
Comics that ask "what if?" | Randall Munroe
What this covers
This TED talk is Randall Munroe, the web cartoonist behind the "what if?" series, working through reader questions with math, physics, and deadpan humor. It has two main parts. The longer one starts from a question about how much data Google stores. It then follows his attempt to answer it from outside, even though Google is secretive and only its employees know the real figures. A shorter, more cinematic part imagines a baseball pitched at 90 percent of the speed of light.
The Google section is a worked example of Fermi estimation, which he compares to a sudoku puzzle: no advanced math, just constraints that narrow the feasible answers. He builds a model with unknowns such as total storage, number of servers, and drives per server. He bounds each unknown with a different source: capital spending, power contracts, and the square footage of each data center. Google's consumption of hard drives comes in as well. Less formal clues include recruitment messages, stray cell phone photos of hardware, and pizza delivery drivers. Ratios from data centers with partial information fill in the ones with less. His estimate is about 15 exabytes. He makes the scale concrete by imagining it on punch cards stacked across New England, and he also finds that the NSA looks smaller than Google.
The baseball sequence is a nanosecond-by-nanosecond account. The air has no time to move aside, so collisions fragment the ball and set off thermonuclear fusion. An expanding plasma cloud then reaches home plate and vaporizes the batter, the catcher, and the backstop before the batter sees any warning. Munroe ends with how Google responded to his article. Its official answer was "No comment," but a package of puzzle-encoded punch cards arrived about two weeks later. He also admits that some questions lie beyond what mathematics can answer.
Randall Munroe demonstrates how mathematics and estimation techniques can answer seemingly absurd hypothetical questions by breaking complex unknowns into measurable components and using available data to constrain solutions.
- Fermi estimation uses multiple independent data sources (spending, power consumption, physical space) to triangulate unknown quantities like Google's data storage
- Mathematical modeling can turn unanswerable questions into tractable problems by identifying measurable proxies and proportional relationships
- Some questions lie outside the domain of mathematical analysis entirely, revealing the boundaries of quantitative reasoning
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
When partial information is available for some data centers (e.g., both spending and power contracted with local government), the ratios between these quantities can be used to infer missing information for data centers where only one data point is known.
“For some data centers, you might get two of these pieces of information. You know how much they spent, and they also, say, because they had to contract with the local government to get the power provided, you might know what they made a deal to buy, so you know how much power it takes. Then you can look at the ratios of those numbers, and figure out for a data center where you don't have that information, you could figure out well, maybe the power is proportional.”
Fermi estimation of Google's data infrastructure can be constrained using spending data, as large data centers require substantial capital investment and the total capital expenditure limits the number of centers that could be built.
“Google has to reveal how much they spend, in general, and that lets you put some caps on how many data centers could they be building, because a big data center costs a certain amount of money.”
At approximately 30 nanoseconds into the event, the home plate is far enough away that light from the collision has not yet reached it, meaning the batter still sees the pitcher about to throw with no warning of the impending disaster.
“Now at this point, about 30 nanoseconds in, the home plate is far enough away that light hasn't had time to reach it, which means the batter still sees the pitcher about to throw and has no idea that anything is wrong.”
Fermi estimation, while not requiring advanced mathematics, is fundamentally similar to solving a sudoku puzzle in that both involve using logical constraints to narrow feasible solutions.
“The math is not all that advanced, and really it's like nothing more than solving a sudoku puzzle.”
Spending data also allows estimation of Google's market share in hard drive consumption, which is sizable given that Google experiences drive failures approximately every one to two minutes and replaces them rather than attempting repair.
“And you can also then put a cap on how much of the world hard drive market are they taking up, which turns out, it's pretty sizable. I read a calculation at one point, I think Google has a drive failure about every minute or two, and they just throw out the hard drive and swap in a new one.”
Additional data sources for estimating Google's data center locations include Google recruitment messages that indicate where they employ people, unauthorized cell phone photos of data center hardware posted by visitors, and pizza delivery driver knowledge of facility locations.
“You could always look at the Google recruitment messages that they post. That gives you an idea of where they have people. Sometimes, when people visit a data center, they'll take a cell-cam photo and post it, and they aren't supposed to, but you can learn things about their hardware that way. And in fact, you can just look at pizza delivery drivers. Turns out, they know where all the Google data centers are, at least the ones that have people in them.”
Power consumption data provides a constraint on the number of servers Google operates, as servers require a minimum amount of electricity; while Google is more efficient than most companies, it still has basic power requirements.
