How Were the Pyramids Built? The Theories, Rated by the Evidence

Picture a job site where a two-and-a-half ton block has to be quarried, dragged or floated into position, and set with fine precision, then repeated 2.3 million times. To finish the Great Pyramid of Giza inside the roughly 27 years of Khufu’s reign, crews would have had to place a block on average once every three minutes, in daylight and dark, for decades.

For a technical audience that is not a history lesson. It is a scheduling problem that predates the crane by thousands of years. And 2026 has been a busy year for people trying to reverse-engineer it. Two peer-reviewed studies and one heavily criticized viral claim have all landed, each answering the same question a different way: how did they actually move and stack the stone?

We checked the figures against the original studies and rated each theory by how well the evidence holds up. Some is strong. Some is a hypothesis waiting on a shovel. One viral story does not survive contact with a radar expert.

Ramps hidden inside the pyramid

The strongest new model treats the build as an optimization problem

The most convincing recent work comes from independent researcher Vicente Luis Rosell Roig, published in npj Heritage Science. Instead of a giant external ramp, which would have needed almost as much material as the pyramid and would have left obvious remains that have never been found, he argues the ramp was built into the faces of the pyramid and filled back in once the structure topped out.

His parametric 3D model runs the build block by block. It starts with as many as 16 ramps near the wide base and tapers to a single path as the pyramid narrows, each route about 3.8 meters across so loaded and empty sledges could pass. Run the numbers and the core build lands near 13.67 years, or roughly 20 to 27 years once quarrying, river transport and seasonal pauses are added. That fits Khufu’s reign.

Why it rates so well: Roig published his code and datasets openly so others can test them, and the model fits real evidence like the postholed quarry ramps at Hatnub. Egyptologist Dr Roland Enmarch of the University of Liverpool called it plausible but flagged that it still needs physical confirmation, such as scans of the structural quirks you would expect where a ramp changed direction at the corners.

Follow the water

Peer-reviewed work says Egyptian crews engineered rivers, not just rock

The busiest research thread right now comes from a French-led team under Xavier Landreau. In a 2024 PLOS ONE study, they argued the older Step Pyramid of Djoser, built around 2680 BCE, may have used a hydraulic lift: water fed into two internal shafts to raise a float carrying stone, with the nearby Gisr el-Mudir enclosure working as a check dam. Built in the same Unreal game engine that video-game studios use, their 3D model suggested the system could raise loads up to about 100 tons.

In 2026 the same group went further. A new npj Heritage Science paper reinterprets two long structures beside Snefru’s Red Pyramid at Dahshur, features assumed for a century to be transport ramps, as a pair of dams roughly 600 to 700 meters long, complete with a zigzag “duckbill” spillway of a type still used in modern flood control. A channel would have carried water close to the pyramid’s western side, hinting that water played a direct role in the build.

Rate it moderate to high, with a caveat. Water transport itself is well documented: papyri from Wadi el-Jarf record crews shipping limestone by boat during Khufu’s reign, and the Sadd el-Kafara, one of the world’s oldest large dams, sits less than 20 km away. The hydraulic-lift and dam readings are peer-reviewed but not yet dug up. The Dahshur structures have never been excavated, partly because they lie inside a military zone.

Ramps and muscle: the textbook answer, minus the myths

The workforce is well understood. The giant external ramp is not.

The version most people learned, thousands of workers dragging stone up one massive external ramp, is part right and part missing. The labor is the solid part. Worker settlements and records point to organized, paid crews, an estimated 20,000 craftsmen and laborers rotating through, not the enslaved masses of the movies.

The external ramp is the weak part. A single straight ramp tall enough to reach the summit would have needed vast quantities of material, and no clear archaeological trace of one has ever turned up. Modeling a single internal spiral ramp does not rescue it either. On the recent math, one ramp working alone would have stretched the Great Pyramid’s build toward half a century, far past Khufu’s reign.

What survives is the toolkit, not the single-ramp picture. At the Hatnub quarry, archaeologists found a real ramp flanked by post holes, where crews hauled alabaster up slopes steeper than 20 percent on rope-and-sledge systems dating to Khufu’s era. Ramps were clearly used somewhere. The open question is what shape they took on the pyramid itself, which is exactly what the internal-ramp and water models are trying to answer.

The claim to skip: an “underground city” under Giza

A viral radar story that experts have taken apart

In March 2025 an Italian-led group calling itself the Khafre Project announced, at a press conference rather than in a journal, that satellite radar had found eight spiral shafts and huge chambers more than 600 meters beneath Khafre’s pyramid, which one member described as an underground city. It spread fast across social feeds.

It has not held up. The work was never peer-reviewed. Zahi Hawass, Egypt’s former antiquities minister, called it completely false. Lawrence Conyers, a radar archaeologist at the University of Denver, said the method cannot see anywhere near that deep and that the underground-city reading is a large exaggeration. Snopes found no credible evidence, and noted that one of the most-shared cross-section images was almost certainly AI-generated.

For a publication that sells its audience honest signal over noise, this one rates at the bottom. It is a useful reminder that a slick render and the word “radar” are not the same as proof, a distinction that matters just as much when you are sizing up a vendor’s AI demo as when you are reading about pyramids.

Why a 4,500-year-old build should matter to you

The winning move then is the winning move now: narrow, sequenced, modeled

Strip away the sand and every credible theory here is describing modern construction management. The Great Pyramid worked because someone broke an impossible job into repeatable, tightly sequenced tasks and kept material flowing. That is the same discipline separating today’s production-grade robots from the ones still parked on innovation decks: the winners do one narrow thing well, run often, and slot into the sequence early.

It is also the discipline modern megaprojects lose when they rush. Britain’s HS2 climbed toward 100 billion pounds partly because, in the words of the man now running its reset, work started too soon, before the design and the sequence were ready. The pyramid builders, by contrast, look more and more like people who modeled the problem before they moved the stone.

The tools have changed more than the problem has. Researchers are cracking these ancient builds with the same kit modern firms buy: parametric 3D models, game engines, satellite and radar sensing, digital elevation data, and open, testable code. If a solo researcher can stress-test a 4,500-year-old schedule on a laptop, the case for modeling your own build before you pour gets hard to argue with.

Related Articles
Construction Robotics in 2026: 4 Workflows Where Robots Are Actually Working
Your Internal AI Cost Will Follow HS2’s £103bn Disaster