cells Enabled by Laser-Induced Graphene and Gold Nanoparticles Composite,” 59 citations) reported an open-circuit potential of 0.77 V and a maximum power density of 483 μW/cm² in vitro, with a 28-day lifetime. The most-cited body-power review (Mokhtari et al., Energies, 2022, 42 citations) sets total deliverable power from human glucose at the microwatt-to-low-milliwatt scale across the entire field — enough for a sensor, not enough for a pacemaker without a battery.

The cleanest tell that this is still pre-commercial: Oxford’s Institute of Biomedical Engineering announced in November 2025 a fresh £2.1 million project to build glucose fuel cells at “the microwatt scale” for implantable devices. New money, new program, fundamental research framing — twenty years after Kurzweil said this would be powering bodies by the 2020s.

Patent volume confirms the trajectory. US filings for “implantable biofuel cell” yielded one patent in 2024 and one in 2025. Across the broader biofuel-cell category — microbial, enzymatic, and abiotic combined — annual grants peaked at 37 in 2012 and have run at 6 to 20 per year since 2017. This is a field with steady academic output, not industrial liftoff.

Decentralized energy: wrong mechanism, right destination.

Kurzweil paired fuel cells with solar in 2005. In 2024 he kept the solar half and silently dropped the fuel cell half. In the new book he writes that “in general, decentralized technologies will define the 2020s and beyond in many areas, including energy production (solar cells), food production (vertical agriculture), and production of everyday objects (3D printing)” and that “solar will dominate sometime during the 2030s.”

The numbers behind that update are striking. Global installed solar PV capacity reached roughly 2,974 GW by end-2025, with 698 GW added in 2025 alone. Rooftop solar grew about 23 percent in 2024, adding nearly 220 GW. In the United States, distributed energy resources — residential solar, behind-the-meter batteries, controllable EV chargers, smart thermostats — were projected to reach 397 GW by end-2025. Virtual power plants, which aggregate those resources into dispatchable capacity, hit 37.5 GW and grew 13.7 percent year-over-year, with the DOE roadmap pointing to 80–160 GW by 2030.

The picture inside the patent record matches. Filings mentioning distributed solar or rooftop photovoltaic ran in single digits in the early 2000s, broke 60 in 2014, and have stayed in the 30–60 range through 2025. The assignee table reads like a list of the actual market: NEXTRACKER (utility-scale tracker mounts), SolarEdge (string-level optimizers and inverters that make distributed solar economic), SunPower, Schneider Electric, Tesla. None of them is a fuel-cell company.

The destination Kurzweil pointed at — “massively distributed energy resources embedded into infrastructure, reducing reliance on centralized plants and fuel transport” — is being reached. Just not by the technology he named. Solar plus lithium plus power electronics plus aggregator software did the work. The fuel cell was a casualty of better economics elsewhere.

MEMS/nanoscale fuel cells: the 2005 claim held up, then plateaued.

Kurzweil’s 2005 status claim — that meaningful progress on MEMS and nanoscale fuel cells already existed — was accurate. The patent record shows steady but small activity in MEMS-scale fuel cell filings every year from 2003 through 2025. The market exists: roughly $500 million in 2025, growing in the low double digits, dominated by Panasonic and Toshiba for portable consumer power and autonomous-sensor applications. The MIT/Technical University of Munich team in 2022 fabricated 150 individual glucose fuel cells on a silicon wafer using a ceria ceramic electrolyte, each 400 nanometers thick, surviving 600 °C sterilization. The engineering is real. The volume is small. Kurzweil’s status claim verified; the implied trajectory toward dominance did not.

Kardashev II: too early, but the slope is wrong.

Human civilization sits at roughly 0.73 on the Kardashev scale today. A peer-reviewed machine-learning forecast in Scientific Reports (2023) projects 0.7449 by 2060. At current trajectories, multiple analyses conclude that even Type I status will take many centuries. Type II — capturing the full output of the sun, on the order of 4×10²⁶ watts — is not a 22nd-century milestone on present trends; it is a Dyson-swarm engineering problem with no plausible buildout schedule. The prediction is not falsifiable yet, but the slope of energy consumption growth is not on the curve required.

The scorecard

What Kurzweil missed (and what he nailed)

The directional call was good. Energy did get more decentralized. Photovoltaics did follow an exponential cost curve. Distributed generation did start displacing centralized plants in the 2020s, exactly when he said it would. The Singularity Is Nearer leans hard into that — the 2024 book includes a Swanson’s Law chart showing five decades of declining module cost per watt and projects solar will cover global electricity needs sometime in the 2030s. Most forecasters in 2005 were not making that call. Kurzweil was.

The mechanism call was wrong, and the silence about it in 2024 is more interesting than the original prediction. The fuel cell — micro, MEMS, nanoengineered, glucose-fed, ATP-fed — was, in 2005, the canonical example of how nanotechnology would reshape energy from the bottom up. It was a beautiful theory. It lost the economics race to crystalline silicon, polycrystalline modules, lithium-iron-phosphate batteries, and power electronics. None of those are nanotech in the Drexlerian sense Kurzweil meant. They are conventional materials science and manufacturing, scaled relentlessly.

The instructive pattern: when Kurzweil bet on a direction (more decentralized, cheaper, more information-dense), he was usually right. When he bet on a mechanism (nanotech assemblers, atomically engineered fuel cells, glucose-powered implants by 2025), he was often wrong. The forecasting lesson — useful well beyond this batch — is that the exponential is real, but the substrate carrying the exponential is much harder to predict than the curve itself. The S-curve always wins; which S-curve wins is the harder question.

One more tell: the fuel cell prediction did not die with a retraction. It died with omission. The 2024 book simply moved on. For a forecaster who has spent a career defending old predictions, the cleanest signal of an abandoned bet is the chapter that no longer mentions it.

Method note

We scored these predictions against three sources. First, the US patent corpus from 2000 through May 2026 — every full-text patent matching terms like “implantable biofuel cell,” “MEMS fuel cell,” “microbial fuel cell,” “rooftop photovoltaic,” and “distributed solar inverter” — with assignee data joined in to identify which companies are actually filing. Second, the OpenAlex catalog of scholarly literature, filtered for high-citation papers on glucose biofuel cells and perovskite solar efficiency. Third, current-year industry reporting on global installed solar capacity, distributed energy resource estimates, virtual power plant deployments, and Kardashev-scale scoring. All numbers in this post come from those sources; quotations from Kurzweil are taken from the 2005 The Singularity Is Near and the 2024 The Singularity Is Nearer.

Sources:

• Implantable, biofuel cells for self-charging medical devices, US 12,329,979 (UC Berkeley, 2025)
• Apparatus and a method for in-vivo power generation, US 10,797,336 (2020)
• Liu et al., Transient, Implantable, Ultrathin Biofuel Cells, Nano Letters (2022)
• Mokhtari et al., Energy Harvesting from the Human Body for Biomedical Applications, Energies (2022)
• Oxford IBME — £2.1M glucose fuel cell project (Nov 2025)
• MIT/TUM ultrathin glucose fuel cell (2022)
• Snapshot of Global PV Markets 2025, IEA PVPS
• Wood Mackenzie — US virtual power plant capacity 37.5 GW in 2025
• Forecasting Kardashev progression through 2060, Scientific Reports (2023)