“You can also look at power. You can look at how much electricity they need, because you need a certain amount of electricity to run the servers, and Google is more efficient than most, but they still have some basic requirements, and that lets you put a limit on the number of servers that they have.”
Square footage of data centers provides another constraint, allowing estimation of how many server racks could physically fit in known data center facilities.
“You can also look at square footage and see of the data centers that you know, how big are they? How much room is that? How many server racks could you fit in there?”
Fermi estimation of complex quantities like total data storage can be achieved by constructing a mathematical model with multiple unknown variables (total storage, number of servers, drives per server) where each variable is constrained by available data, allowing progressive narrowing of solution space.
“And you can do this same thing with a lot of different quantities, you know, with guesses about the total amount of storage, the number of servers, the number of drives per server, and in each case using what you know to come up with a model that narrows down your guesses for the things that you don't know. It's sort of circling around the number you're trying to get.”
Collisions between the relativistic baseball and air molecules would knock away nitrogen, carbon, and hydrogen from the ball, fragmenting it into tiny particles and triggering waves of thermonuclear fusion in the surrounding air.
“the collisions with these air molecules would knock away the nitrogen, carbon and hydrogen from the ball, fragmenting it off into tiny particles, and also triggering waves of thermonuclear fusion in the air around it”
After 70 nanoseconds, the expanding plasma cloud reaches home plate and begins disintegrating the bat, batter, plate, catcher, umpire, and backstop, carrying them backward through the structure as all materials vaporize.
“after 70 nanoseconds, the ball will reach home plate, or at least the cloud of expanding plasma that used to be the ball, and it will engulf the bat and the batter and the plate and the catcher and the umpire and start disintegrating them all as it also starts to carry them backward through the backstop, which also starts to disintegrate”
When a baseball is pitched at 90 percent the speed of light, the air molecules do not have time to move out of the way, causing the ball to smash directly into them rather than having air flow around it as normal.
“when an object flies through the air, the air will flow around the object, but in this case, the ball would be going so fast that the air molecules wouldn't have time to move out of the way. The ball would smash right into and through them”
The thermonuclear fusion and collisions would produce a flood of x-rays spreading outward in a bubble with exotic particles and plasma, moving slightly faster than the ball itself from the pitcher's mound.
“This would result in a flood of x-rays that would spread out in a bubble along with exotic particles, plasma inside, centered on the pitcher's mound, and that would move away from the pitcher's mound slightly faster than the ball.”
Observed from a distance, the relativistic baseball impact would appear as a bright flash of light fading over several seconds, followed by a blast wave shredding trees and houses, and eventually a mushroom cloud rising over the ruined city.
“So if you were watching this whole thing from a hill, ideally, far away, what you'd see is a bright flash of light that would fade over a few seconds, followed by a blast wave spreading out, shredding trees and houses as it moves away from the stadium, and then eventually a mushroom cloud rising up over the ruined city.”
Google is highly secretive about its operations, including the total amount of data it stores and the number of data centers it operates; only Google employees have this information.
“Google's pretty secretive about their operations, so no one really knows how much data Google has, and in fact, no one really knows how many data centers Google has, except people at Google itself.”
A punch card can store approximately 80 characters of data and can be stored at a density of about 2,000 cards per box.
“a punch card can hold about 80 characters, and you can fit about 2,000 or so cards into a box”
Munroe's primary motivation for engaging in mathematical and Fermi estimation work is not mathematical computation for its own sake, but rather the ability to use mathematical symbol manipulation to discover surprising facts about the world.
“I love calculating these kinds of things, and it's not that I love doing the math. I do a lot of math, but I don't really like math for its own sake. What I love is that it lets you take some things that you know, and just by moving symbols around on a piece of paper, find out something that you didn't know that's very surprising.”
Mathematics provides the power to answer many seemingly ridiculous questions, though not all questions are amenable to mathematical or scientific analysis.
“I have a lot of stupid questions, and I love that math gives the power to answer them sometimes. And sometimes not.”
There are some questions that mathematics cannot answer, revealing the boundaries of mathematical and scientific reasoning.
“I think there are some questions that math just cannot answer.”
The NSA is often considered a potential contender for the world's largest data holder, but using Fermi estimation methods to analyze publicly available information about NSA data centers suggests their operations are smaller than Google's.
