# Humanity Stats > The live hub for watching civilization evolve — population, energy, exploration milestones, AGI progress and country-level data across Earth, the Moon, Mars and the Solar System. Humanity Stats is a real-time instrument for watching civilization evolve: live population and resource counters, exploration milestones, AGI progress, the Kardashev energy scale, and country- and region-level data — rendered across owned 3D engines for Earth, the Moon, Mars, the Solar System and the galaxy. - Canonical site: https://humanitystats.com - Sitemap: https://humanitystats.com/sitemap.xml - Languages: English (default, unprefixed) plus French, Spanish, German, Portuguese, Italian, Chinese, Japanese, Korean, Russian and Hindi. Non-English pages live under a locale prefix (e.g. /fr/...) with translated slugs. ## Live & real-time - [Live](https://humanitystats.com/live): real-time dashboard of the planet's current state. - [Strait of Hormuz monitor](https://humanitystats.com/live/hormuzstrait-monitor): live tracking of one of the world's most critical maritime chokepoints. ## Earth - [Earth](https://humanitystats.com/earth): interactive 3D globe showing the live state of the planet. - [World](https://humanitystats.com/world): owned country globe — bind any statistic to every nation. - [Explore](https://humanitystats.com/explore): drill from countries down to regions and departments with per-region stats. - [Submarine cables](https://humanitystats.com/earth/submarine-cables): the undersea cable network that carries the internet, mapped on the globe. - [Population 3D bar chart](https://humanitystats.com/earth/population-3d-bar-chart): world population rendered as extruded bars over the globe. - [All countries](https://humanitystats.com/stats/countries): every country on Earth in one board, rankable by population, GDP, GDP per capita or life expectancy — and the index that links to all of the profiles below. - Country profiles: one indexable page per country at https://humanitystats.com/stats/countries/ (e.g. https://humanitystats.com/stats/countries/france) with its population, economy and comparative standing. ## Energy - [Energy reserves and production](https://humanitystats.com/energy): proven oil and gas reserves against actual production for every major nation — the potential a country holds versus the share it is currently exploiting, tracked over time. ## Moon & Mars - [Moon](https://humanitystats.com/moon): 3D Moon with exploration context. - [Moon landing sites](https://humanitystats.com/moon/landing-sites): every crewed and robotic landing, located on the surface. - [Mars](https://humanitystats.com/mars): 3D Mars and the state of its exploration. - [Mars landing sites](https://humanitystats.com/mars/landing-sites): rovers, landers and mission sites mapped on Mars. ## Worlds (real NASA surface maps, in 3D) - [Worlds](https://humanitystats.com/worlds): every solid body in the Solar System with a global mission mosaic — 14 of them — streamed as live NASA Trek tile pyramids onto our own spheres, labelled with the IAU Gazetteer of Planetary Nomenclature, and ranked by a settlement difficulty index computed from measured gravity, temperature, pressure, radiation, water, solar flux and access cost. The honest habitable fraction of every one of these surfaces is zero; the index says how far from Earth each one sits and which constraint is the binding one. - [Mars](https://humanitystats.com/worlds/mars): The only other world where a settlement is a serious engineering question. Viking Orbiter VIS mosaic at 232 m/px; settlement difficulty 26.6/100, the constraint that most sets it apart being radiation; 0 human and 10 robotic landings. Every named place on it: https://humanitystats.com/worlds/mars/features. - [The Moon](https://humanitystats.com/worlds/moon): The only other world a human has ever stood on. LRO LROC WAC mosaic at 100 m/px; settlement difficulty 29.1/100, the constraint that most sets it apart being pressure; 6 human and 24 robotic landings. Every named place on it: https://humanitystats.com/worlds/moon/features. - [Titan](https://humanitystats.com/worlds/titan): The only moon with a real atmosphere, rivers, lakes and rain — none of it water. Cassini ISS, 9873 frames at 938 nm mosaic at ~4 km/px; settlement difficulty 32.6/100, the constraint that most sets it apart being available energy; 0 human and 1 robotic landings. Every named place on it: https://humanitystats.com/worlds/titan/features. - [Venus](https://humanitystats.com/worlds/venus): Earth-sized, Earth-gravity, and the most hostile surface in the Solar System. Magellan SAR mosaic at 75 m/px; settlement difficulty 43/100, the constraint that most sets it apart being water; 0 human and 10 robotic landings. Every named place on it: https://humanitystats.com/worlds/venus/features. - [Ceres](https://humanitystats.com/worlds/ceres): A dwarf planet made largely of water ice, sitting in the asteroid belt. Dawn Framing Camera mosaic at ~140 m/px; settlement difficulty 43.8/100, the constraint that most sets it apart being gravity; 0 human and 0 robotic landings. Every named place on it: https://humanitystats.com/worlds/ceres/features. - [Callisto](https://humanitystats.com/worlds/callisto): The most cratered object known — and the one place in the Jupiter system a crew could survive. Voyager ISS / Galileo SSI mosaic at 1 km/px; settlement difficulty 44.6/100, the constraint that most sets it apart being available energy; 0 human and 0 robotic landings. Every named place on it: https://humanitystats.com/worlds/callisto/features. - [Mercury](https://humanitystats.com/worlds/mercury): The closest world, and the only inner planet nobody has ever landed on. MESSENGER MDIS mosaic at 166 m/px; settlement difficulty 46/100, the constraint that most sets it apart being pressure; 0 human and 0 robotic landings. Every named place on it: https://humanitystats.com/worlds/mercury/features. - [Phobos](https://humanitystats.com/worlds/phobos): A 22-km rock in a decaying orbit, and the cheapest handhold in the Mars system. Viking Orbiter VIS mosaic at ~5 m/px; settlement difficulty 48.7/100, the constraint that most sets it apart being water; 0 human and 0 robotic landings. Every named place on it: https://humanitystats.com/worlds/phobos/features. - [Enceladus](https://humanitystats.com/worlds/enceladus): A 500-km snowball venting its ocean into space through cracks at the south pole. Cassini ISS mosaic at 100 m/px; settlement difficulty 54.8/100, the constraint that most sets it apart being available energy; 0 human and 0 robotic landings. Every named place on it: https://humanitystats.com/worlds/enceladus/features. - [Ganymede](https://humanitystats.com/worlds/ganymede): The largest moon in the Solar System — bigger than Mercury, and the only one with a magnetic field. Voyager / Galileo SSI / JunoCam mosaic at 359 m/px; settlement difficulty 55/100, the constraint that most sets it apart being radiation; 0 human and 0 robotic landings. Every named place on it: https://humanitystats.com/worlds/ganymede/features. - [Vesta](https://humanitystats.com/worlds/vesta): A protoplanet with an iron core that stopped growing 4.5 billion years ago. Dawn Framing Camera mosaic at ~60 m/px; settlement difficulty 56.2/100, the constraint that most sets it apart being water; 0 human and 0 robotic landings. Every named place on it: https://humanitystats.com/worlds/vesta/features. - [Ryugu](https://humanitystats.com/worlds/ryugu): A 900-metre rubble pile that has been brought back to Earth in a jar. Hayabusa2 ONC-T mosaic at ~0.5 m/px; settlement difficulty 58.1/100, the constraint that most sets it apart being water; 0 human and 4 robotic landings. Every named place on it: https://humanitystats.com/worlds/ryugu/features. - [Europa](https://humanitystats.com/worlds/europa): A saltwater ocean twice the size of Earth's, under 15 km of ice. Voyager ISS / Galileo SSI mosaic at 270 m/px; settlement difficulty 60.9/100, the constraint that most sets it apart being radiation; 0 human and 0 robotic landings. Every named place on it: https://humanitystats.com/worlds/europa/features. - [Io](https://humanitystats.com/worlds/io): Four