Musk, Chess.com and the 10^120 Question: When a Billionaire Declares Chess Has Been Solved

When the stadium is empty, the true value of people begins to speak.

In 2026, I sat in a small VTC studio, microphone in hand, commentating on a match between two players, Davis and Hendley. The board had only 64 squares. But it took me nearly thirty more years to understand that the number 64 said nothing about the game's complexity. What said everything was another number, so long that algorithms had to write it in exponent form: 10^120.

On August 14, 2026, Elon Musk posted a line on X. It was short, but enough to spark a debate I had followed for nearly forty years in various forms. Musk argued that chess would eventually be fully solved by AI. Chess.com's official account replied. The exchange lasted a few rounds, drew tens of millions of views, and within 48 hours became a topic across both the chess and tech communities.

Chess.com responded that according to combinatorial theory, the number of possible games exceeds the number of atoms in the observable universe. Musk countered with a different argument: the number of legal positions is only about 10^44, and a future super-AI could compress information in ways humans have never imagined. He also called the Chess.com account operator an "intern." Chess.com replied in Vietnamese: "It's down to skill."

That is the surface. Beneath the surface is a longer story, and I want to tell it with data, because data never forgets what the crowd has forgotten.

Context: from Deep Blue to AlphaZero, and the gap no one has filled

In May 2026, IBM's Deep Blue defeated Garry Kasparov in a six-game rematch in New York. It was the first time a machine beat the reigning world champion in an official match with standard time controls. The world called it the start of the machine chess era. But there is a detail few reports of the time emphasized: Deep Blue did not understand chess. It evaluated roughly 200 million positions per second using specialized hardware and won through search speed. Its architecture was brute force. There was nothing called "solving."

Twenty years later, in December 2026, DeepMind's AlphaZero appeared. The difference was ontological: AlphaZero was not trained on human game data. It played itself for four hours, knowing only the rules, and reached a level surpassing Stockfish — the strongest engine at the time. AlphaZero's neural-network evaluation of positions changed how humans understand chess structure. Moves considered heresy for centuries suddenly became reasonable.

But there is one thing AlphaZero did not do: it did not prove chess was solved. It was merely stronger than every human opponent and every other engine. That is the gap Musk and Chess.com are talking about — two different things using the same word.

In 2026, Musk wrote that chess was "too simple for real life," because there is no fog of war, no technology tree, and the starting state is symmetric. I read that line many times. To a chess professional, this reasoning sounds strange. Symmetry does not make chess simpler. It makes it harder, because any small deviation compounds into an accumulating advantage over time. Fog of war hides information in other games, but chess does not need to hide anything to create complexity. It creates complexity through branching.

I once believed in emotion, until a number knocked at 3 a.m.

What that number is, I will present later.

The core: two numbers, two definitions, and a long-running confusion

Chess has two complexity metrics often conflated in public debates.

The first is the space of legal positions. According to widely cited calculations in combinatorial theory literature, the total number of legal chess positions falls around 10^44. This is the number Musk uses. Each such number represents a board configuration with pieces in specific positions — indifferent to the order of moves that led there, only to the final position.

The second is game-tree complexity, also called the Shannon number, named after Claude Shannon — the founder of information theory and the first to compute chess complexity in his 2026 paper. This number falls around 10^120, the number of distinct move sequences possible from the first move to the end. This is the number Chess.com uses.

Both numbers are correct. But they answer two different questions, and the answer to "can chess be solved" depends on which question you are asking.

A game is considered "solved" when the optimal result from every legal position can be determined, assuming both sides play perfectly. This definition has concrete consequences. For Go, no one has ever claimed to solve it. For checkers, people have — and this is the window into understanding the issue.

In 2026, Jonathan Schaeffer's team at the University of Alberta announced they had fully solved international checkers through retrograde analysis. It took over eighteen years, hundreds of thousands of distributed CPU hours, and the final conclusion was that perfect checkers is a draw if both sides play optimally. The position space of checkers is about 10^20. Its game-tree complexity is about 10^31. That is the limit computing technology at the start of the 21st century could handle.