“There's a couple of other contenders, especially everyone always thinks of the NSA. But using some of these same methods, we can look at the NSA's data centers, and figure out, you know, we don't know what's going on there, but it's pretty clear that their operation is not the size of Google's.”
Contrary to his expectations, Google sent Munroe a package containing a set of punch cards about two weeks after he published his article.
“But then a little while later I got a message, a couple weeks later, from Google, saying, hey, someone here has an envelope for you. So I go and get it, open it up, and it's punch cards.”
Google's official answer to Munroe's article estimating their data storage capacity was 'No comment.'
“which is their official answer to my article, and it said, 'No comment.'”
If Google's estimated 15 exabytes of data (10 exabytes online plus 5 exabytes offline) were stored on punch cards and stacked in boxes in New England, the resulting stack would cover the entire region to a depth of approximately 5 kilometers, which is roughly three times deeper than glaciers during the last ice age 20,000 years ago.
“putting them in, say, my home region of New England, it would cover the entire region up to a depth of a little less than five kilometers, which is about three times deeper than the glaciers during the last ice age about 20,000 years ago”
The punch cards sent by Google were Google-branded and contained a multi-layer puzzle: encoded punch card data concealing code, which contained another code, which contained equations whose solutions revealed a message.
“Google-branded punch cards. And on these punch cards, there are a bunch of holes, and I said, thank you, thank you, okay, so what's on here? So I get some software and start reading it, and scan them, and it turns out it's a puzzle. There's a bunch of code, and I get some friends to help, and we crack the code, and then inside that is another code, and then there are some equations, and then we solve those equations, and then finally out pops a message from Google”
Munroe's estimate of Google's total data storage is approximately 10 exabytes across all operations, plus an additional 5 exabytes or so of offline storage in tape drives.
“I came up with my estimate, which I felt pretty good about, that was about 10 exabytes of data across all of Google's operations, and then another maybe five exabytes or so of offline storage in tape drives”
Google's estimated 10 exabytes of online data storage represents a significantly larger amount of data than any other organization in the world currently maintains, making Google the world's largest data holder as far as is publicly known.
“This is a staggering amount of data. It's quite a bit more than any other organization in the world has, as far as we know.”
A reader asked Munroe: If all digital data in the world were stored on punch cards, how big would Google's data warehouse be? This question prompted Munroe to estimate Google's total data storage.
“But one week, a couple months ago, I got a question that was actually about Google. If all digital data in the world were stored on punch cards, how big would Google's data warehouse be?”
A reader asked Munroe: 'If people had wheels and could fly, how would we differentiate them from airplanes?', submitting this question with the subject line 'Urgent.'
“This is a question I got from a reader, an anonymous reader, and the subject line just said, 'Urgent,' and this was the entire email: 'If people had wheels and could fly, how would we differentiate them from airplanes?' Urgent.”
Under Major League Baseball rules 6.02 and 5.09, a batter struck by a pitch is eligible to take first base; therefore, in the relativistic baseball scenario, the batter would be considered hit by pitch and eligible for first base, if it still existed.
“So the Major League Baseball rules are a little bit hazy, but — under rule 6.02 and 5.09, I think that in this situation, the batter would be considered hit by pitch and would be eligible to take first base, if it still existed.”
Munroe receives hypothetical questions from website readers weekly, which he attempts to answer using math, science, and comics.
“I have a feature on my website where every week people submit hypothetical questions for me to answer, and I try to answer them using math, science and comics.”
Some readers submit serious or concerning questions to Munroe, such as asking for the scientifically fastest way to hide a body or requesting help with homework assignments.
“I get people writing in with a lot of other strange questions. I've had someone write and say, scientifically speaking, what is the best and fastest way to hide a body? Can you do this one soon? And I had someone write in, I've had people write in about, can you prove whether or not you can find love again after your heart's broken? And I've had people send in what are clearly homework questions they're trying to get me to do for them.”
Munroe attempted to ask Google directly about their data operations but was unsuccessful in getting information.
“I've tried, I've met them a few times, tried asking them, and they aren't revealing anything.”
Munroe published his analysis of Google's data storage on his website and stated he did not expect Google to respond due to the company's secrecy.
“I posted it on my website. I wrote it up. And I didn't expect to get an answer from Google, because of course they've been so secretive, they didn't answer of my questions, and so I just put it up and said, well, I guess we'll never know.”
Munroe spent one or two days researching to gather data for estimating Google's data storage.
“I went through all of this information, spent a day or two researching.”