hundred active volcanoes, and the worst radiation environment with a solid surface. Galileo SSI / Voyager ISS mosaic at 1 km/px; settlement difficulty 76.7/100, the constraint that most sets it apart being radiation; 0 human and 0 robotic landings. Every named place on it: https://humanitystats.com/worlds/io/features. - [Feature classes](https://humanitystats.com/worlds/feature-types): the 53 Latin descriptor terms the IAU uses to classify every named surface feature — mons, patera, chasma, corona, tessera — each defined, counted across all 14 mapped worlds, and ranked by the largest example anywhere. One page per term at https://humanitystats.com/worlds/feature-types/ (e.g. https://humanitystats.com/worlds/feature-types/paterae), and one page per term per world at https://humanitystats.com/worlds//type/ (e.g. https://humanitystats.com/worlds/venus/type/coronae). - Where to land: every world has a computed ranking of its named places by settlement difficulty at https://humanitystats.com/worlds//sites (e.g. https://humanitystats.com/worlds/moon/sites). Each site is scored on seven factors that genuinely vary across a surface — annual mean sunlight integrated over the orbit, distance to the latitudes holding ice, radiation geometry inside a magnetosphere, equilibrium temperature at that latitude, what the landform class implies for landing, line of sight to Earth on a tidally locked world, and distance to a previous landing — and reported as a delta against the world's own index. Gravity, pressure and delta-v are constant across a body and are not counted twice. - Named places: every IAU-approved feature has its own page at https://humanitystats.com/worlds// (e.g. https://humanitystats.com/worlds/mars/olympus-mons) with its size, coordinates, etymology, nearest neighbours, its own settlement difficulty score against the world average, and a 3D view framed on it. ## Making Mars habitable - [How to terraform Mars](https://humanitystats.com/mars/terraforming): the 23 systems humanity would have to build to live on Mars, each in interactive 3D with the real numbers, the case for it and the case against it. Sourced from Jakosky & Edwards (2018), Kite et al. (2024), Wordsworth et al. (2019), Zubrin & McKay, DuPont & Murphy (2021) and the 2025–2026 terraforming roadmaps. - [Orbital Mirrors](https://humanitystats.com/mars/orbital-mirrors): A mirror the size of a small country, aimed at the pole. (designed, 125 km radius · +5 K). - [Engineered Nanorods](https://humanitystats.com/mars/aerosol-nanorods): Glitter, nine microns long, that turns the sky into a blanket. (tested, 5,000× better than any gas). - [Super-Greenhouse Gas Factories](https://humanitystats.com/mars/greenhouse-gas-factories): Deliberately building the pollution Earth spent decades banning. (designed, 150 Mt over 20 years). - [Silica Aerogel Shields](https://humanitystats.com/mars/aerogel-domes): Two centimetres of frozen smoke, and the ground beneath is habitable. (tested, 2–3 cm → +50 K). - [Melting the Polar Caps](https://humanitystats.com/mars/polar-ice-caps): The planet keeps its spare atmosphere frozen at the south pole. (designed, Doubles the air — to 12 mbar). - [The Atmosphere](https://humanitystats.com/mars/atmosphere): Six millibars of the wrong gas, and blood that boils at body heat. (flown, 610 Pa — 0.6% of Earth). - [Making Oxygen](https://humanitystats.com/mars/oxygen): A toaster-sized box on Perseverance already did it, 122 grams at a time. (flown, MOXIE: 12 g/h, 122 g total). - [The Nitrogen Problem](https://humanitystats.com/mars/nitrogen): The gas nobody thinks about is the one Mars cannot supply. (speculative, Need 2–3 × 10¹⁷ kg). - [Importing Volatiles](https://humanitystats.com/mars/comet-bombardment): If the planet lacks air, throw air at it. (speculative, 3.89 × 10¹⁵ kg per millibar). - [Artificial Magnetosphere](https://humanitystats.com/mars/magnetic-shield): Mars lost its magnetic field four billion years ago. This is the replacement. (designed, 3,400 km superconducting loop). - [Radiation](https://humanitystats.com/mars/radiation): Measured, not