Kurzweil Scorecard: Ten Energy Bets From 2005, and the One That Paid Out Wasn’t on His List

In the “Powering the Singularity” chapter of The Singularity Is Near (2005), Ray Kurzweil named the technologies he thought would supply the coming explosion of computation and mind-machine merger. He spent pages on hydrogen fuel-cell cars, microbial fuel cells, enzymatic ethanol cells, carbon-nanotube “nanoscale batteries,” and a startled endorsement of desktop fusion via ultrasonic cavitation. Crystalline silicon photovoltaics, the actual winner, got almost none of the real estate.

Batch 36 gives us ten of those energy claims in one place. Twenty-one years later, we can grade them against the patents being granted today and the papers being cited.

The predictions

Every prediction in this batch comes from Kurzweil’s energy survey — nine from “Powering the Singularity” and two from “Chapter Nine: Response to Critics”. Taken together, they paint the picture of an author who believed the energy question would be solved by an eclectic mix of nano-enabled hardware, biological catalysts, and lab-bench fusion. Each was “testable” in the sense that by the mid-2020s, either a product category would exist or it wouldn’t.

In The Singularity Is Nearer (2024), Kurzweil quietly reframed the whole thesis. He now leads with “photovoltaic module costs per watt have been exponentially declining for almost five decades, a trend that is sometimes known as Swanson’s law” and argues that “solar will dominate sometime during the 2030s” (ch. “The Law of Accelerating Returns”). Hydrogen cars, sonofusion, and Rhodoferax microbes are not mentioned. The pivot is the story.

Where we actually are

Hydrogen fuel-cell cars. Kurzweil cited a 2004 Department of Energy report that claimed nanotechnology could enable “strong light storage tanks and fuel cells twice as efficient as gasoline engines producing only water as waste” (ch. “Powering the Singularity”). Global FCEV sales fell to 4,102 units in the first half of 2025, and the US market dropped 59% to 132 vehicles. Shell shut its three remaining US hydrogen stations in February 2025 and cancelled all planned builds. The 2026 Toyota Mirai represents 0.0008% of Toyota’s sales. The fuel cell worked; the refueling network, the hydrogen price, and the cost of the drivetrain did not.

Ethanol fuel cell with dehydrogenase enzymes. St. Louis University researchers, including Nick Akers, did demonstrate a stable enzymatic ethanol fuel cell in 2003. It never scaled. Shelley Minteer, the field’s most prolific academic, has now commercialized the technology through a Salt Lake City startup called 32ATPs — for medical implant power, not for transportation or grid energy. In our patent corpus, the search phrase for enzymatic fuel cells produces at most one granted US patent per year since 2011. The cell is real. The category is not.

Microbial glucose fuel cells. Chaudhuri and Lovley’s 2003 Rhodoferax ferrireducens paper is cited as proof of an 81-percent-efficient bug-powered generator. Twenty-three years later, US 12,347,907, granted July 2025, claims “anode electrode structures for microbial fuel cell (MFC) devices… for treating wastewater and generating electrical energy through a bioelectrochemical waste-to-energy conversion process” — with a conductive core wrapped in sheets of conductive textile. Global market reports put MFC revenue at roughly $16 billion in 2024, growing toward $41 billion by 2034. But the successful deployments are 1,000-liter reactors on municipal wastewater, where the electricity is a byproduct of cleaning sewage, not the point. Kurzweil sold this as an energy source. In practice it became a wastewater-treatment efficiency play.

Carbon nanotube batteries. Kurzweil wrote that “nanotubes have already shown promise as nanoscale batteries, adding energy storage to their roles in computation, communication, power transmission, and structural materials” (ch. “Powering the Singularity”). Carbon nanotubes did become critical to battery technology — but as conductive additives inside conventional lithium-ion electrodes, not as standalone “nanoscale batteries.” The CNT-for-Li-battery market was $809 million in 2024 and is projected at $6.15 billion by 2031 at a 34.5% CAGR. Vertically aligned CNT electrodes developed by European manufacturers triple energy density in research cells. The material won. The architecture Kurzweil described did not.

