Kurzweil Scorecard: Powering the Singularity

Ray Kurzweil devoted a whole chapter of The Singularity Is Near (2005) to the
energy question, because his 2030s future needs a lot of it. Nanobots in every
neuron, molecular manufacturing in orbit, servers the size of sugar cubes —
none of it runs without watts. He sketched a twenty-year arc: thin-film solar
at pennies per square meter, space-based power beaming billions of watts to
Earth, glucose fuel cells feeding blood-borne nanobots, and self-healing
heat-recycling nanomachinery operating just below Earth’s thermal limit.

Twenty years on, the watts are arriving. The inventions attached to them
aren’t.

The predictions

This batch covers twelve energy predictions from Powering the Singularity
and two adjacent chapters. Most are indexed to the 2020s or 2030s. Three are
“claims” — things Kurzweil asserted were already happening in 2005, which we
can check in the historical record — and nine are forecasts about what
comes next.

Kurzweil’s 2005 thesis was essentially that the same exponential that powered
chips would power power itself. Nanotechnology would collapse the cost of
capturing sunlight and the mass of photovoltaic cells. Space, freed up by
cheap launch and a carbon-nanotube elevator, would become the obvious place
to put panels. And because molecular machines would be small and efficient,
the total thermal footprint of a nanobot-rich civilization would fit
comfortably inside Earth’s energy budget.

In The Singularity Is Nearer (2024) he didn’t walk any of this back. If
anything, he doubled down: “the implication is that solar will dominate
sometime during the 2030s” (ch. “Renaissance”), defending Swanson’s Law as
an unbroken five-decade trendline.

Where we actually are

Solar got cheap. Nanosolar didn’t. Kurzweil’s 2005 case study was
Nanosolar, whose founder Martin Roscheisen projected titanium-oxide
nanoparticle cells at fifty cents per watt “by 2006”
(ch. “Powering the Singularity”). The company raised over $400 million,
shipped fewer than 50 megawatts, and was auctioning off equipment by August
2013. Its CIGS nanoparticle ink was outrun by a decidedly non-nano
competitor: boring, monocrystalline silicon from Chinese giga-factories.

The outcome — cheap solar — happened anyway. Our patent index shows the
perovskite tandem cell patent family growing steadily, with 44 grants in
2024 and 40 already in 2025. The literature follows the same curve: 618
perovskite-tandem papers in 2025 alone, up from 256 in 2020. LONGi
certified a 34.85% two-terminal silicon-perovskite tandem at NREL in
April 2025, using a lithium fluoride / ethylenediammonium diiodide bilayer
passivation — the kind of atomic-scale interface engineering Kurzweil was
gesturing at, just without a company called Nanosolar doing it.

Verdict: wrong mechanism, right outcome.

Space-based solar is a real program, not a real grid asset. In January
2023, Caltech’s Space Solar Power Demonstrator reached orbit aboard a
Momentus Vigoride. The MAPLE experiment beamed power between internal
receivers on March 3, 2023, then directed a beam at Earth that was detected
on the roof of the Gordon and Betty Moore Laboratory in Pasadena on May 22
— the first wireless transfer of space-based solar energy to Earth. Japan’s
OHISAMA mission is scheduled to repeat the feat with a 180-kg satellite
beaming microwaves to a 600 m² ground array in Suwa. China’s Chongqing
University is running ground-based balloon tests. The U.S. Air Force
Research Laboratory has its Arachne experiment in the queue.

Our patent index reflects the renewed interest: a trickle through the
2010s, then US 11,929,708 and related Caltech filings on cable holders,
compactible structures, and — most directly — US 12,021,162, “Ultralight
photovoltaic power generation tiles,” granted June 25, 2024. Its abstract
describes “a photovoltaic concentrator tile for a space-based solar power
(SBSP) system,” with integrated cell, reflector, and power transmitter on
compactible structures — the actual architecture Caltech flew. That patent
is not vaporware. It is the building block of a working demonstrator.

What it is not is billions of watts from NASA. NASA is a minor actor
here; Caltech, JAXA, and ESA have the live hardware. The power levels
demonstrated to date are milliwatts, not gigawatts. Kurzweil’s prediction
gets direction right, institution wrong, and magnitude off by six orders
of magnitude with a decade of runway left.

Verdict (microwave transmission): on track.
Verdict (NASA gigawatt-class satellites by 2030s): behind schedule,
wrong mechanism.