Chess has 10^44 legal positions and 10^120 game-tree complexity. This is not a linear gap. It is the gap between an ant and the width of a galaxy.

So what is right about Musk's argument? He is right that the definition of "solved" depends on computing power, and that power is growing exponentially. He is right that a future super-AI could have information-compression methods we cannot yet imagine — this does not violate physics.

But it is a claim about the future, not a technical argument. It rests on a belief that technology will overcome every combinatorial barrier, a belief that the history of computer science itself has never provided concrete evidence for.

One more fact is needed to understand why this exchange exploded at this particular time. In early 2026, an AI-versus-AI chess tournament was held, in which xAI's Grok 4 — Musk's own product — lost to OpenAI's o3. The result was widely covered by tech media. I followed that result and noted a detail: after that defeat, Musk's public comments about chess changed. Previously he said chess was simple. Afterward he shifted to saying chess would be solved. Two different arguments, but the same direction: downplaying the game's complexity.

I am not saying this to infer motive. I am only placing facts side by side, so readers can see the pattern themselves.

A press is not noise. It is a question that data is whispering.

The other side: when a number cannot explain a life

If you read only the numbers above, the story seems simple: Musk is technically wrong, Chess.com is technically right, done.

But there are three things that stop me there.

First, Chess.com's argument — though technically correct — also uses a number strategically. 10^120 is not just a scientific fact; it is a branding tool. Chess.com is the world's largest chess platform, with tens of millions of regular users. Every time they respond to a figure with over 200 million followers on X, they are not just defending chess. They are drawing attention to their platform. I have tracked chess platform traffic for years. Every public debate like this pulls a wave of new registrations over one to two weeks. That is not bad. But it means: the debate about chess's mathematical nature is simultaneously a marketing campaign.

Second, Musk's information-compression argument cannot be refuted with existing data. Claude Shannon in 2026 calculated chess complexity at 10^120 assuming an average branching factor of about 30 moves per turn and an average game of about 40 turns. That assumption is reasonable but not absolute truth. If a super-AI learns to eliminate branches that never appear in optimal outcomes — something neither combinatorial theory nor computational physics has proven impossible — the effective number could be much smaller. The problem is no one knows how much. And that is exactly why Musk's argument cannot be defeated with current reasoning.

Third, and this is the point I most want to emphasize: if chess is solved, what actually happens?

A solved game does not disappear. Checkers has been solved for nineteen years. There are still professional checkers players. There is still a world checkers championship. There is still prize money. The number of checkers players did not collapse after 2026. It continued to exist, because the value of a game does not come from whether the optimal result is known in advance. It comes from the gap between one player and another — a gap every chess engine has known about for a long time but cannot erase.

This leads me to a paradox I have never seen clearly stated in public debates: if chess were solved, it would become the first game in history where solving made it more attractive to the public, not less.

There are players the world forgot, but data never forgets them.

And a game declared solved is the same. Those who play it will not be forgotten. They simply play in a different frame of reference.

The contrarian angle: the other side of the board

I have spent many years tracking data on online chess platforms. There is a metric few notice: the share of games ending by human decision — resignation or timeout — rather than checkmate from exploiting a clear blunder.

About 90% of games at every level follow this script. Players are not defeated by algorithms. They are defeated by their own fatigue, stress, and wrong decisions in the moment. This is not a matter of skill.

Even at the super-grandmaster level, the share of games ending in clear mistakes under psychological instability remains significant. This means: even if chess is theoretically solved, humans still play chess with a portion of intelligence that is not in any optimal lookup table. They play with emotion, with guesswork, with style, with gamble.

Musk is right on a very small point: if solving chess is the technical goal of a super-AI, then chess will be solved. But if you reverse the question — whether solving chess makes the game less valuable to humans — every piece of historical evidence says the opposite.