modelled: a rover has been counting since 2012. (flown, 0.21 mSv/day at the surface). - [Dust](https://humanitystats.com/mars/dust): The planet is covered in a toxic powder that gets into everything. (flown, Global storm every ~3 Mars years). - [Water](https://humanitystats.com/mars/water): Mars is not dry. It is frozen, and most of it is underground. (flown, Caps alone ≈ 20–30 m global layer). - [A Second Ocean](https://humanitystats.com/mars/oceans): The northern lowlands are shaped like a seabed because they were one. (speculative, Needs >611 Pa and >273 K). - [Ecopoiesis](https://humanitystats.com/mars/ecopoiesis): Let something alive do the work the machines cannot afford. (tested, ~1,000 yr to breathable O₂). - [Regolith to Soil](https://humanitystats.com/mars/soil): Mars has dirt. Dirt is not soil, and Martian dirt is poisonous. (tested, Perchlorate at 0.5–1 wt%). - [Food](https://humanitystats.com/mars/food): Resupply takes two years. The greenhouse is not optional. (tested, ~40–60 m² of crop per person). - [Habitats](https://humanitystats.com/mars/habitats): The best house on Mars is a cave that already exists. (designed, ~3 m of regolith halves the dose). - [Power](https://humanitystats.com/mars/power): Everything on this list is a power problem wearing a costume. (tested, KRUSTY: 1–10 kWe, tested). - [ISRU](https://humanitystats.com/mars/isru): Do not bring the return trip with you. Manufacture it there. (flown, ~1,000 t of propellant per return). - [0.38 g](https://humanitystats.com/mars/gravity): The one variable nobody can test without going. (speculative, 3.72 m/s² — 38% of Earth). - [Getting There](https://humanitystats.com/mars/getting-there): The planets line up every 26 months, and not a day sooner. (flown, Window every 25.6 months). - [Paraterraforming](https://humanitystats.com/mars/paraterraforming): Cover the parts you need instead of converting a world. (designed, The only near-term-feasible path). ## Solar System & the Kardashev scale - [Solar System](https://humanitystats.com/solar-system): owned, data-driven Sun, planets and orbits. - [Kardashev scale progression](https://humanitystats.com/kardashev-scale-progression): humanity's place on the scale of civilizations ranked by energy use. - [Type I — planetary](https://humanitystats.com/kardashev-scale-progression/type-1): a civilization harnessing all the energy of its home planet. - [Type II — stellar](https://humanitystats.com/kardashev-scale-progression/type-2): harnessing the full output of its star (Dyson sphere). - [Type III — galactic](https://humanitystats.com/kardashev-scale-progression/type-3): harnessing the energy of an entire galaxy. - [Type IV — universal](https://humanitystats.com/kardashev-scale-progression/type-4): energy on the scale of the observable universe, climbed rung by rung across the real large-scale structure — Local Group (https://humanitystats.com/kardashev-scale-progression/type-4.1), Virgo Supercluster (https://humanitystats.com/kardashev-scale-progression/type-4.2), Laniakea (https://humanitystats.com/kardashev-scale-progression/type-4.3) and the cosmic web (https://humanitystats.com/kardashev-scale-progression/type-4.4). - [Type V — multiversal](https://humanitystats.com/kardashev-scale-progression/type-5): a theoretical multiverse-scale civilization, climbed through the inflationary foam — our bubble wall (https://humanitystats.com/kardashev-scale-progression/type-5.1), the first crossing (https://humanitystats.com/kardashev-scale-progression/type-5.2), a cluster of universes (https://humanitystats.com/kardashev-scale-progression/type-5.3) and the foam itself (https://humanitystats.com/kardashev-scale-progression/type-5.4). ## Megastructures - [Megastructures](https://humanitystats.com/mega-structures): the engineering a civilization builds as it climbs the Kardashev scale, each rendered in interactive 3D. - [Dyson Sphere](https://humanitystats.com/mega-structures/dyson-sphere): A shell of collectors that swallows a whole star. - [Dyson Swarm](https://humanitystats.com/mega-structures/dyson-swarm): Millions