The space elevator is still science fiction. Kurzweil’s 2029 vision of
orbital factories fed by a carbon-nanotube tether hasn’t survived contact
with materials science. Our patent index shows exactly zero recent grants
for “space elevator carbon nanotube.” The International Space Elevator
Consortium itself has conceded that defect-free carbon nanotubes at scale
remain out of reach — a single meter-long tube takes about eleven days
to grow, and structural defects cap usable strength at roughly two-thirds
of theoretical. The consortium’s 2024 material of choice has quietly
shifted to graphene super-laminate, which has also not been produced at
tether scale. Obayashi Corporation’s internal target is 2050. Nobody
serious is defending 2029.

Verdict: behind schedule.

Glucose fuel cells: the vampire bot grew up. The 2005 “vampire bot”
claim — University of Texas scientists building a glucose-oxygen fuel cell
that runs on human blood — checks out as history. It has also kept
evolving. US 10,147,963 and its continuation US 10,797,336 describe a
bioenzyme that converts glucose in cerebrospinal fluid or urine to a
hydrogen-rich fuel, feeding a proton-exchange biofuel cell with a
laccase-coated cathode. That is a remarkably literal implementation of
what Kurzweil described. A 2024 Nature Communications line of work on
carbon-foam enzymatic fuel cells hit 285 μW/cm² in vivo, and a 2025
noble-metal nanozyme biofuel cell delivered 104 μW — enough, the authors
argued, to run a pacemaker. Nobody has powered a blood-borne nanobot, for
the simple reason that blood-borne nanobots don’t exist, but the energy
source Kurzweil specified for them works.

Verdict: verified historical claim; on track as a component.

World energy is not doubling. Kurzweil forecast that worldwide energy
requirements would roughly double by 2030. In The Singularity Is Nearer
he pegs 2000 at 218 TWh of renewable electricity and 2021 total primary
energy at 165,320 TWh-equivalent — roughly 595 EJ. A 2005-to-2030 doubling
would land above 1,000 EJ. The 2024 figure is around 620 EJ and trending
flat relative to GDP. Efficiency gains, particularly LEDs and heat pumps,
broke the link he assumed between economic growth and energy demand.

Verdict: behind schedule on magnitude, right on direction.

Renewables eat the growth. The complement to the previous prediction
is the composition one: most new energy demand, Kurzweil wrote, would
come from nanoscale solar, wind, and geothermal. The IEA’s October 2025
Renewables 2025 report projects that renewables will meet over 90% of
electricity demand growth through 2030, with solar and wind capacity
doubling the previous five years’ additions. Global renewable capacity is
forecast to grow by 4,600 GW between 2025 and 2030. Kurzweil undersold it.

Verdict: ahead of schedule.

The scorecard

What Kurzweil missed (and what he nailed)

The pattern in this batch is sharp. Where Kurzweil was describing a physical
phenomenon — sunlight is abundant, photovoltaic modules follow a learning
curve, wireless power is a solvable RF engineering problem — he was right,
often by a wider margin than he claimed. Solar is ahead of the schedule in
his book. Microwave power transmission got a flight demo two years before
the earliest bound he gave it.

Where he was wrong, he was wrong in a consistent way: the vehicles he
picked for the physics. Nanosolar, not crystalline silicon. Carbon nanotube
tethers, not prosaic rockets. Molecular nanotechnology manufacturing,
not Chinese wafer fabs. Blood-borne nanobots, not pacemakers. In every
case the outcome he cared about — cheap solar, space-beamed power, implant
power, a clean grid — is arriving. The molecular-manufacturing pathway he
assumed would deliver it is not.

The cleanest tell is US 12,021,162. That patent is the space-based solar
architecture Kurzweil described. It uses no MNT, no self-replicating
assemblers, and no carbon-nanotube elevator. It is a photovoltaic
concentrator tile you fold up, stick on a Falcon 9, and unfold in LEO.
The substrate shift he predicted didn’t happen. The application that was
supposed to need it arrived anyway.

The one real disappointment is energy demand growth. Kurzweil expected the
Singularity to require roughly double the watts. It turns out efficiency
gains in lighting, heat, and compute kept demand growing at a fraction of
that rate — a genuine forecasting miss that, if anything, is good news
for the thesis of the rest of the book.

Method note

Patent counts come from our full-text index of 9.3M U.S. patent documents;
literature counts come from an index of 357M OpenAlex scientific works.
Specific patent abstracts and claims were pulled and read directly. Web
searches covered Caltech’s SSPD-1 / MAPLE results, JAXA’s OHISAMA mission,
LONGi’s 2025 NREL-certified tandem records, Nanosolar’s post-mortem, IEA
Renewables 2025 projections, the International Space Elevator
Consortium’s current tether assessment, and 2024–2025 glucose biofuel
cell power densities. Kurzweil’s original phrasings come from the
prediction corpus derived from The Singularity Is Near (2005); the
2024 restatements come from the full text of The Singularity Is Nearer.