This is where I find the Musk-Chess.com exchange carries a deeper meaning than the surface. Both sides stand on opposite ends of a board that one side left long ago.

Musk stands on the technical side: if everything can be compressed, then chess will eventually be compressed. This is a valid view on the future of computation.

Chess.com stands on the human side: if chess is compressed into a lookup table, the game millions play every day does not change. The lookup table does not play chess with you. It only stands outside the board, looking in.

Both are correct within their scope. But only one side is talking about something observable right now.

I once believed in emotion, until a number knocked at 3 a.m. But I also once believed in data, until I saw a grandmaster cry after a loss in a tournament with no prize money. Both times, I learned something.

The transfer market does not buy the past. It buys what data has forgiven. For chess, data forgave human players long ago. What now needs forgiving is the belief that a number can define a game. It cannot.

Blind spots, and what needs to be seen with the naked eye

The debate between Musk and Chess.com reflects a reality the chess industry increasingly faces: chess is increasingly viewed as a benchmark for AI. People do not talk about games or players. They talk about which model is stronger than which.

This has upsides. Chess's reputation as a test of intelligence retains value. AI companies invest in chess, organize tournaments, sponsor events. The money flow is not bad.

But the other side exists too. If the public starts to believe chess is a simple game, already solved, with nothing left to explore, then those considering sponsoring professional chess will make different decisions. Sponsors do not want to attach their brand to something considered obsolete. Even when that is not true, perception creates reality.

This is why Chess.com's fast, sharp response matters more than a social-media exchange appears to. They are not just defending a number. They are defending a space where tens of thousands make a living, where hundreds of millions find joy.

But I also note: responding with numbers has its limits. If the chess community only answers with 10^120, outsiders will see one side reading data and one side claiming the future. The debate has no conclusion. And a debate with no conclusion is often remembered by which side spoke more eloquently.

A more sustainable answer does not lie in the number. It lies in the people playing. Children learning chess for the first time. Older adults turning to chess to keep their minds sharp. Players in club-level tournaments where prize money does not cover train fare, yet the games are still played with full effort.

When the stadium is empty, the true value of people begins to speak.

I wrote this for football years ago, in an article about the pandemic when stadiums had no spectators. It applies to chess in a different way. There is no audience watching lookup tables. There is only an audience watching people sitting before a board, making decisions in uncertainty, and bearing responsibility for those decisions.

Looking ahead: signals for the next rounds

There is one fact I consider most significant in this whole story, and it is not in the exchange between Musk and Chess.com.

AI-versus-AI tournaments are beginning to form as a commercial category of their own. This is new. Previously, chess was used to compare machine to human. Now, chess is used to compare machine to machine. The result no longer matters to chess players. But it matters to AI companies, because each win is a pricing signal for their product.

In the coming years, more AI chess tournaments are likely. This creates a new revenue stream for the platforms hosting them. And it also creates a risk: the public keeps growing used to seeing chess as a technical test, rather than a human intellectual sport.

If that happens, elite chess tournaments will need to invest more in telling players' stories. In building characters. In making viewers understand why a move by a 25-year-old from a small town in India, or by a 40-year-old from Norway, is different. That is something no lookup table can do. That is the only competitive advantage human chess still holds, and it cannot be compressed.

As for the question Musk posed — whether chess will be solved — I leave it open. Not because I lack a technical answer. But because the technical answer is no longer the most important question.

The more important question is: when a game reaches the limits of its exploration, what will humans do with it? The history of checkers after 2026 gives one answer. Nineteen years after it was fully solved, checkers is still played. There are still winners, losers, people crying, people laughing after every game. The only thing that changed is that checkers grandmasters had to play differently to equalize both sides' chances.

Chess may follow that path. Or it may not. But whichever path it takes, the board still has 64 squares. And there will still be people sitting across from each other, looking at the board, and making decisions no number can predict.

I light a candle for data. But I always let the flame of emotion light the way for the question.