of mirrors, each orbiting the star on its own. - [Urban Drone Shield](https://humanitystats.com/mega-structures/urban-drone-shield): A swarm of drones that shades a city from a killer heatwave. - [Space Elevator](https://humanitystats.com/mega-structures/space-elevator): A cable to orbit — no rockets required. - [Orbital Ring](https://humanitystats.com/mega-structures/orbital-ring): A hard ring the whole planet lives on. - [Planetary Collector Shells](https://humanitystats.com/mega-structures/planetary-collector-shells): The planet wrapped in its own captured sunlight. - [Orbital Satellite Network](https://humanitystats.com/mega-structures/orbital-satellite-network): The first ring of machines a civilization hangs in the sky. - [Orbital Solar Belts](https://humanitystats.com/mega-structures/orbital-solar-belts): Belts of collectors reclaiming the sunlight a planet wastes. - [Artificial Star](https://humanitystats.com/mega-structures/artificial-star): A man-made star, caged and parked in orbit. - [Solar Relay](https://humanitystats.com/mega-structures/solar-relay): A sun-facing collector that beams starlight home. - [Galactic Energy Network](https://humanitystats.com/mega-structures/galactic-network): A hundred billion stars wired as one power grid. ## Cosmic objects (nebulae, stars, remnants, black holes, galaxies) - [Types of stars, black holes and galaxies](https://humanitystats.com/cosmic-objects): the full cosmic zoo — 34 objects, each rendered in interactive 3D, with a life-cycle map and a side-by-side comparison of mass, radius, temperature and lifespan. - [Molecular cloud](https://humanitystats.com/cosmic-objects/molecular-cloud): the cold dark nebula every star condenses out of, visible only by the starlight it blocks. - [Emission nebula](https://humanitystats.com/cosmic-objects/emission-nebula): an H II region glowing because the young stars inside have ionized it, and eroding itself in the process. - [Protoplanetary disk](https://humanitystats.com/cosmic-objects/protoplanetary-disk): the disc of gas and dust around a newborn star, with the gaps carved by forming planets. - [Main-sequence star](https://humanitystats.com/cosmic-objects/main-sequence-star): the stable hydrogen-fusing phase that ~90% of stars, including the Sun, are in. - [Red dwarf](https://humanitystats.com/cosmic-objects/red-dwarf): the most common star in the galaxy, able to burn for trillions of years. - [Red giant](https://humanitystats.com/cosmic-objects/red-giant): what a Sun-like star becomes when core hydrogen runs out. - [Blue supergiant](https://humanitystats.com/cosmic-objects/blue-supergiant): the brightest and shortest-lived stars, destined for supernova. - [Wolf-Rayet star](https://humanitystats.com/cosmic-objects/wolf-rayet-star): the hottest stars known, stripping their own atmospheres away at thousands of km/s. - [Brown dwarf](https://humanitystats.com/cosmic-objects/brown-dwarf): too heavy to be a planet, too light to fuse hydrogen. - [Planetary nebula](https://humanitystats.com/cosmic-objects/planetary-nebula): the shell a dying Sun-like star exhales, lit by its exposed core. - [White dwarf](https://humanitystats.com/cosmic-objects/white-dwarf): a stellar core crushed to Earth size, held up by electron degeneracy pressure. - [Black dwarf](https://humanitystats.com/cosmic-objects/black-dwarf): the cold end state of a white dwarf; the universe is far too young for one to exist. - [Neutron star](https://humanitystats.com/cosmic-objects/neutron-star): more than a Sun's mass in a sphere ~22 km across — the densest matter outside a black hole. - [Pulsar](https://humanitystats.com/cosmic-objects/pulsar): a rotating neutron star whose beams sweep past Earth, some 700 times a second. - [Magnetar](https://humanitystats.com/cosmic-objects/magnetar): the strongest magnets known, up to 10^15 gauss, source of giant flares and fast radio bursts. - [Quark star](https://humanitystats.com/cosmic-objects/quark-star): a hypothetical remnant denser than a neutron star, where neutrons have dissolved into free quarks. - [Supernova