Sonofusion (desktop fusion by acoustic cavitation). Kurzweil argued the Taleyarkhan sonofusion claims at Oak Ridge were “more credible than earlier cold-fusion claims” and approvingly noted the 2004 DOE cold-fusion review. On July 18, 2008, a Purdue investigative committee found Taleyarkhan guilty of research misconduct for “falsification of the research record,” sanctioning his graduate-faculty status for three years after he arranged for a student to appear as co-author to fabricate the appearance of independent replication. Our patent corpus contains exactly one US patent mentioning acoustic-cavitation fusion since 2011. Sonofusion is dead science. The 2004 DOE cold-fusion panel itself split roughly evenly on whether excess heat was real, concluded the evidence was “not convincing,” and did not fund the field.

Nanotubes as power for microdevices. This is the softest claim in the batch and the easiest to verify: yes, CNT-based microscale power sources — supercapacitors, flexible electrodes, and hybrid biofuel-cell anodes — are a real research area, with 142 US patents on photovoltaic-adjacent perovskite architectures alone since 2015 and a steady drip of CNT electrode claims. Verified as a direction, still niche as a product.

Solar-cell startup interest. “By 2004, world events had sparked increased interest in solar-cell energy start-ups.” Verified at the time. But Kurzweil’s named exemplars — Konarka, Nanosolar — both collapsed. Konarka filed for Chapter 7 in May 2012. Nanosolar laid off 75 percent of its workforce by February 2013 and auctioned equipment in August 2013. At least thirteen US thin-film solar makers filed for bankruptcy in 2012 alone. What killed them was the technology he wasn’t betting on: crystalline silicon, whose module costs were dropping along Swanson’s law curve. In 2024, solar PV supplied 7% of global electricity and reached 10% of consumption for the first time, growing 30% year-over-year. LONGi’s crystalline-silicon-perovskite tandem hit 34.85% efficiency, certified by NREL in April 2025. JinkoSolar reached 34.76% in December. US 12,598,856, granted April 7, 2026, describes “a pair of tandem solar cells [where] a step surface or trench within the top subcell… is at least partially filled with another material such as an insulator support or electrically conductive support to transfer stress away from the absorber layer.” These are the patents pushing single-junction silicon past its theoretical ceiling — and they are where Kurzweil’s framing most badly missed.

Titanium dioxide photocatalysis. Verified and still progressing. US 12,453,959 (October 2025) claims a TiO2 photocatalyst doped with d-block and p-block metal impurities manganese and beyond. US 12,515,199 (January 2026) recycles waste carbon-fiber-reinforced polymer into a cocatalyst paired with semiconductor materials for “photocatalytic water splitting.” The path from TiO2-for-contaminant-destruction to TiO2-for-hydrogen-production is an unexpected second act that Kurzweil did not anticipate but that honors his original claim.

The scorecard

What Kurzweil missed (and what he nailed)

The direction was right: clean, abundant electricity has arrived faster than mainstream forecasts in 2005 expected. Solar generation grew 30% in 2024 alone. Twenty-one years after this chapter was published, the answer to “how will we power the Singularity?” is unambiguously photovoltaics plus batteries, and the Swanson-law cost curve Kurzweil now celebrates in Nearer has done more heavy lifting than any of the exotic options he originally championed.

The mechanisms were almost all wrong. Hydrogen passenger cars are collapsing. Enzymatic fuel cells migrated to implantable medical devices. Microbial fuel cells found a home inside wastewater plants. Carbon nanotubes are quietly stitched into lithium-ion cathodes rather than functioning as batteries themselves. Sonofusion was a research-integrity scandal within three years. Of his ten 2005 bets, exactly one — titanium dioxide photocatalysis — is still progressing on its original terms, and even that is being repurposed for something he didn’t mention: green hydrogen production by water splitting.

The pattern is useful. Kurzweil is best when he reasons about exponential price-performance curves that are already established — which is why Nearer leads with Swanson’s law rather than his 2005 favorites. He is worst when he reasons about specific product categories and named startups, because the winners at any given moment often lose to whatever gets cheaper faster. Predicting exponentials is tractable. Predicting which company harvests the exponential is much harder, and in 2005 Kurzweil picked mostly the wrong ones.

Method note

Patent counts come from 9.3M US grant and pregrant documents with full-text search against titles, abstracts, claims, and descriptions. Scientific-paper citation counts come from a 357M-record academic corpus. Specific patent numbers were pulled and read in full to confirm they claim what the headline suggests. News events (Shell’s 2025 hydrogen withdrawal, Purdue’s 2008 Taleyarkhan sanctions, Konarka’s 2012 bankruptcy, LONGi’s 34.85% efficiency certification) were verified against primary sources this session. No figures are recycled from prior knowledge; everything above was checked.