remnant](https://humanitystats.com/cosmic-objects/supernova-remnant): the expanding shell of an exploded star, and where the heavy elements come from. - [Stellar-mass black hole](https://humanitystats.com/cosmic-objects/stellar-black-hole): 3 to 150 solar masses, ray-traced with real gravitational lensing. - [Intermediate-mass black hole](https://humanitystats.com/cosmic-objects/intermediate-black-hole): the missing link between stellar and supermassive, confirmed by GW190521. - [Supermassive black hole](https://humanitystats.com/cosmic-objects/supermassive-black-hole): Sagittarius A* and M87*, the engines at the centre of galaxies. - [Quasar](https://humanitystats.com/cosmic-objects/quasar): a feeding supermassive black hole outshining its entire galaxy. - [Blazar](https://humanitystats.com/cosmic-objects/blazar): the same object with its relativistic jet aimed at Earth, rendered down the jet axis. - [Spinning (Kerr) black hole](https://humanitystats.com/cosmic-objects/kerr-black-hole): frame dragging, the ergosphere, and why a spinning hole's shadow is not a circle. - [Binary black hole](https://humanitystats.com/cosmic-objects/binary-black-hole): two holes lensing each other on the way to a merger — the objects LIGO hears. - [Primordial black hole](https://humanitystats.com/cosmic-objects/primordial-black-hole): hypothetical relics of the first second of the universe, and a dark matter candidate. - [Wormhole](https://humanitystats.com/cosmic-objects/wormhole): a traversable throat rendered from the exact Ellis-Bronnikov geodesics, showing a second universe through it. - [White hole](https://humanitystats.com/cosmic-objects/white-hole): the time reversal of a black hole — identical geometry, opposite horizon, almost certainly not real. - [Boson star](https://humanitystats.com/cosmic-objects/boson-star): a horizonless compact object that lenses light without casting a shadow, and a dark matter candidate. - [Spiral galaxy](https://humanitystats.com/cosmic-objects/spiral-galaxy): why spiral arms are density waves rather than structures, and so survive billions of rotations. - [Barred spiral galaxy](https://humanitystats.com/cosmic-objects/barred-spiral-galaxy): the Milky Way's own shape, and what the central bar does to the gas. - [Elliptical galaxy](https://humanitystats.com/cosmic-objects/elliptical-galaxy): red, dead and enormous — the merger remnants that host the largest black holes. - [Irregular galaxy](https://humanitystats.com/cosmic-objects/irregular-galaxy): unstructured, gas-rich and forming stars fast, like the Magellanic Clouds. - [Globular cluster](https://humanitystats.com/cosmic-objects/globular-cluster): up to a million stars in a sphere, older than the galaxies that host them. - [Cosmic web](https://humanitystats.com/cosmic-objects/cosmic-web): the largest structure in the universe — filaments of galaxies around vast empty voids. ## Artificial intelligence The structured record of how AI is actually being built: every notable model, the data centers and GPU clusters training them, the chips inside, the companies funding them and the benchmarks they are measured on. Around 5,100 pages, each generated from the underlying records rather than written by hand. - [AI progress](https://humanitystats.com/ai): the hub — training compute over time, who is building what, and where it physically sits. - [Compute map](https://humanitystats.com/ai/map): every located AI data center and GPU cluster on one world map, sized by compute. - [GPU cluster map](https://humanitystats.com/ai/gpu-clusters-map): every known large GPU cluster and AI supercomputer plotted at its published coordinates, sized by compute, with totals by country and by owner. - [AI data center map](https://humanitystats.com/ai/data-centers-map): every tracked AI data center with a known address, geocoded and sized by compute, with power, capital cost and per-country totals. - [Methodology](https://humanitystats.com/ai/methodology): where each number comes from, how it is normalized, and what its uncertainty is. ### Datasets - [AI models](https://humanitystats.com/ai/models): Training compute, parameters, dataset size, hardware and accessibility for the machine learning systems that define the field, from the 1950s to today. - [AI data centers](https://humanitystats.com/ai/data-centers): Compute capacity, electrical power, capital cost and build-out timeline for the biggest AI training and inference sites in the world. - [GPU clusters](https://humanitystats.com/ai/gpu-clusters): Every known large GPU cluster and AI supercomputer, with its chip count, compute throughput, power draw and the date it came online. - [AI hardware](https://humanitystats.com/ai/hardware): GPUs, TPUs and custom accelerators with their throughput at each precision, memory bandwidth, power draw, process node and price. - [AI chips](https://humanitystats.com/ai/chips): The chip types tracked through the AI supply chain, with their compute in H100-equivalents, memory, power and approximate unit cost. - [AI companies](https://humanitystats.com/ai/companies): Revenue, valuation, funding, headcount and compute spend for the labs and platforms at the centre of the industry. - [Benchmarks](https://humanitystats.com/ai/benchmarks): Every evaluation tracked, from saturated classics to the frontier tests models still fail — plus the Epoch Capabilities Index that ties them together. ### Rankings - [All rankings](https://humanitystats.com/ai/rankings): the leaderboards, each built from the datasets above. - [The largest AI data centers in the world](https://humanitystats.com/ai/rankings/largest-ai-data-centers): Ranked by installed AI compute, measured in H100-equivalents. - [AI data centers ranked by electrical power](https://humanitystats.com/ai/rankings/most-powerful-ai-data-centers): Ranked by electrical power draw — the constraint that now decides where AI can be built at all. - [The most expensive AI data centers ever built](https://humanitystats.com/ai/rankings/most-expensive-ai-data-centers): Ranked by total capital cost in 2025 dollars — compute, construction and everything around it. - [The AI models trained with the most compute](https://humanitystats.com/ai/rankings/most-training-compute): Ranked by the raw compute spent on the training run, in floating-point operations. - [The most expensive AI models ever trained](https://humanitystats.com/ai/rankings/most-expensive-ai-models): Ranked by the estimated cost of the compute consumed by the training run, in 2023 dollars. - [The largest GPU clusters and AI supercomputers](https://humanitystats.com/ai/rankings/largest-gpu-clusters): Ranked by compute in H100-equivalents, across every known cluster and AI supercomputer. - [The fastest AI accelerators](https://humanitystats.com/ai/rankings/fastest-ai-chips): Ranked by tensor FP16/BF16 throughput, the precision most training actually runs at. - [The most power-efficient AI accelerators](https://humanitystats.com/ai/rankings/most-efficient-ai-chips): Ranked by compute delivered per watt — what decides how much AI a megawatt of grid capacity actually buys. - [The biggest AI companies by revenue](https://humanitystats.com/ai/rankings/biggest-ai-companies-by-revenue): Ranked by most recently reported annualized revenue. - [The most valuable AI companies](https://humanitystats.com/ai/rankings/most-valuable-ai-companies): Ranked by the valuation set at the most recent funding round. - [The benchmarks with the most models scored](https://humanitystats.com/ai/rankings/best-ai-benchmark-scores): Ranked by how many model versions have been scored — a proxy for which evaluations the field actually treats as load-bearing. ### Cross-cuts - [Organizations](https://humanitystats.com/ai/organizations): Every lab, company and university with a tracked model, ranked by the largest training run it has run. - [Countries](https://humanitystats.com/ai/countries): Where models are trained and where the compute physically sits, ranked by the largest training run produced. - [Comparisons](https://humanitystats.com/ai/compare): head-to-head pages putting any two models, chips or data centers side by side. Data source: the AI datasets are built on [Epoch AI](https://epoch.ai)'s public research data, used under CC BY. Figures are Epoch's; the normalization, aggregation and prose are ours, and the methodology page states which is which. ## Humanity's progress - [Steps](https://humanitystats.com/steps): milestones on the path of civilization. - [Path to AGI](https://humanitystats.com/steps/agi): tracking progress toward artificial general intelligence. - [Fermi paradox](https://humanitystats.com/fermi-paradox): if the galaxy should be full of civilizations, where is everybody — the Drake equation made interactive, with the Great Filter. - [Humanity Engine](https://humanitystats.com/humanity-engine): models of how humanity moves and develops. - [World Flows](https://humanitystats.com/humanity-engine/economy-flow): real flows — submarine cables, flights, shipping and population. - [Trend diffusion](https://humanitystats.com/humanity-engine/trend-diffusion): how trends spread across the world. - [Encounters](https://humanitystats.com/humanity-engine/encounters): real gatherings located on the globe, then entered — the crowd resolved into interacting people. - [Green entrepreneurs ranking](https://humanitystats.com/rankings/entrepreneurs-who-have-contributed-most-to-the-reduction-of-potential-co2-in-the-atmosphere): people who have most reduced potential atmospheric CO2. ## The cellular view (crowds, person by person) - [Cellular view](https://humanitystats.com/humanity-engine/cellular-view): what actually happens inside a crowd, simulated person by person and grounded in published crowd, proxemics and alcohol research. - [A nightclub](https://humanitystats.com/humanity-engine/cellular-view/nightclub): Loud enough to destroy speech at arm's length, wet enough to remove the reason not to try. - [A street crossing](https://humanitystats.com/humanity-engine/cellular-view/street-crossing): Denser than any dancefloor, and almost nothing forms — because nobody stops. - [A queue](https://humanitystats.com/humanity-engine/cellular-view/queue): Sober, stationary strangers for hours. The most connective room here, by a distance. - [A metro carriage](https://humanitystats.com/humanity-engine/cellular-view/metro-carriage): The tightest routine packing of humans anywhere, and the least social space on Earth. - [A beer tent](https://humanitystats.com/humanity-engine/cellular-view/beer-tent): The highest alcohol throughput of any venue here, and quiet enough to still hear a stranger. - [A festival crowd](https://humanitystats.com/humanity-engine/cellular-view/festival-crowd): Extreme density, and everyone facing the same direction — away from each other. - [A warehouse rave](https://humanitystats.com/humanity-engine/cellular-view/warehouse-rave): Louder than a club and packed tighter. Whatever forms here forms at touching distance. - [A pilgrimage](https://humanitystats.com/humanity-engine/cellular-view/pilgrimage): The densest sober crowds on Earth. Bonds here form on shared purpose alone. - [A conference](https://humanitystats.com/humanity-engine/cellular-view/conference): The sparsest and quietest room here — and nearly every collision is deliberate. - [A street carnival](https://humanitystats.com/humanity-engine/cellular-view/street-carnival): A moving crowd. Encounters are constant and brief, because nobody occupies the same metre twice. - [A march](https://humanitystats.com/humanity-engine/cellular-view/protest-march): A crowd that already agrees, moving together. Shared purpose raises contact; the column keeps everyone walking. - [A stadium concourse](https://humanitystats.com/humanity-engine/cellular-view/stadium-concourse): Fifteen minutes of half-time churn. Wet, loud and busy, and everyone is trying to get somewhere. - [A covered market](https://humanitystats.com/humanity-engine/cellular-view/covered-market): Thousands of brief, functional exchanges an hour, and almost none of it survives the aisle. ## Optional - [Changelog](https://humanitystats.com/changelog): what's new on Humanity Stats. - [Credits](https://humanitystats.com/credits): data sources and acknowledgements.